Fielding Player WonLost Records
Fielding, including pitcher fielding, accounted for
18.8% of all Player decisions across all seasons for which I calculate
Player wonlost records. An increase in the numbers of strikeouts and home runs in recent seasons has made fielding slightly less important; since 2000, fielding has accounted for only
17.6% of all Player decisions.
Fielding decisions are accumulated in seven
Components:
Component 1: Stolen Bases
Component 2: Wild Pitches and Passed Balls
Component 5: Hits vs. Outs
Component 6: Singles v. Doubles v. Triples
Component 7: Double Plays
Component 8: Baserunner Outs
Component 9: Baserunner Advancement
The table below shows how fielding decisions break down by Component and by Position over the full Retrosheet Era.
Breakdown of Fielding Decisions by Component by Position
Position 
Component 1 
Component 2 
Component 5 
Component 6 
Component 7 
Component 8 
Component 9 
Total All Position
 (excl. P, C)

Pitcher 


3.5%  0.1%  0.1%  0.1%  1.1%  4.9% 

Catcher 
2.8%  0.9%  0.7%  0.0%  0.1%  0.1%  0.2%  4.7% 

First Base 


4.3%  0.0%  0.3%  0.3%  0.9%  5.9%  6.5% 
Second Base 


9.7%  0.0%  2.3%  0.4%  1.2%  13.6%  15.0% 
Third Base 


9.6%  0.1%  0.1%  0.2%  1.3%  11.4%  12.6% 
Shortstop 


11.4%  0.0%  2.1%  0.3%  1.5%  15.3%  16.9% 
Left Field 


7.8%  2.8%   1.6%  3.2%  15.3%  17.0% 
Center Field 


7.7%  1.6%   1.4%  3.4%  14.1%  15.6% 
Right Field 


7.7%  2.2%   1.7%  3.3%  14.8%  16.3% 
Total by Component 
2.8%  0.9%  62.4%  6.9%  4.9%  6.1%  16.0% 


The numbers in the bottom row show the percentage of total fielding decisions accumulated by
Component. The numbers in the two rightmost columns show the distribution of total fielding decisions accumulated by position (the latter column excluding Pitchers and Catchers).
Component 5, which measures whether a ball is a hit or an out, given where and how it is hit, accounts for just over 60% of all Fielding decisions. Most defensive metrics based on playbyplay data – e.g.,
UZR,
PMR,
+/,
TotalZone – match up with this measure, that is, they only look at whether ballsinplay become hits or outs. Of course, by my estimate, this means that such measures miss nearly 40% of all defensive value.
The allocation of fielding decisions across the defensive positions is discussed a bit more by me in my discussion of the general
allocation by component of Player decisions. Relative fielding across positions is discussed later in this article, as is the relative fielding of center fielders vs. corner outfielders.
For most of the
Components for which Fielding Player decisions are awarded, fielders
share these decisions with their
pitchers. The exact extent to which fielders and pitchers share these decisions varies by fielding position, by
component, and by
season. The average percentage of defensive Player decisions assigned to fielders by position and component over the Retrosheet Era are shown in the next table.
Percentage of Total Defensive Decisions assigned to Fielders
Position 
Component 1 
Component 2 
Component 5 
Component 6 
Component 7 
Component 8 
Component 9 
Catcher 
48.7%  26.8%  100.0%  0.0%  10.7% 
100.0% 
100.0% 
First Base 


61.4%  11.7%  33.1% 
100.0% 
100.0% 
Second Base 


67.8%  44.0%  93.0% 
100.0% 
100.0% 
Third Base 


71.5%  46.5%  66.2% 
100.0% 
100.0% 
Shortstop 


75.7%  25.6%  91.2% 
100.0% 
100.0% 
Left Field 


62.1%  88.3% 

100.0% 
100.0% 
Center Field 


68.3%  73.4% 

100.0% 
100.0% 
Right Field 


62.9%  68.7% 

100.0% 
100.0% 
Combining the results from the above two tables, fielders' overall share of responsibility on Components 12 and 59 is as follows.
Pitcher 
100%^{*} 
Catcher 
44.1% 
First Base 
63.1% 
Second Base 
73.8% 
Third Base 
73.9% 
Shortstop 
79.5% 
Left Field 
75.0% 
Center Field 
77.3% 
Right Field 
72.9% 
^{*}Obviously, pitchers as pitchers can't really "share" decisions with pitchers as fielders. Defensive player decisions associated with plays in which the pitcher is the fielder of record are all counted as "fielding" decisions. This distinction is purely semantic.
Fielding Player WonLost Records: Best and Worst
Before getting into too much boring detail about the math underlying Fielding wonlost records, let’s look at some fielding records for majorleague players.
Here are the top and bottom 10 fielders by position for their careers as measured by net Fielding Wins (Fielding Wins minus Fielding Losses) for the seasons for which Retrosheet has playbyplay data available (1921  2018).
Pitcher
The top player on the above list,
Greg Maddux, holds the majorleague record for most Gold Gloves with 18. So that's encouraging. On the other hand, 16time Gold Glove winner
Jim Kaat does not make my top 10 list.
In fact, Jim Kaat actually scores as very slightly below average for his career, with a career fielding winning percentage of 0.499. Kaat was brilliant when he was young, leading the major leagues in net fielding wins (among pitchers) in
1962 (the year he won his first Gold Glove), and amassing 0.9 net fielding wins through age 29. His fielding slipped as he got older, however, and was mostly below 0.500 after he reached the age of 30 with an overall fielding winning percentage over this time period of 0.437.
Catcher
The numbers shown here for catchers only include traditional fielding measures:
stolen bases,
wild pitches, and catchers' ability to
field ballsinplay. These numbers do not attempt to measure playcalling ability or pitchframing.
The top six catchers listed here won 13, 3, 6, 7, 6, and 10 Gold Gloves, respectively. Perhaps the biggest surprise here is that
Johnny Bench, winner of 10 Gold Gloves only ranks 6th. A comparison between
Gary Carter and
Johnny Bench is interesting, I think, in this regard. The vast majority of catcher fielding decisions (about twothirds) is what I call
Component 1, stolen bases and caught stealings.
Johnny Bench caught 14,488.1 (regularseason) innings in his career. For his career, Bench allowed 610 stolen bases, caught 469 baserunners stealing, and picked off an additional 62 men. Gary Carter caught about 20% more innings in his career, 17,369.0. He caught wouldbe basestealers at a considerably lower rate than Bench (35% vs. 43%), but at a rate that was better than leagueaverage (32% during Carter’s career). But Gary Carter caught 62% more wouldbe basestealers than Johnny Bench (810 CS, 51 PO). Why? Because Carter faced more than twice as many basestealing attempts as Bench (2,359 vs. 1,141). Bench's arm was so good and so wellrespected that teams mostly didn't try running on him nearly as often as they ran on other catchers.
For most seasons for which I have calculated player wonlost records, the actual stolen base success rate has tended to be very close to the breakeven success rate. This means that, on average, never stealing a base is very close in net value to stealing bases at a leagueaverage success rate. The same, then, is true for catchers as well: never having anybody attempt a stolen base has roughly the same net value as throwing out baserunners at a leagueaverage rate. In other words, shutting down the opponents’ running game doesn’t really show up as that big of a positive in Johnny Bench’s fielding record – not that having the 6thbest fielding record of the past 65+ years is at all negative, of course. There could also be ancillary benefits to completely shutting down an opponents’ running game; but in my system, any such benefits aren’t necessarily going to show up directly in Bench’s fielding record, but could instead be showing up in the pitching and/or fielding records of his teammates.
First Base
Fielding Player wonlost records for first basemen do not include any attempt to estimate the ability of first basemen to reach errant throws from other infielders.
Comparing the above list to a list of Gold Glove winners shows a few misses: 9time winner
Don Mattingly and 7time winner
Bill White, among others, don’t make the list. Perhaps more surprising, 11time winner
Keith Hernandez, who is considered by many to be the finest defensive first baseman ever, is only 6th on the list. In the case of Hernandez, I think there could be a similar phenomenon to what I observed above with respect to
Johnny Bench: Hernandez was such a good fielder that opposing teams avoided testing him, e.g., bunting less often or more toward third base than expected, thereby limiting his opportunities.
Overall, however, the top 10 players generally all had reputations as good fielders during their careers with 8 of the 10 winning at least two Gold Gloves in their careers.
On the other side, the bottom 10 list includes some notoriously bad fielders, such as
Frank Thomas and
Pedro Guerrero, and several longcareer belowaverage players who make the list more for being fairly bad for a long time (
McGriff,
McCovey) than for necessarily being truly awful.
Second Base
Lou Whitaker won 3 Gold Gloves in his career. Majorleague baseball began awarding Gold Gloves in 1957, when
Nellie Fox was 29 years old and in his 8th season as a starter. Even with the late start, Fox proceeded to win 3 of the first 4 Gold Gloves at second base (including the first one in 1957 when only one Gold Glove was awarded for all of majorleague baseball). Seeing those two at the top of the list here, therefore, is encouraging and not terribly surprising. The rest of the top 10 list also consists of players with strong defensive reputations.
There are a few notable omissions, however. Probably the most significant names missing are 10time Gold Glove winner
Roberto Alomar and 8time Gold Glovers
Frank White and
Bill Mazeroski.
Frank White rates among the
top 25 players in net fielding wins among players for whom I have calculated Player wonlost records. His career record is hurt by sub.500 seasons at the beginning and end of his career. From
1975 through
1988, White amassed a Fielding wonlost record at second base of 68  62.0, 0.525 (6.4 net fielding wins).
Mazeroski rates as above average in fielding wonlost records. Mazeroski’s record is very likely understated a bit here, however, because Retrosheet has fairly spotty records (e.g., uncertainty even regarding which fielders made some outs) for several games through Mazeroski's career. If I were to judgmentally create a list of the best fielders of the Retrosheet Era, Bill Mazeroski would definitely be a strong candidate to receive a positive judgmental boost. I discuss Mazeroski a bit more below when I compare my Fielding wonlost records to other fielding measures, specifically DRA and DRS.
Roberto Alomar, on the other hand, simply isn’t that wellregarded by my system. He scores out as slightly below average overall for his career, with a fielding record of 83.0  82.7. The reason why Alomar scores out as a net negative for his career is because his fielding got distinctly worse starting around 2000 (age 32). Based purely on net wins, he scores as deserving of a Gold Glove in
1994 and close to one in several other seasons. Even before 2000, however, Alomar had several seasons where his fielding record was below 0.500. Outside of Gold Gloves, my assessment of Roberto Alomar’s fielding is actually pretty much in line with most other analysts.
The worst fielding second baseman of the Retrosheet Era by this measure is
Steve Sax, who had infamous throwing issues through much of his career. The top 10 list here includes several other players who I remember as having goodhit, nofield reputations at second base through their career, including
Jorge Orta,
Juan Samuel,
Dan Uggla, and
Todd Walker. The list also includes 4time Gold Glove winner
Craig Biggio although, like Alomar, my rating of Biggio's secondbase defense is not out of line with other statistical measures.
Third Base
Brooks Robinson won 16 Gold Gloves and was elected to the Hall of Fame in his first year of eligibility, largely on his reputation as the greatest defensive third baseman in majorleague history. My Player wonlost records agree that Brooks Robinson was the best defensive third baseman of the past 65+ years, and by a substantial margin.
The rest of the top 10 are players with strong defensive reputations. Eight of the 10 won multiple Gold Gloves. The exceptions were
Clete Boyer and
Aurelio Rodriguez, both of whom spent their fielding prime in the American League during the time when Brooks Robinson held a monopoly on Gold Gloves.
Boyer put together a 16year, 1,725game career, despite a career batting line of .242/.299/.372 (OPS+ of 86). He finally left the American League in
1967 at the age of 30 and managed to win the only Gold Glove of his career at age 32 in
1969.
Rodriguez managed to put together an even longer career than Boyer  17years, 2,017games  despite being an even worse hitter  career batting line of .237/.275/.351 (OPS+ of 76). Like Boyer, Rodriguez did win one Gold Glove. In Rodriguez's case it came in
1976, when he had the distinction of being the first American League third baseman not named Brooks Robinson to win the award in 17 years, since
Frank Malzone won the first three such awards from 1957  1959.
Shortstop
Pee Wee Reese’s career mostly predated Gold Gloves, but he probably would have won his share. The next five players on the above list won 13, 8, 2, 9, and 4 Gold Gloves, respectively. Of course,
Derek Jeter also won as many Gold Gloves as
Dave Concepcion (5). Fielding wonlost records are not the only fielding metric that shows Derek Jeter as being a bad fielding shortstop.
Left Field
Barry Bonds won 8 Gold Gloves in left field.
Greg Luzinski was a comically bad outfielder who became a fulltime DH at age 30.
Center Field
I am missing 158 games from the early part of
Joe DiMaggio's career. As those games are added, I would expect to see DiMaggio move up this list a bit. Although I have at least basic playbyplay data for every game of
Willie Mays's career, I suspect that I may also be underrating him a bit here. For many of the early years of Mays's career, some of the playbyplay data which I have used here is fragmentary and limited. In these cases, I could be inadvertently sharing credit that ought to be Mays's with some of his teammates (some early playbyplay data does not identify the player responsible for recording all batting outs) or, perhaps more likely, blaming Mays unfairly for hits allowed by his teammates (it is not uncommon for early playbyplay data to have no information regarding hits in terms of where these hits landed). Another reason for Mays's relatively low (but still top 10!) ranking is that he scored fairly poorly later in his career. Through 1968 (when Mays was 37 years old), Mays had amassed 1.6 net Fielding wins in center field, which would push him into the top 5 in the above table.
Right Field
Ichiro Suzuki,
Al Kaline, and
Roberto Clemente each won at least 10 Gold Gloves.
Jesse Barfield only won 2 Gold Gloves in his relatively short career (only 6 seasons where he qualified for a batting title) but had the best outfield arm I ever saw (Clemente was a bit before my time).
Carl Furillo’s career mostly predated the Gold Glove award, but his nickname was the “Reading Rifle” and he was good enough to play over 300 games in center field in his career, so his appearance at the top of this list is no real surprise, either.
There are, however, a few multiple Gold Glove winners missing from the above list, including
Dwight Evans (8),
Dave Winfield (7),
Larry Walker (7), and
Tony Gwynn (5), although all four of these players rate as aboveaverage rightfielders for their career.
The next several sections of this article take an analytical look at several aspects of Fielding wonlost records. The article then concludes with comparisons of Fielding wonlost records to some other sabermetric fielding measures.
Comparing Fielding WonLost Records Across Multiple Positions
For my work, I calculate
Positional Averages for players based largely on a comparison of offensive performance by position. As an alternative, however, one could try to compare defensive value across fielding positions by analyzing the performance of a single player at multiple positions.
For example, across all seasons for which I have estimated Player wonlost records, players who played both left field and center field within the same season had an average winning percentage of 0.488 in center field and 0.510 in left field. From this, one could reasonably conclude that center field is a more difficult position to play and one could also use this difference as a basis for adjusting these winning percentages to reflect a common base.
Comparisons of this type were done for all of the infield and outfield positions. Pitchers and catchers are not considered here. In the case of pitchers, this is because pitchers virtually never play a different position. This is also true, although to a lesser extent, of catchers. More problematic, however, in the case of catchers, is the fact that the skill set needed to be a good majorleague catcher isn’t really the same skill set needed to be a good fielder at any other position (the same is true to a lesser extent, of course, when comparing infielders to outfielders, and, really, is true to at least some extent in every case here).
Average Winning Percentage at Position X

1B 
2B 
3B 
SS 
LF 
CF 
RF 
1B 
 0.527  0.522  0.541  0.505  0.504  0.504 
2B 
0.491   0.492  0.497  0.486  0.487  0.485 
3B 
0.481  0.496   0.500  0.480  0.477  0.477 
SS 
0.482  0.489  0.487   0.484  0.482  0.484 
LF 
0.487  0.503  0.496  0.509   0.510  0.501 
CF 
0.484  0.492  0.490  0.498  0.488   0.490 
RF 
0.480  0.490  0.489  0.494  0.493  0.506  
This table is read as follows. For a player who played both first base and second base, the average winning percentage at first base is shown in the top row,
0.527 – this is the average winning percentage of second basemen when they are playing first base. The average winning percentage of first basemen when they are playing second base is shown in the first column,
0.491. In all cases here, average winning percentages are calculated as weighted averages where the weights used are the
harmonic mean between the player decisions at the two fielding positions being compared.
The average “normalized” winning percentage for a player at position Y when playing other positions can then be calculated as the weighted average of the numbers down the relevant column. The weights used to calculate these averages were the number of games upon which the comparison was based, which, as noted above, was the harmonic mean of the number of Player decisions accumulated at the two positions being compared.
Doing so produces the following average winning percentages by fielding position:
1B 
0.484 
2B 
0.495 
3B 
0.494 
SS 
0.500 
LF 
0.491 
CF 
0.506 
RF 
0.495 
This says that, on average, a first baseman amasses an average winning percentage of
0.484 at other positions. These numbers are only comparable, however, if we assume that the players being considered here are 0.500 fielders. Averaging across the rows, we can calculate the average winning percentage at first base of players who also played other positions: in this case,
0.512. Doing this for every position produces the following baseline winning percentages by position to which the above percentages should be compared:
1B 
0.512 
2B 
0.493 
3B 
0.491 
SS 
0.488 
LF 
0.503 
CF 
0.489 
RF 
0.496 
The first set of winning percentages was adjusted via the
Matchup Formula based on this latter set to ensure a combined winning percentage of 0.500 across all positions. These results are as follows:
1B 
0.472 
2B 
0.501 
3B 
0.504 
SS 
0.512 
LF 
0.488 
CF 
0.517 
RF 
0.499 
In words, if a set of first basemen with an average winning percentage of
0.512 amass an average winning percentage of
0.484 at other positions, then we would expect a set of first basemen with an average winning percentage of 0.500 to amass an average winning percentage of
0.472 at other positions.
Based on these winning percentages, the defensive spectrum looks something like this:
1B < LF < RF < 2B < 3B < SS < CF
Several aspects of these results are noteworthy. First, the range of winning percentages is fairly narrow, outside of first basemen. The other interesting comparison, I think, is that third base appears to be a tougher position to play than second base. Bill James discusses this in his
Win Shares book, where he discusses the historical shift of the defensive spectrum with second base becoming more important than third base over time. As Bill James puts it:
"Third basemen need quicker reactions, since they are nearer the batter, and they need a stronger arm, since they are further from first base. Without the double play, third base is obviously the more demanding position." (Win Shares, p. 183)
The results here confirm this. Second base is, in one sense, the more valuable position, with approximately 19 percent more player decisions accumulated at second base than at third base, a difference which comes almost entirely from
Component 7 (double plays). Yet, comparing how well fielders do when they play both second base and third base in the same season, third base is the more difficult position.
The final table here compares these results with relative Fielding winning percentages implied by average offensive performances by position, which I derived in
this article.
Adjusting Fielding Winning Percentage by Fielding Position
Position 
Implied by Relative Fielding 
Implied by Offensive Performance 

1B 
0.472  0.394 
2B 
0.501  0.534 
3B 
0.504  0.501 
SS 
0.512  0.547 
LF 
0.488  0.470 
CF 
0.517  0.488 
RF 
0.499  0.467 
The most striking difference between relative Fielding winning percentages implied by offensive performances and those based on comparing players who played more than one position is the former results in a much wider spread of implied fielding talent across positions. There are also several differences in the relative difficulty implied by position. Perhaps most strikingly, offensive performances by position imply that middle infielders are much better fielders than center fielders.
So, which methodology produces better results?
For my work, I have chosen to calculate my positional averages based on relative offensive performances by position. I do this for several reasons which, I believe, make this a better choice for my purposes.
First, the mathematics here, attempting to normalize winning percentages across fielding positions, is fairly murky. In contrast, simply setting the positional average equal to the average winning percentage compiled at that position seems to me to be much cleaner and more elegant mathematically.
Second, I believe that limiting the analysis only to players who have played more than one position in the same season, as is done here, may lead to issues of selection bias. That is, we are not looking at the full population of all majorleague players here – since most majorleague players never played a game at shortstop, for example – or a random sample of majorleague players. Instead, we are looking at a selected sample of majorleague players, who were selected, in part, on the basis of exactly what we’re attempting to study: with very few exceptions, the only majorleague players who are selected to play shortstop are those whose manager thought they were capable of playing a majorleague caliber shortstop (and the few exceptions likely only played an inning or two in an emergency situation, so they will be weighted very lightly in the above calculations).
I think that this is probably the primary reason why the winning percentages found here are generally closer to 0.500 than those implied by differences across offensive performances. The players considered here are selfselected for their ability to play multiple positions similarly well. Truly bad players at “offensefirst” positions – think
Frank Thomas at 1B,
Manny Ramirez in LF – are so bad that nobody would ever consider trying to play
Frank Thomas at 3B or
Manny Ramirez in CF. But, at the other end, great defensive players at “defensefirst” positions are so great defensively that, for example,
Ozzie Smith never played a single inning of majorleague baseball at any defensive position besides SS;
Willie Mays never played a corner outfield position until he was 34 years old.
Finally, I believe that setting positional averages based on actual empirical winning percentages is more consistent with what I am attempting to measure with my Player wonlost records. Player wonlost records are a measure of player
value. At the bottomline theoretical level, every team must field a player at all nine positions. If one team has a second baseman that is one win above average and another team has a left fielder who is one win above average, then these two teams will win the same number of games (all other things being equal). Hence, in some sense, not only is it a reasonable assumption to view an average second baseman as equal in value to an average left fielder, it is, in fact, a necessary assumption.
Use of Location Data in Calculating Player WonLost Records
For balls in play, there are three pieces of information that are potentially useful in determining the value of particular plays and to whom that value should be credited (or debited): (i) the first fielder to make a play on the ball, (ii) the type of hit (bunt, ground ball, fly ball, line drive), and (iii) the location of the ball. The extent to which these three pieces of information are available in
Retrosheet playbyplay data varies considerably through the years.
(i) First Fielder
The first fielder to touch the ball is the most important consideration for determining credit. The first fielder to touch the ball is identified for virtually all plays for the last 25  30 years of Retrosheet data. For earlier years, the first fielder to touch base hits is frequently unknown. As data goes back even further, there are even some outs for which the fielder of record is unknown.
For my work, the identity of the first fielder is used for assigning credit whenever this information is available. When this information is not available, credit is allocated across all fielders in the proportion in which fielders get credit across similar plays for which the fielder is known.
(ii) Hit Type
The second level of detail on balls in play is the type of hit: bunt, ground ball, fly ball (or pop up), line drive. This information is available from Retrosheet for all balls in play for the years 1989  1999 and for seasons since 2003. For other years, hit types are generally only available on outsinplay, not hits.
As with firstfielder information, hittype information is used in calculating Player wonlost records whenever this information is available. When this information is not available, credit is allocated based on the expected distribution of hit type based on the final play result.
(iii) Location
For the years 1989  1999, the
location of all ballsinplay are identified in Retrosheet's playbyplay data. I do not use this location data directly in calculating Player wonlost records, however. Instead, I use location data for these seasons to
calculate expected ex ante probabilities for ballinplay events. That is, based on 1989  1999 location data, I calculate what the probability of an out would have been on a play that ended up as, say, a line drive double to the left fielder.
After a great deal of research and consideration, I decided to use location data only in this indirect way even for those seasons for which Retrosheet provides location data (i.e., 1989  1999). I made this decision for several reasons. For one thing, using location data only indirectly leads to a more consistent methodology across all seasons for which I estimate Player wonlost records. But also, it was not clear to me, in looking at results from those years for which location data are available, that the location data actually improved the results.
Location data are fundamentally subjective, by its very nature. Relying on individual pieces of subjective data will inevitably introduce errors and possible biases into the valuation of these individual plays. Relying on location data only indirectly, however, and by relying on all of the location data  11 years' worth  in assessing every play, should allow these individual errors to balance out and offset in such a way as to vastly reduce any potential biases or errors.
Consider, for example, the impact of using
STATS data versus BIS data for calculating
UZR fielding statistics. Simply changing the data source leads to
wildly different stories about some players' defense: was
Andruw Jones the best fielder in baseball from 2003  2008 (+112 runs in UZR using BIS data) or a slightly belowaverage center fielder (5 using STATS)? If the results are that unstable across different location measurements of the same plays, then it's hard to see exactly how much information location data are bringing to the party at all.
Beyond the question of whether the actual locations being reported are accurate, however, another issue with using location data is that I think that relying too heavily on location data builds on a fundamental assumption that I am not entirely sure is true. This is that balls hit to the same location are more similar than balls that end up with the same end result. That is, a fielding system based on location data treats two fly balls to medium rightcenter field as equivalent  implicitly assuming that all fly balls to medium rightcenter field are created equally. My fielding system here treats two fly ball doubles fielded by the right fielder as equivalent  implicitly assuming that all flyball doubles fielded by the right fielder are equivalent.
I am not saying the latter of these implicit assumptions is necessarily right, so much as I wonder whether the former implicit assumption is actually more right. And if our focus is purely on player
value rather than player talent (as it is in my system), then, in fact, in many ways it makes more sense to me to view one flyball double to right field as being equal in value to any other flyball double to right field than to view a flyball double to medium rightcenter field as being equal in value to a fly out to the same location.
The Impact of Location Data on my Fielding System
The problem with evaluating fielding systems, in general, is that we don't really know what the "right" answer is  after all, if we knew the right answer, we'd just use that.
One thing that I can compare, however, is how my results compare to what they would have been had I used location data for those years for which it is available, 1989  1999. For those seasons, I calculated Player wonlost records both ways. I then calculated a
weighted correlation of winning percentages by fielding position between the two methods. I calculated correlations two ways: for overall (
contextneutral,
teammateadjusted) fielding win percentage and for
Component 5 win percentage, which is based purely on whether a ball in play becomes a hit or an out and corresponds most directly to other locationbased fielding systems. The results were as follows:
Weighted Correlation,
Fielding Winning Percentages:
Location Data v. No Location Data

Total 
Component 5 
Pitcher 
82.64% 
82.06% 
Catcher 
75.63% 
72.69% 
First Base 
86.08% 
84.18% 
Second Base 
77.07% 
71.45% 
Third Base 
89.82% 
89.00% 
Shortstop 
80.11% 
75.48% 
Left Field 
92.65% 
89.04% 
Center Field 
88.71% 
83.53% 
Right Field 
93.24% 
88.56% 
Note: Catcher figures exclude SB, WP; Totals calculated using the same pitcherfielder splits for both sets of numbers.
These correlations are extremely high, which is quite encouraging.
It occurs to me that one way in which fielding records based on location data (e.g., fly balls to medium right field) can be compared to fielding records based on event data (e.g., flyball doubles fielded by the right fielder) is by comparing the extent to which player win percentages
persist.
As I explain in the article linked in the previous sentence, I
divide credit on balls in play between pitchers and fielders based on "persistence equations" which measure the extent to which player winning percentages on evennumbered plays can be explained as a function of player winning percentage on oddnumbered plays within the same season: i.e.,
WinPct_{Even} = 0.500 + b•(WinPct_{Odd}  0.500)
The coefficient b in the Persistence Equation measures the persistence of fielding winning percentage between the two samples (even plays v. odd plays). Broadly speaking, a higher value of b in a persistence equation suggests that more of a real skill is being measured. Hence, if the values of b in persistence equations based on location data were consistently higher than the values of b in persistence equations based on event data, this could suggest that locationbased data were capturing more fielding skill than eventbased data.
The next table, then, compares the values of b from persistence equations for
Components 5,
6, and
7 for both pitchers and fielders based on locationbased and eventbased data.
Persistence Coefficients,
Location Data v. No Location Data

Component 5 
Component 6 
Component 7 

Pitcher 
Fielder 
Pitcher 
Fielder 
Pitcher 
Fielder 

Location 
Event 
Location 
Event 
Location 
Event 
Location 
Event 
Location 
Event 
Location 
Event 
Catcher 
22.88% 
23.82% 
33.10% 
6.40% 
72.07% 
2.92% 
25.20% 
9.59% 
7.14% 
16.05% 
7.87% 
14.19% 
First Base 
9.41% 
8.66% 
46.56% 
42.91% 
19.26% 
18.69% 
20.78% 
18.99% 
72.64% 
70.95% 
33.95% 
34.26% 
Second Base 
9.27% 
13.97% 
41.40% 
28.68% 
66.98% 
80.44% 
34.87% 
62.68% 
7.55% 
7.83% 
29.06% 
29.53% 
Third Base 
3.64% 
5.90% 
35.41% 
41.14% 
73.76% 
38.69% 
23.24% 
14.80% 
91.22% 
88.32% 
36.17% 
33.44% 
Shortstop 
19.03% 
12.03% 
48.99% 
31.59% 
90.35% 
90.44% 
34.77% 
24.43% 
2.92% 
2.16% 
16.64% 
15.40% 
Left Field 
14.51% 
16.20% 
37.93% 
38.94% 
4.38% 
0.09% 
40.42% 
23.96% 




Center Field 
3.19% 
12.18% 
37.15% 
39.51% 
9.88% 
11.58% 
16.94% 
23.26% 




Right Field 
15.88% 
18.55% 
39.19% 
36.81% 
12.79% 
11.01% 
36.94% 
25.07% 




There are a lot of numbers there. Let me try to walk through some of the numbers to give an idea of what they are saying. I will then provide some summary data.
Let's start with catchers. For
Component 5, which measures whether a ballinplay is converted into an out or a hit, on plays made by the catcher, the persistence in pitcher win percentages is very similar for locationbased data (22.88%) versus eventbased data (23.82%). For fielders (i.e., catchers), however, the persistence is much stronger using locationbased data (33.10%) than for eventbased data (6.40%). This is clearly a vote in support of the superiority of locationbased data.
Plays made by the third baseman, on the other hand, suggest that eventbased data are superior, with somewhat greater persistence coefficients for both pitchers (5.90% vs. 3.64% using locationbased data) and fielders (41.14% vs. 35.41%).
The next table combines the results across fielders. The numbers here are weighted averages, with the total share of fielding decisions by component used for weights (these weights differ slightly between locationbased and eventbased records; the numbers used here are the average of the two).

Component 5 
Component 6 
Component 7 

Pitcher 
Fielder 
Pitcher 
Fielder 
Pitcher 
Fielder 

Location 
Event 
Location 
Event 
Location 
Event 
Location 
Event 
Location 
Event 
Location 
Event 
All Positions 
10.94% 
12.41% 
40.92% 
36.08% 
11.76% 
8.15% 
33.24% 
24.04% 
8.12% 
8.55% 
23.90% 
23.59% 
Location minus Event 
1.47% 

4.85% 

3.61% 

9.19% 

0.43% 

0.31% 

So, on average, location data do not seem to provide any additional information in evaluating Component 5 performance for pitchers. This makes sense to me. Pitchers have some control over what happens to balls in play, but at a much more generalized level of detail  e.g., fly balls vs. ground balls, perhaps how hard a ball is hit  than by location  e.g., I'm skeptical that a pitcher can control whether a ground ball is hit in the 56 hole or more directly in the 6 hole. Hence, it makes sense to me that focusing on events rather than locations provides somewhat more information than focusing on location (although the difference here, 1.5%, is not really enough to favor either of these two over the other with any degree of certainty).
In contrast, locationbased data does appear to provide some additional information in evaluating Component 5 performance for fielders, although the difference between the two persistence coefficients, on average, is relatively small (less than 5%).
For
Component 6  whether hits are converted into singles, doubles, or triples  location data seems to provide more persistent measures for both pitchers and fielders. This makes sense as, for example, the depths of fly ball hits likely makes a big difference as to how many bases they go for. Still, the differences here are not enormous (less than 10% for both pitchers and fielders).
For
Component 7  whether ground ball outs are converted into double plays  on the other hand, location data appears to add essentially no value.
Overall, the results seem generally supportive of the idea that calculating fielding records based on events rather than locations is probably not much, if any, worse than locationbased fielding records.
I look at the best and worst fielders as measured by Fielding wonlost records later in this article. I also compare my results to those of other fielding systems.
Dividing Credit between Pitchers and Fielders
In many cases, it is not clear exactly who should get credit for a particular play. For example, pitchers and catchers share responsibility for
Component 1 (basestealing) Player decisions. The allocation of Player decisions in these cases is done based on the relative skill level apparent by the relevant players.
The technique outlined here is used to divide responsibility between pitchers and catchers for
Component 1 (basestealing) and
Component 2 (wild pitches and passed balls) Player decisions, between
pitchers and fielders for
Components 5 (hits vs. outs),
6 (single vs. double vs. triple), and
7 (double plays), and between
batters and baserunners for Components
7,
8 (baserunner outs), and
9 (baserunner advancements).
The division of Component 1 Player decisions between pitchers and catchers is used here as an illustration of the general technique.
1. Basic Theory
How does one determine how to divide credit between pitchers and catchers for Component 1 (basestealing) Player decisions?
Let’s begin by asking, what if somebody deserved no credit for a particular component of Player decisions but we allocated Player decisions to them anyway? For example, what if we assigned Component 1 Player decisions to the defensive team’s right fielder? What would we expect Component 1 Player decisions to look like in that case? Essentially, we would expect every right fielder to have a Component 1 winning percentage of 0.500 plus or minus some random variation.
Suppose we were to try to predict a right fielder’s Component 1 winning percentage over some time period based on his Component 1 winning percentage over some other time period. We would expect, in such a persistence equation, for there to be no predictive ability of this component.
Alternately, what would we expect Component 1 Player decisions to look like if we assigned them to players who had different levels of talent in terms of affecting the opponents’ basestealing? In such a case, we would expect a player’s Component 1 winning percentage to be equal to his “true” winning percentage (his “truetalent”) plus or minus some random variation and for a player’s Component 1 winning percentage over some time period to have significant predictive capacity over other time periods.
In other words, the extent to which a player’s winning percentage at some point in time is predictive of his winning percentage at some other point is suggestive of the extent to which there is a true skill involved in a particular component. Based on this, Player wins and losses are allocated in proportion to the extent to which a player’s winning percentage has predictive power.
2. Mathematics
The basis for dividing shared Player decisions is Persistence Equations. I divide the plays that took place in a particular season into two pools: odd and even. To evaluate the persistence of skills, I then fit a simple equation which attempts to explain winning percentage by component on even plays as a function of the same factor for odd plays:
(Win %)_{Even} = b•(Win %)_{Odd} + (1b)•(Win %)_{Baseline}
where (Win %)
_{Baseline} represents a baseline toward which Component winning percentage regresses over time.
The coefficient b in the persistence equation measures the persistence of Component winning percentage between the two samples (even plays v. odd plays) and, hence, the extent to which Component winning percentage is a true “skill” for the relevant set of players being evaluated.
This equation is estimated using a
Weighted Least Squares technique which weights observations by the
harmonic mean of the number of games over which the even and odd winning percentages have been compiled squared.
3. Complication: Controlling for the Talent of the Other Players Involved
Earlier, I identified a defensive team’s right fielder as an example of a player for whom we would expect his Component 1 winning percentage to simply be randomly distributed. In fact, however, some of you might have seen a flaw in my example.
In 2004, the
Montreal Expos allowed only 58 stolen bases on the season, while catching 41 opposing baserunners attempting to steal. Based on this, the Montreal Expos compiled a teamwide Component 1.1 (basestealing by runners on first base) winning percentage of
0.642. Of course, this means that Expos rightfielders would have a combined Component 1.1 winning percentage of
0.642, not 0.500, not because Expos right fielders had some innate ability to prevent the other team from stealing bases, but because they had the good fortune to be teammates with
Brian Schneider, who amassed an unadjusted Component 1.1 winning percentage of
0.660 at catcher.
On the other hand, the
2002 New York Mets allowed 151 stolen bases against only 53 caught stealing, leading to a teamwide contextneutral Component 1.1 winning percentage of
0.432, due, in part, to the notorious problems of their catcher,
Mike Piazza, who allowed 125 stolen bases (which led the National League) against 27 caught stealing in 121 games caught, for a contextneutral Component 1.1 winning percentage of
0.321.
Unfortunately, this problem with attempting to measure “truetalent” Component 1 winning percentage is not limited to outfielders, where we know that no such talent exists. In fact, on average, the contextneutral Component 1.1 winning percentage for Montreal Expos pitchers in 2004 was
0.642, not necessarily because Expos pitchers were particularly adept at holding runners on base, but, in large part, because Brian Schneider was their catcher. Yet, pitchers do have some ability here. The key is to separate the ability of Montreal Expos pitchers from the ability of Montreal Expos catchers.
The first step before one can accurately assess “truetalent” Component 1 winning percentages is to adjust player winning percentages for the context in which these percentages were amassed. Specifically, pitchers’ Component 1 winning percentages are adjusted to control for the Component 1 winning percentages of their catchers, and catchers’ Component 1 winning percentages are adjusted to control for the Component 1 winning percentages of their pitchers. Similar adjustments are done for all Components for which Player Game Points are to be shared.
This is done iteratively. First, pitchers’ Component 1 winning percentages are adjusted to control for the Component 1 winning percentages of their catchers. This is done using the
Matchup Formula.
After pitchers’ winning percentages are adjusted based on catcher winning percentages, catcher winning percentages are then adjusted based on these newlyadjusted pitcher winning percentages. Ideally, one would probably prefer to continue the iterative process until all Component 1 winning percentages do not change between iterations. For computational simplicity, I simply repeated this process three more times for both pitchers and catchers.
Returning to the earlier examples, the adjusted Component 1.1 winning percentages for Montreal Expos pitchers was
0.533 in 2004 (versus
0.642 unadjusted), while Montreal Expos catchers put up a combined adjusted Component 1.1 winning percentage of
0.643 (versus
0.642 unadjusted). Here, because Expos pitchers and catchers were both aboveaverage in this component in 2004, their combined winning percentage ends up being greater than either of their individual winning percentages. The whole is greater than the sum of the parts.
For the 2002 New York Mets, their pitchers’ adjusted winning percentage was
0.519 (versus
0.432 unadjusted) while Mets’ catchers had an adjusted winning percentage of
0.415 (0.321 for Mike Piazza and
0.709 for other Mets’ catchers). Mets pitchers weren’t bad at preventing stolen bases in 2002; they simply had the misfortune of pitching to one of the worst catchers in modern times at stopping an opponent’s running game.
The Persistence Equations by which Shared Player Wins and Losses are calculated are estimated using component winning percentages which have been adjusted in this way for the winning percentages of players’ teammates.
4. Example Persistence Equations
Persistence equations are estimated using all of the seasons for which I have estimated Player wonlost records, which model player winning percentage for the Component of interest on evennumbered plays as a function of player winning percentage for the Component of interest on oddnumbered plays:
(Component Win Pct)_{Even} = b•(Component Win Pct)_{Odd} + (1b)•(WinPct)_{Baseline}
where (WinPct)
_{Baseline} represents a baseline winning percentage toward which Component winning percentages regress over time.
The results for Component 1.1, Component 1 (basestealing) for the baserunner on first base, are shown below.
Persistence of Component 1 Winning Percentage: Baserunner on First Base
Pitchers: n = 39,108, R^{2} = 0.0483
WinPct_{Even} = (25.94%)•WinPct_{Odd} + (74.06%)•0.5000
(53.20)
Catchers: n = 8,193, R^{2} = 0.0042
WinPct_{Even} = (23.50%)•WinPct_{Odd} + (76.50%)•0.5000
(21.62)
The number n is the number of players over whom the equation was estimated, that is, who accumulated any Player wins and/or losses on both odd and evennumbered plays. The value R
^{2} measures the percentage of variation in the dependent variable (WinPct
_{Even}) explained by the equation (i.e., explained by WinPct
_{Odd}).
The baseline, toward which WinPct
_{Even} regresses  (Win %)
_{Baseline} in the persistence equation  is set equal to 0.500. This is done for all of the persistence equations which I use to allocate shared credit. I did this based on emprical experimentation with alternatives, including freely estimating (Win %)
_{Baseline}. I thought the results when (Win %)
_{Baseline} was constrained to 0.500 worked best.
The numbers in parentheses are
tstatistics. Tstatistics measure the significance of b, that is, the confidence we have that b is greater than zero. The greater the tstatistic, the more confident we are that the true value of b is greater than zero. Roughly speaking, if a tstatistic is greater than 2, then we can be at least 95% certain that the true value of b is greater than zero (assuming that certain statistical assumptions regarding our model hold).
For baserunners on first base, Component 1 win percentage is significantly persistent for both pitchers and catchers with tstatistics far greater than two for both sets of players. The persistence is somewhat weaker for catchers
(23.5%) than for pitchers
(25.9%), although the two numbers are very close. The percentage of Component 1 Player decisions with a runner on first base (Component 1.1) which are attributed to pitchers is set equal to the pitcher persistence coefficient
(25.9%) divided by the sum of the persistence coefficients for pitchers and catchers
(25.9% + 23.5%). This leads to
52.5% of Component 1.1 decisions being allocated to pitchers and
47.5% of Component 1.1 decisions allocated to catchers.
5. Changes in Component Splits over Time
There is no reason to believe that the split of credit between positions should be constant over time. On the other hand, if a distinct persistence equation is estimated every year, this could well produce significant yeartoyear shifts because of statistical quirks from small sample sizes. Ideally, what we would like to do is allow for gradual changes in component splits over time, but do so in a way that reduces the likelihood of flukish yeartoyear changes.
To accomplish this, I estimate unique Persistence Equations for every season, but I use all of my data in all of these equations. I simply weight the data based on how close to the season of interest it is. Each observation is multiplied by a Year
_{Weight}, which is equal to the following:
Year_{Weight} = 1  abs(Year  Year_{Target}) / 100
where "Year" is the year in which the observation occurred, and Year
_{Target} is the year for which shares are being estimated. So observations in the target year get a Year
_{Weight} of 1.0, observations one year before or after the target year get a Year
_{Weight} of 0.99, observations two years removed from the target year get a Year
_{Weight} of 0.98, etc.
The result is a set of share weights that vary by year but do so fairly gradually. For example, the share of credit for Component 1.1 (basestealing by runners on first base) attributed to pitchers varies by season within a range of
46.0% to
56.0%.
6. Final Proportions of Shared Player Game Points
The specific Persistence Equations used to separate shared responsibilities are summarized in my
Component writeup.
Separate persistence equations and, hence, separate share weights, are calculated for specific fielders and by specific baserunners, so that, for example,
Component 5 shares for first basemen and third basemen will differ. Also, as noted above, these share weights vary by season. Splits by season are presented on the pages for specific leagues (e.g.,
2010 National League).
Average breakdowns of shared components over the full Retrosheet Era are summarized in the table below. The numbers below are averages across all fielders/baserunners and across all seasons, so do not necessarily apply precisely for any specific players or seasons. As noted above, detailed discussions of shared credit by component can be found in a
separate article on my website.
Shared Components based on Persistence Equations
Controlling for Abilities of Teammates: example, Doug Mirabelli
In 2000,
Doug Mirabelli committed 5 passed balls in 80 games for the
San Francisco Giants, good for a (teammateunadjusted) contextneutral
Component 2 winning percentage of
0.592.
In 2003, Doug Mirabelli committed 14 passed balls in only 55 games for the
Boston Red Sox, posting a (teammateunadjusted) contextneutral Component 2 winning percentage of
0.480.
Did Doug Mirabelli really get that much worse in just three years? Well, he did age from 29 in 2000 to 32 in 2003, so some of that could be agerelated decline. But, more significantly for Mirabelli, in 2003, he was the personal catcher for knuckleballer
Tim Wakefield, who had a career (contextneutral, teammateadjusted) Component 2 winning percentage of
0.255.
In order to make Player WonLost records meaningful as measures of player talent, it is necessary to control for the ability of one’s teammates. This is done using the
Matchup Formula as just described in this article.
The case of Doug Mirabelli, sometime personal catcher for knuckleballer Tim Wakefield, is instructive in this regard.
Doug Mirabelli’s teammateunadjusted contextneutral
Component 2 wonlost records over his career are as follows:
Year 
Team 
Wins 
Losses 
Win Pct 

1996  SFN  0.02  0.04  0.330 
1997  SFN  0.00  0.00  1.000 
1998  SFN  0.01  0.01  0.560 
1999  SFN  0.07  0.01  0.871 
2000  SFN  0.18  0.12  0.592 
2001  TEX  0.06  0.06  0.496 
2001  BOS  0.13  0.17  0.436 
2002  BOS  0.11  0.11  0.491 
2003  BOS  0.10  0.11  0.480 
2004  BOS  0.10  0.18  0.359 
2005  BOS  0.06  0.07  0.479 
2006  SDN  0.02  0.02  0.561 
2006  BOS  0.13  0.20  0.387 
2007  BOS  0.08  0.12  0.411 
CAREER   1.08  1.21  0.470 
Outside of Boston over these years, Mirabelli’s Component 2 winning percentage was over 0.500 in five of seven seasons, with an overall winning percentage of
0.586. In contrast, Mirabelli’s Component 2 winning percentage was below 0.500 in six of his seven seasons in Boston, with an overall Component 2 winning percentage in Boston of
0.426. Overall, Mirabelli rates as a fairly poor catcher at preventing wild pitches and passed balls, with an overall Component 2 winning percentage of
0.470.
When Mirabelli’s Component 2 wonlost record is adjusted to control for the pitchers who Mirabelli caught, however, the results are the following:
Year 
Team 
Wins 
Losses 
Win Pct 

1996  SFN  0.02  0.04  0.320 
1997  SFN  0.00  0.00  1.014 
1998  SFN  0.02  0.01  0.565 
1999  SFN  0.07  0.01  0.852 
2000  SFN  0.18  0.13  0.585 
2001  TEX  0.06  0.06  0.492 
2001  BOS  0.14  0.16  0.470 
2002  BOS  0.11  0.10  0.529 
2003  BOS  0.11  0.10  0.525 
2004  BOS  0.13  0.16  0.457 
2005  BOS  0.08  0.05  0.608 
2006  SDN  0.02  0.02  0.531 
2006  BOS  0.15  0.18  0.451 
2007  BOS  0.10  0.10  0.503 
CAREER   1.18  1.11  0.516 
Adjusting for the pitchers he caught, Doug Mirabelli turns out to have been slightly above average at preventing wild pitches and passed balls through his career. Outside of Boston over these years, Mirabelli’s Component 2 winning percentage remains fairly consistent after adjusting for his teammates, at
0.576. With Boston, on the other hand, Mirabelli’s combined Component 2 winning percentage improves dramatically from
0.426 unadjusted to
0.493 adjusted.
In words, adjusting for Mirabelli’s teammates brings his Component 2 winning percentages closer together over time. Mathematically, the standard deviation of Mirabelli’s winning percentages falls from
0.110 unadjusted – i.e., Mirabelli’s Component 2 winning percentages fell mostly in a range of
0.470
+/ 0.110 (0.360  0.580) – to
0.090 adjusted – i.e., Mirabelli’s Component 2 winning percentages range from
0.426 to 0.606 (0.516
+/ 0.090).
Mirabelli was still a bit worse in Boston than elsewhere. Of course, outside of one month in 2006 in San Diego, his career outside of Boston came at ages 25 – 30, while his Boston career was from ages 30 – 36. So based on age alone, we would have expected him to probably be a little less agile at blocking wouldbe wild pitches in Boston than in San Francisco and Texas.
It seems clear to me that the latter set of numbers more accurately reflect Doug Mirabelli’s ability to prevent wild pitches and passed balls.
Center Fielders versus Corner Outfielders
Over the Retrosheet Era, total Fielding Decisions (Player wins plus Player losses) for each of the three outfield positions were as follows:

Component 5 
Component 6 
Component 8 
Component 9 
Total Fielding 

Left Field 
14,684  5,350  2,955  6,031  29,021 
Center Field 
14,589  3,004  2,644  6,459  26,695 
Right Field 
14,480  4,071  3,134  6,231  27,915 
At first glance, this looks a little curious. Why do left fielders and right fielders accumulate more fielding decisions than center fielders? What does this mean, exactly? Is a good defensive right fielder more valuable than a good defensive center fielder? Arguably. Should teams play their best defensive outfielder in right field rather than center field? Probably not.
The reason for this apparent anomaly is not because corner outfielders are better, or even necessarily more valuable, than center fielders. Rather, this is the result of two issues that are worth thinking about with respect to Player WonLost records. First, there is a wider range of fielding talent across corner outfielders than across center fielders, and, second (and somewhat related), there is a greater range of possible outcomes on balls hit to left or right field than on balls hit to center field.
The table below shows the number of plays in the 2007 American League (i.e., games played at AL ballparks) for which the various outfielders are the fielder of record (i.e., are the first fielder to touch the ball):

Total Plays
 Total Outs
 Singles^{*}
 Doubles
 Triples
 % Outs
 % XBH

Left Field
 10,252
 4,611
 3,965
 1,618
 58
 45.0%
 29.8%

Center Field
 11,667
 5,996
 4,421
 1,060
 190
 51.4%
 22.1%

Right Field
 9,609
 4,527
 3,629
 1,278
 175
 47.1%
 28.7%

^{*}“Singles” include batters reaching on error.
For simplicity, suppose that singles have a net fielding win value of 0.0364 and extrabase hits have a net fielding win value of 0.0625 (these are reasonably close to the average net win values for these plays in recent seasons).
^{*} Let’s also normalize the above numbers to be per 100 plays.
^{*}Base hits will likely not have the exact same value to all fields because of differences in baserunner advancement. The numbers here should therefore be thought of as illustrative, not definitive.

Total Plays
 Outs
 Singles
 ExtraBase Hits

Left Field
 100.00
 44.98
 38.68
 16.35

Center Field
 100.00
 51.39
 37.89
 10.71

Right Field
 100.00
 47.11
 37.77
 15.12

So, for example, left fielders allow 38.68 singles per 100 plays. At 0.0364 wins per play that works out to 1.41 wins for left fielders on singles allowed. Full numbers are shown in the table below.
Net Wins on: 
Singles
 ExtraBase Hits
 Total Losses
 Total Wins
 Wins per Out

Left Field
 1.4078
 1.0218
 2.4295
 2.4295
 0.0540

Center Field
 1.3793
 0.6696
 2.0489
 2.0489
 0.0399

Right Field
 1.3747
 0.9451
 2.3198
 2.3198
 0.0492

Let me walk through the numbers briefly. As noted above, left fielders allow 38.68 singles per 100 plays with a value of 0.0364 wins (0.0364 losses) per single, for a total of 1.4078 wins (1.4078 losses). Left fielders allow 16.35 extrabase hits per 100 plays with a value of 0.0625 wins per extrabase hit, for a total of 1.0218 wins on extrabase hits. Adding these together, left fielders accumulate approximately 2.43 losses per 100 plays. Since fielding wins and losses are set to be equal in the aggregate for every position by construction, this means that left fielders also accumulate 2.43 wins per 100 plays, which works out to 0.0540 wins per out by the left fielder.
Note what this shows. Plays made by the left fielder are worth more player decisions on average – 0.049 decisions per play
^{*} – than plays made by right fielders – 0.046 – than plays made by center fielders – 0.041. This is true for two reasons. First, center fielders allow fewer extrabase hits than corner outfielders – 10.7 per 100 plays vs. 15.8 per 100 plays for corner outfielders, and extrabase hits have the highest value in terms of total player decisions per play. Second, center fielders allow fewer hits than corner outfielders – 48.6 per 100 plays vs. 54.0 per 100 plays – which makes outs to center field less valuable – because they’re more common – than outs to the corners. The overall result is that an average play made by a left fielder is worth about 19% more player decisions than an average play made by a center fielder, which is more than enough difference to offset the fact that center fielders were involved in 14% more plays than left fielders (in the 2007 American League).
^{*}2.4295 wins plus 2.4295 losses equals 4.86 total decisions per 100 plays, or 0.0486 decisions per play.
The primary reason for this, I believe, is that there is a much wider range in the abilities of corner outfielders as compared to center fielders. Mathematically, this can be measured by looking at the standard deviation of winning percentages by corner outfielders. Over the Retrosheet Era, the
standard deviation of seasonlevel winning percentages for center fielders (fielding only) is
4.5%, versus
5.1% for right fielders and
5.2% for left fielders. In other words, the spread in winning percentages for corner outfielders (which can be taken as an approximation of the spread in the fielding talent of corner outfielders) is approximately
13% greater than the spread in centerfielder winning percentages (fielding talent).
In words, virtually all center fielders are good fielders, whereas, while some corner outfielders are excellent fielders (e.g.,
Ichiro Suzuki), others are notoriously bad fielders (e.g.,
Manny Ramirez). The result is that the value of a corner outfielder who is capable of converting balls in play into outs and in preventing extrabase hits is greater than the value of a center fielder that can do the same, because such a corner outfielder is rarer. Curious, but I think it’s true.
Fielding Player WonLost Records vs. Other Fielding Measures
The final two sections of this article compare Fielding Player wonlost records to some other sabermetric fielding measures.
Fielding Player WonLost Records vs. DRA and DRS
To the best of my knowledge, there are two other fielding systems which rely on largely the same data source as I do (
Retrosheet playbyplay data) and have publicly presented career fielding records. The first of these is Defensive Runs Saved (DRS), which were originally presented by Sean Smith in a
Hardball Times article and are available now online at
BaseballReference. The second system is Defensive Run Average (DRA), which was created by Michael Humphreys, who explained the system in his wonderful book,
Wizardry: Baseball’s AllTime Greatest Fielders Revealed.
In his book, Humphreys presents career fielding numbers (measured in net runs) for all players who played a significant time (typically, more than 3,000 innings) at six defensive positions: second base, third base, shortstop, and each of the three outfield positions. At the time of publication, Humphrey’s book included statistics through 2009.
BaseballReference presents DRS values for players from 1953 to the present.
To compare my results to Smith’s and Humphrey’s numbers, therefore, I compared career results for all of the players listed by Humphrey whose career started in 1953 or later. I then excluded any data after 2009. This left a total of 958 players, ranging from 152 left fielders to 168 second basemen.
Raw Results
The first table summarizes the results for DRA (Humphreys), DRS (Smith), and (
contextneutral, teammateadjusted) Net Fielding Wins (eWins minus eLosses).


Net Fielding Runs (per 1000 innings) 
Net Fielding eWins (per 1000 innings) 


DRA 
DRS 


Position 
No. of Players 
Mean 
Std. Dev. 
Mean 
Std. Dev. 
Mean 
Std. Dev. 
2B 
168  0.48  6.15  0.36  4.78  0.022  0.197 
3B 
155  0.36  5.59  0.29  5.42  0.013  0.235 
SS 
163  0.21  5.75  0.22  5.70  0.000  0.244 
LF 
152  0.28  6.87  0.17  4.95  0.033  0.335 
CF 
167  0.27  6.37  0.36  5.51  0.031  0.263 
RF 
153  0.41  6.10  0.30  5.73  0.011  0.311 
The first thing that we have to do before we can compare DRA and DRS to Player wins is to put them on the same scale. DRA and DRS are expressed in runs while Player wins are, of course, expressed in wins. Traditionally, in sabermetric measures, one win is equivalent to approximately 10 runs. Looking at the standard deviations in the above table, however, the ratio of DRA/DRS to Player wins is closer to 20. In other words, even if you converted DRA and DRS to wins, using a conventional runtowin translation, the spread of players' DRA and DRS is roughly double the spread of players' net fielding wins.
Why is the Spread on Player Fielding Wins lower than Defensive Runs?
I believe that the spread on my net fielding wins is less than the spread of other fielding measures because I assign more credit on ballsinplay to pitchers, whereas standalone fielding measures implicitly assign all of the credit on ballsinplay to fielders, since that's all that they are measuring.
Specifically, looking at my
Components 4 (excluding home runs),
5,
6,
7,
8, and
9, I assign 54.5% of the (defensive) credit for these to pitchers and only 45.5% of the (defensive) credit to fielders.
Putting Things on the Same Scale
In order to really compare DRA, DRS, and what I'll start calling NFW (net fielding wins), it is necessary to put them all on the same scale. To do this, I created
"zscores" associated with all three statistics. The basic formula for a zscore of variable x is (x  m) / s, where m is the mean of the statistic and s is the standard deviation. I calculated zscores for each player for all three fielding stats using a value of m equal to zero (since all three of these statistics are constructed to be relative to league average by construction) and the standard deviations from the above table.
For example,
Al Cowens scores at 2.16 DRA (per 1000 innings in RF), 1.22 DRS, and 0.236 NFW. From the previous table, the standard deviations associated with these three numbers
are 6.10, 5.73, and 0.311, respectively. This translates, therefore, into zscores for Al Cowens in right field of 0.35 for DRA, 0.21 for DRS, and 0.76 for NFW.
I did this for every player referenced in the earlier table. I then calculated simple
correlations between DRA, DRS, and NFW by position.
Position 
DRA v. DRS 
DRA v. NFW 
DRS v. NFW 

2B 
0.741 
0.715 
0.788 
3B 
0.776 
0.686 
0.859 
SS 
0.763 
0.766 
0.829 
LF 
0.689 
0.700 
0.776 
CF 
0.749 
0.636 
0.729 
RF 
0.763 
0.648 
0.773 
I'm not always exactly sure how to interpret correlations. If we thought that one of the other two measures (DRA or DRS) was a very bad measure of fielding, for example, then we probably would prefer a fairly low correlation. On the other hand, if we thought that one of the other two measures was a perfect measure of fielding, then we could view NFW's correlation with it as a measure of how close to perfect Player wonlost records are at measuring fielding.
Of course, neither of these hypotheticals are true. DRA and DRS are both quite good, but nevertheless imperfect, measures of fielding.
Given that, the correlations here, which are all very high, strike me as very good. I'm probably not doing something terribly wrong here and, perhaps, I'm even doing something a little more right than some other people.
My Player wonlost records (NFW) correlate more strongly with DRS (Sean Smith's numbers, as found at
BaseballReference.com) than with DRA (Michael Humphreys' numbers from his book
Wizardry). This makes sense, since both DRS and NFW are constructed play by play, whereas DRA data are calculated (rather well) at a seasonal level.
The next section of this article looks more closely at how DRA, DRS, and NFW compare on a positionbyposition basis.
PositionbyPosition Analysis
Second Base
For the 168 players evaluated here, the average difference in zscores between DRA and NFW (DRA minus NFW) is 0.035. The average absolute difference in zscores between DRA and NFW is 0.596. For DRS, the corresponding numbers are 0.039 and 0.493.
There are a total of nine players for whom the difference in zscores is greater than one (in absolute value) for both DRA vs. NFW and DRS vs. NFW. These players are shown in the next table.
The first player in this table,
Bill Mazeroski, is a good candidate for a somewhat closer look.
Bill Mazeroski
Bill Mazeroski was elected to the Baseball Hall of Fame in 2001 on the basis of two things: hitting a World Series winning home run in
1960 and being considered by many to be the best defensive second baseman in majorleague history.
Michael Humphreys ranks Mazeroski as the secondbest defensive second baseman in MLB history and his career DRA record rates as a zscore of 1.322. He scores even better in DRS, with a career zscore of 1.688. In contrast, his Player fielding record, while not bad, is much more pedestrian, with a career zscore of only 0.316. The next table compares Mazeroski's seasonbyseason zscores for DRA (Humphreys), DRS (Smith), and NFW (Thress).


Fielding ZScore 
Season 
Innings 
DRA (Humphreys) 
DRS (Smith) 
NFW (Thress) 
1956 
671 
1.697 
0.623 
1.002 
1957 
1,238.1 
1.182 
1.351 
1.559 
1958 
1,344 
3.147 
3.580 
0.961 
1959 
1,167.1 
0.000 
0.358 
1.154 
1960 
1,341.2 
0.364 
0.780 
0.218 
1961 
1,333 
1.464 
1.255 
0.882 
1962 
1,410.1 
1.499 
1.038 
0.064 
1963 
1,229 
3.970 
3.915 
2.411 
1964 
1,438.2 
1.922 
1.890 
1.516 
1965 
1,131.2 
1.437 
3.327 
2.999 
1966 
1,452.1 
1.456 
1.296 
1.582 
1967 
1,437.1 
1.132 
1.164 
1.157 
1968 
1,255.2 
1.166 
2.499 
1.445 
1969 
545.1 
0.597 
1.151 
0.168 
1970 
888.1 
2.197 
2.355 
1.596 
1971 
345.1 
1.413 
1.211 
2.377 
1972 
107 
0.000 
1.955 
5.802 
Retrosheet has released playbyplay data for every game since 1937, i.e., for every game of Bill Mazeroski's career. For some games in the 1950s and 1960s (and earlier), however, these playbyplay data were deduced from box score and newspaper accounts. In these cases, the information available is very rudimentary, including, in many cases, a lack of specificity on outs. That is, there are many plays for which it is known that the batter made an out, but there is no information on which fielder(s) recorded the out. The
Pirates are one team for whom this data is particularly sparse in some of these seasons. The result of this is that, in many cases, I do not know which fielder recorded certain outs for the Pirates. When this happens, I spread the fielding credit for these plays in proportion to leaguewide out distributions for known plays. This probably results in me undercrediting good fielders and overcrediting bad fielders on a team for plays made (and overdebiting good fielders and underdebiting bad fielders for hits allowed).
Retrosheet's playbyplay data generally gets more reliable over time. And, in fact, as the above table indicates, my view of Bill Mazeroski's fielding (a) improves and (b) becomes much more consistent with DRA and DRS in the latter part of his career. From 1956  1962, Mazeroski's zscores are 1.339 for DRA, 1.254 for DRS, and 0.496 for Net Fielding wins. From 1963  1972, the zscores are 1.324, 2.043, and 1.019, respectively. From 1965  1972, the three zscores are 0.749, 1.752, and 1.289.
In this case, DRA and DRS may be better measures of Bill Mazeroski's career fielding record, or at least the first half of it.
Third Base
For the 155 players evaluated here, the average difference in zscores between DRA and NFW (DRA minus NFW) is 0.007. The average absolute difference in zscores between DRA and NFW is 0.594. For DRS, the corresponding numbers are 0.004 and 0.413.
There are a total of 4 players for whom the difference in zscores is greater than one (in absolute value) for both DRA vs. NFW and DRS vs. NFW. These players are shown in the next table.
Former Dodger
Jim "Junior" Gilliam appears on each of the previous two lists. Both DRA and DRS view Gilliam as an average fielder at both second and third base. Player wonlost records, on the other hand, view Gilliam as an excellent fielder at both positions. Player wonlost records also rate Gilliam as having been
excellent in his (more limited) time in left field. DRS actually agrees that Gilliam was an excellent left fielder (+21 runs in 203 innings), while Humphreys did not report Gilliam's DRA for LF because he had too few innings.
Gilliam was before my time, retiring two years before I was born, so I will leave it to others to judge whether Gilliam was an average or excellent defensive infielder.
Shortstop
For the 163 players evaluated here, the average difference in zscores between DRA and NFW (DRA minus NFW) is 0.036. The average absolute difference in zscores between DRA and NFW is 0.553. For DRS, the corresponding numbers are 0.037 and 0.474.
There are a total of 6 players for whom the difference in zscores is greater than one (in absolute value) for both DRA vs. NFW and DRS vs. NFW. These players are shown in the next table.
Left Field
For the 152 players evaluated here, the average difference in zscores between DRA and NFW (DRA minus NFW) is 0.057. The average absolute difference in zscores between DRA and NFW is 0.619. For DRS, the corresponding numbers are 0.064 and 0.533.
There are a total of 8 players for whom the difference in zscores is greater than one (in absolute value) for both DRA vs. NFW and DRS vs. NFW. These players are shown in the next table.
Decomposition of Fielding Value
Most fielding measures focus primarily on what is almost certainly the most significant aspect of fielding: how well a fielder turns balls in play into outs. It is my understanding that this is the primary focus of both DRA and DRS. Both DRA and DRS do, however, attempt to incorporate infielders' ability to turn double plays and outfielders' ability to throw out runners and/or prevent baserunner advancement.
Player wonlost records assign Fielding wonlost records within five
components.
Component 5 measures whether balls in play become hits or outs and is, therefore, perhaps most directly comparable to other fielding systems.
Component 6 measures whether hits become singles, doubles, or triples. To the best of my knowledge, no other fielding system attempts to measure anything comparable to this.
Component 7 measures whether ground balls are converted into double plays in double play situations (runner on first, less than two outs). I believe that both Humphreys (DRA) and Smith (DRS) make some attempt to incorporate similar information within their systems.
Component 8 measures whether fielders are able to put baserunners out on the bases.
Component 9 measures the extent to which baserunners are able to advance more or less than average on a particular play. Many fielding systems (including both DRA and DRS, I believe) make at least some effort to incorporate these latter two factors for outfielders. My system goes a step farther, however, and calculates Component 8 and 9 player wonlost records for infielders as well.
Some of the differences, then, between how Player wonlost records view some players' fielding visavis DRA and DRS (and other systems) is that Player wonlost records are incorporating additional aspects of these players' fielding skills.
For example, of the eight players on the above list, only two of them 
Gates Brown and
Pete Rose  would appear on a comparable list comparing Net
Component 5 Fielding wins to DRA and DRS.
Curiously, though, Net Component 5 Fielding Wins are actually less strongly correlated to DRA and DRS than total Net Fielding Wins for outfielders and while only two of the eight players above differ by more than one zscore in Net Component 5 Wins from both DRA and DRS, there are four other players who also differ by at least one zscore in Component 5 Wins but do not differ by as much when total Net Fielding Wins are considered. This is likely because, while DRA and DRS do not (in my opinion) model all of the other aspects of fielding as accurately as Player wonlost records, they nevertheless do capture some of these aspects, and do so reasonably well. Still, I do believe that this is an example of how the imperfect correlations between Player wonlost records and other fielding systems are indicative that Player wonlost records are doing a better job of measuring many aspects of fielding.
Right Field
For the 153 players evaluated here, the average difference in zscores between DRA and NFW (DRA minus NFW) is 0.102. The average absolute difference in zscores between DRA and NFW is 0.661. For DRS, the corresponding numbers are 0.018 and 0.544.
There are a total of 10 players for whom the difference in zscores is greater than one (in absolute value) for both DRA vs. NFW and DRS vs. NFW. These players are shown in the next table.


Fielding ZScore 
Player 
Innings 
DRA (Humphreys) 
DRS (Smith) 
NFW (Thress) 
Bobby Murcer 
7,106 
1.429 
1.598 
0.130 
Dante Bichette 
7,264 
0.135 
0.481 
1.624 
Jay Johnstone 
4,195 
0.976 
0.583 
0.613 
Jerry Morales 
4,501 
1.019 
1.009 
0.094 
Jim King 
5,344 
0.981 
0.196 
0.909 
Jose Cruz, Sr. 
4,076 
0.643 
0.771 
0.913 
Michael Tucker 
5,513 
0.297 
0.095 
1.336 
Ollie Brown 
7,130 
0.046 
0.294 
1.135 
Ron Fairly 
5,094 
0.675 
0.480 
0.858 
Ron Swoboda 
3,955 
0.083 
0.265 
1.108 
Center Field
For the 167 players evaluated here, the average difference in zscores between DRA and NFW (DRA minus NFW) is 0.076. The average absolute difference in zscores between DRA and NFW is 0.666. For DRS, the corresponding numbers are 0.053 and 0.573.
There are a total of 16 players for whom the difference in zscores is greater than one (in absolute value) for both DRA vs. NFW and DRS vs. NFW. This is the most players for any of the six positions compared here. These players are shown in the next table.


Fielding ZScore 
Player 
Innings 
DRA (Humphreys) 
DRS (Smith) 
NFW (Thress) 
Amos Otis 
15,715 
0.948 
0.393 
1.092 
Bill Virdon 
12,816 
0.588 
0.652 
0.564 
Brian L. Hunter 
5,548 
0.792 
0.655 
2.404 
Chet Lemon 
12,425 
1.376 
1.403 
0.078 
Daryl Boston 
3,470 
0.407 
0.314 
1.612 
Don Demeter 
4,355 
0.829 
0.209 
1.042 
Gary Maddox 
13,736 
1.279 
1.296 
0.111 
Herm Winningham 
3,496 
0.718 
0.572 
0.644 
Jerry Martin 
3,076 
1.632 
1.713 
0.416 
Jimmie Hall 
3,572 
0.307 
0.610 
1.886 
Johnny Grubb 
3,376 
1.022 
1.883 
0.376 
Larry Hisle 
4,140 
1.061 
1.404 
0.284 
Mitch Webster 
3,318 
1.277 
0.712 
0.341 
Rich Becker 
3,751 
0.251 
0.097 
1.111 
Tommie Agee 
7,777 
1.372 
1.355 
0.148 
Willie Wilson 
10,721 
0.717 
0.491 
0.686 
Two center fielders in the above table perhaps warrant some further discussion:
Amos Otis and
Garry Maddox.
Amos Otis
I rate
Amos Otis much more highly than either Humphreys or Smith. In my original version of this comparison, the difference was even more stark. At that time, Amos Otis ranked
first in career net fielding wins in center field among all players for whom I had calculated Player wonlost records. I revised my Player wonlost records somewhat this past spring and Amos Otis does not look quite so good (his zscore here fell from 1.838 to 1.092). But Humphreys' and Smith's systems, on the other hand, think that Amos Otis was a belowaverage defensive centerfielder over the course of his career, so Otis still shows up on the above table.
Humphreys quotes Bill James calling Otis "a 'magnificent' fielding center fielder", and he did win 3 Gold Gloves in his career (in 1971, 1973, and 1974). But everybody can surely think of at least one fielder who won a Gold Glove award or two that he didn't deserve. And anyway, there's a pretty large gap between "3time Gold Glover" and "best fielder of the past 65 years".
In my original article, I expressed "doubt" that Amos Otis was really the "best centerfielder of the past 65 years" and I am actually somewhat relieved that my revised results agree that he was not quite that good (although he's still top 10). That said, my revised results still think much more highly of the "magnificent" fielding of 3time Gold Glove winner Amos Otis than DRA and DRS.
I think that one reason why my system loves Amos Otis's defense so much has to do with his home ballpark in Kansas City. The next table shows
team ballpark factors for the
Kansas City Royals in the 1970s (the seasons when Amos Otis was their everyday centerfielder). Numbers here are expressed in relation to the batting team with 100 being average, so, for example, a Doubles factor of 102 would mean that doubles are 2% more common in Royals games than in the AL in general (because of the ballparks, not the players).
Season 
Runs 
Doubles 
Triples 
Homers 
1970  98.7  102.1  104.4  93.7 
1971  100.0  104.9  106.1  93.8 
1972  99.7  98.4  103.6  92.8 
1973  101.3  105.6  103.7  99.9 
1974  101.0  106.2  106.7  96.2 
1975  100.7  105.4  105.7  95.9 
1976  98.7  103.4  99.0  97.4 
1977  101.7  106.2  106.8  93.8 
1978  101.5  105.4  107.4  95.4 
1979  101.4  103.3  107.0  100.8 
The numbers bounce around a bit from year to year but, in general, Kansas City's ballpark boosted runscoring by boosting doubles and triples while suppressing home runs. The result is a higherthanaverage number of balls in play in Kansas City with a higherthanaverage number of these balls falling in for hits in general, and for extrabase hits in particular.
Because hitsinplay were more plentiful in Kansas City, the value of outs on ballsinplay there were greater than average. By measuring value using
ballparkspecific win probabilities, my Player wonlost records (fielding, batting, baserunning, and pitching) implicitly adjust for ballpark context. So, my Player wonlost records like Amos Otis's defense better because he played in a ballpark where outfield hits were more common, making it a more difficult ballpark to play centerfield.
I think that this is a real advantage of my fielding (and batting, baserunning, and pitching) wonlost records.
Garry Maddox
One of the more troubling results I encountered when I was first evaluating my Player wonlost records was the fielding record of
Garry Maddox. Garry Maddox won 8 consecutive Gold Gloves from 1975 through 1982 and was considered the gold standard of centerfield defense.
My Player wonlost records, on the other hand, show Garry Maddox to have been an average defensive centerfielder over the course of his career.
Now, as anyone familiar with Gold Gloves knows, they are not necessarily the best measure of fielding prowess  far from it in many cases. And similarly, one of the lessons of modern fielding metrics is that looks can frequently be deceiving when it comes to judging majorleague fielding ability.
But DRA and DRS both agree with the consensus of the time: Garry Maddox was a great centerfielder. He led his league in DRS among centerfielders 4 times (1976, 197880) and finished second 3 other times (1975, 1977, 1981). For his career, his zscore in DRA is 1.279 and for DRS it's 1.296. But for Fielding Player wonlost records, his net fielding wins earn a zscore of 0.111.
This concerned me: it seemed like an obvious mistake on the part of my Player wonlost records. But then I read Michael Humphreys' entry on Garry Maddox in his book:
"[Maddox] was at best an average fielder when he came up with the Giants. Traded to the Phillies, he played next to possibly the worst outfielder of all time: Greg "The Bull" Luzinski. On almost all teams, the centerfielder takes all chances in the outfield that he can, including soft flies that could be handled in the gaps by the corner outfielders. But with The Bull, Maddox may have taken what would normally be fly ball chances of the left fielder. Maddux had only one good season when he wasn't playing next to Luzinski, the strikeshortened 1981." (Wizardry, p. 302)
Here's how Garry Maddox's record looks in the three measures I'm comparing here season by season. The seasons where Maddox was not teamed with Luzinski are bolded.


Fielding ZScore 
Season (Team) 
Innings 
DRA (Humphreys) 
DRS (Smith) 
NFW (Thress) 
1972 
810.1 
1.162 
0.000 
0.223 
1973 
1,236 
1.396 
0.294 
0.921 
1974 
1,124.1 
0.140 
0.485 
1.959 
1975 (SFN) 
122.2 
2.558 
5.923 
7.186 
1975 (PHI) 
840.1 
4.294 
1.729 
1.564 
1976 
1,240 
2.531 
2.930 
1.256 
1977 
1,204.2 
2.475 
1.960 
0.050 
1978 
1,324.1 
2.251 
2.606 
0.920 
1979 
1,194.1 
3.021 
3.954 
0.067 
1980 
1,246.2 
1.384 
1.749 
0.743 
1981 
750 
2.720 
1.453 
2.821 
1982 
903 
0.174 
0.402 
0.239 
1983 
736.1 
0.213 
1.727 
0.958 
1984 
496.2 
1.264 
0.731 
0.268 
1985 
492.1 
0.319 
0.738 
0.460 
1986 
14 
22.41 
25.95 
19.95 
Total (w/ Luzinski) 
7,050.1 
2.559 
2.525 
0.100 
Total (w/o Luzinski) 
6,685.2 
0.047 
0.054 
0.123 
My numbers for Maddox are much more stable with and without
Greg Luzinski as a teammate. But does that mean that my numbers are the ones that are right?
One possible problem that my system might be having with the Luzinski/Maddox outfields could be if Maddox ended up tracking down a fair number of hits that were Luzinski's fault. Under my system, one key (perhaps "the key") defining characteristic of ballsinplay is the first fielder to touch the ball. Specifically, certain assumptions about the probability that a play could have been turned into an out and by whom are calculated based on who the first fielder was to touch the ball. So, for example, if a double is fielded by the center fielder, the system assigns more "blame" for that double to the center fielder than to the adjoining fielders.
If Garry Maddox routinely ran down hits that were Greg Luzinski's fault, my system might be underrating Maddox (in
Components 5 and
6) and offsettingly overrating Greg Luzinski. While this seems plausible to me, in fact, I basically agree with Humphreys' and Smith's assessment of Luzinski's fielding. For his career in left field, he gets zscores of 2.192 in DRA, 1.860 in DRS, and 2.348 from me. If anything, I am scoring Luzinski a bit more harshly than Humphreys and Smith.
Let me pick out one season. I don't claim this season is representative, it's just the first one that I looked at. According to Player wonlost records, the
1979 Phillies accumulated a total of approximately 1.2 net fielding wins overall and the Phillies outfield accumulated approximately 0.6 net fielding wins. This ranked them
6th in the National League that year in net fielding wins. According to
BaseballReference.com, on the other hand, the Phillies led the National League in Defensive Runs Saved with +54, with their starting outfield scoring a combined +24 (+26 by Maddox, +18 by
Bake McBride, and 20 by Luzinski).
At the team level, we should be able to get a pretty good sense of how good a team's defense is by looking at the team's Defensive Efficiency Rating (DER, the percentage of ballsinplay turned into outs). According to
BaseballReference, the Phillies ranked 6th in the NL in DER in 1979 at 0.703 vs. a leaguewide value of 0.700. Those numbers are perfectly in line with my assessment of Phillies' team fielding. And while it's a much worse measure of just fielding, it might also be worth noting that the Phillies ranked 9th in the 12team NL in runs allowed per game (4.40 vs. leagueavg of 4.22).
I'm reluctant to shout this result from the rafters and claim that I have established definitively that Garry Maddox was wildly overrated as a defensive centerfielder. But he might have been.
Infielders versus Outfielders
There were a total of 16 infielders (counting
Junior Gilliam twice) whose zscores differed by more than 1.0 when comparing my NFW to both DRA and DRS, about 5 per position and 3.3% of the total infielders that I compared. For outfielders, the number is 34 players, 11 per position and 7.2% of all outfielders that I compared.
Conclusions
Overall, I'm quite pleased with the results here. The overall correlation, across all six positions (958 players) investigated here, between Humphreys' DRA and my Fielding wonlost records  expressed in terms of zscores  was 0.692. The zscores associated with these two systems differed by more than 1 in 191 cases (19.9%). The correlation between Smith's DRS and Fielding wonlost records was 0.792 with disagreements of 1 or more in 108 cases (11.3%). For a little context, the correlation between DRA and DRS was 0.747 and the two systems disagreed by more than 1 zscore in 147 cases (15.3%).
As discussed above, my results disagree with DRA and DRS more often for outfielders than for infielders. Even here, however, my overall correlations with DRA and DRS are 0.660 and 0.759, respectively. Moreover, as I discuss above, much of the lower correlation in this case is because I incorporate outfielders' (and infielders') individual ability to prevent extrabase hits. In this case, therefore, I believe this modestly lower correlation is an indication of the extent to which I am doing a somewhat better job of fully measuring the overall fielding value of these players.
This is not to say that my Fielding wonlost records are necessarily better than alternative fielding measures (including DRA and DRS). But I am confident that my Fielding records stack up very well with the best alternative fielding measures out there.
Fielding Player WonLost Records vs. Ultimate Zone Rating
Perhaps the most prominent modern fielding measure is Ultimate Zone Rating, which was originally conceived by Mitchel Lichtman and is now reported regularly at
fangraphs.com. For this article, I compared my Fielding wonlost records to UZR data for ten years, 2003 – 2012, for all players who played at least 3,000 innings at a given position over those 10 years, for the seven fielding positions other than pitcher and catcher. The total population of players that I compared here was 288 total players (counting some players more than once if they played 3,000 innings at multiple positions), ranging from 30 leftfielders to 49 centerfielders.
I chose 2003 as my cutoff for comparison (Fangraphs reports UZR data starting in 2002) because Retrosheet data since 2003 is very consistent in providing hittype information (e.g., ground ball, fly ball, line drive), but not detailed location data, for all ballsinplay, hits as well as outs. I stopped in 2012 because that was the last season for which I had data when I first did this comparison (which I have since updated based on my most recentlycalculated Player wonlost records). Overall, 2003  2012 makes for a nice even tenyear sample period.
Conceptual Difference between UZR and Fielding Player WonLost Records
Basic Fielding: Outs vs. Hits on BallsinPlay
UZR is described in great detail on Fangraphs' website
here. The basic concept is that UZR calculates a probability of a ballinplay being converted into an out, based on the location of the ball, how hard it's hit, the handedness of the batter, the groundball tendencies of the pitcher, and various other things. Fielders are then given credit or blame over and above this said probability, so, for example, for a ballinplay that had a 75% chance of being an out, if the ball becomes a hit, the responsible fielder(s) are debited with 0.75 plays not made (0  0.75); if the same ballinplay became an out, the responsible fielder is credited with 0.25 plays made (1  0.75). Plays are then converted to runs based on the average run value of ballsinplay based on the location, et al.
The calculation of my fielding Player wonlost records was described earlier in this article. This aspect of fielding: whether ballsinplay are converted into outs or not corresponds to what I call
Component 5 out of the
nine components of my Player wonlost records.
The key difference between Player wonlost records and UZR is that while UZR's baseline for evaluating a play is detailed information about the location of a ballinplay (as well as how hard it was hit, by whom, and against whom), the baseline for evaluating a play in calculating Player wonlost records is what the final result of the play was  out vs. hit, who fielded it, and what type of hit it was (bunt, ground ball, fly ball, line drive). This is largely because of data limitations with respect to
Retrosheet playbyplay data. I discussed this difference and my treatment of location data in general earlier in this article.
In effect, UZR assumes that two hardhit fly balls to medium center field are created equal. Player wonlost records assume that two flyball doubles fielded by the center fielder are created equal.
Hits vs. Fielding Errors
UZR treats errors somewhat differently from hits. According to the
UZR Primer at Fangraphs, errors are assumed to have been easy plays with high probabilities of being outs. Hence, UZR penalizes fielders more heavily for errors than for hits on ballsinplay.
In contrast, Fielding Player wonlost records treat errors the same as base hits. The key distinguishing characteristic of plays in my system is whether the batter reaches base (and, eventually, what base he reaches).
Michael Humphreys discusses the correct treatment of errors in evaluating fielding in his book
Wizardry: Baseball’s AllTime Greatest Fielders Revealed and shows that the true cost to the fielding team of an error or a hit allowed are identical for any given play (see, e.g.,
Wizardry, pp. 7778). I agree with Humphreys and believe that this is one way in which my Player wonlost records are clearly superior conceptually to UZR.
Additional Components of Fielding
In addition to the basic "range runs" and "error runs" described above, UZR also calculates run values for fielders based on their ability to turn double plays (infielders) and their ability to control baserunner advancement (outfielders). Fangraphs reports these values separate from the UZR estimates based purely on whether ballsinplay are converted to outs or not, but combines them into a single final number which it reports as a player's total UZR.
In addition to
Component 5, I also calculate four additional
components that are credited (at least partly) to fielders.
Component 6 gives credit or blame on hitsinplay based on how many bases the batter takes. That is, it distinguishes between singles, doubles, and triples among hitsinplay. To the best of my knowledge, there is no parallel to my Component 6 in UZR (or any other fielding metric of which I am aware). UZR uses run values based on the average hit value of a ball, based on its location, hit type, etc., but makes no distinctions between hits which actually end up as singles versus otherwiseidentical ballsinplay which actually end up as doubles. Component 6 Player wonlost records are shared between pitchers and fielders at all fielding positions. Component 6 wonlost records are much more significant, however, for outfielders (for whom they account for approximately 14.9% of Fielding decisions) than for infielders (for whom they account for approximately 0.4% of Fielding decisions).
Component 7 gives credit or blame to infielders (and pitchers) for turning double plays on ground balls in doubleplay situations. This is essentially comparable to UZR's doubleplay runs. For Player wonlost records, Component 7 fielding decisions are shared between the fielder who fields a ground ball and the pivot man on the double play (pivot men only receive fielding losses for plays where they receive the ball in time to record a force out but are unable to complete the double play). So, for example, on a classic 643 double play, both the shortstop and second baseman will earn Component 7 fielding wins. It was not clear to me in reading the
UZR Primer exactly who is credited with doubleplay runs on a 643 double play. Component 7 accounts for approximately 10.4% of infielder Fielding decisions.
Finally, Components
8 and
9 give credit or blame to fielders for baserunner outs and baserunner advancement, respectively. This is comparable to UZR's Arm runs. The difference here is that Components 8 and 9 are allocated across all fielders, while UZR Arm runs are only allocated to outfielders. Components 8 and 9 combine to account for approximately 13.1% of infielder Fielding decisions and 32.8% of outfielder Fielding decisions.
Components
5,
6, and
7 are shared between fielders and pitchers, while Components
8 and
9 are allocated entirely to fielders. Because of this, "arm ratings" make up a relatively larger share of Fielding Player wonlost records than they do of total UZR.
Raw Results
For this article, I compare two measures of UZR and (
contextneutral, teammateadjusted) Net Fielding wins (eWins minus eLosses): total UZR runs vs. total Net Fielding wins, and (Range + Error) UZR runs vs. net Component 5 Fielding wins. For outfielders, I also calculate a third measure of Net Fielding wins which excludes Component 6  since UZR has no counterpart  which I compare to total UZR runs. To be clear, on this last measure, I exclude Component 6 only to allow for an applestoapples comparison to UZR. The fact that my Fielding records include this measure of the exact value of the hits allowed by fielders while UZR relies only on average hit values across similar plays is, in my opinion, a clear advantage of Fielding Player wonlost records over UZR as an overall measure of player fielding.
The first table summarizes the results for Ultimate Zone Rating (UZR) and net Fielding Wins (eWins minus eLosses).


UZR Net Fielding Runs (per 1,000 innings) 
Net Fielding eWins (per 1,000 innings) 


Total UZR 
UZR, Range+Error 
Total Fielding eWins 
Component 5 eWins 
Total eWins, excl. Comp. 6 
Position 
No. of Players 
Mean 
Std. Dev. 
Mean 
Std. Dev. 
Mean 
Std. Dev. 
Mean 
Std. Dev. 
Mean 
Std. Dev. 
1B 
36  0.12  4.31  0.09  4.24  0.007  0.188  0.011  0.157 


2B 
41  0.12  4.66  0.05  4.46  0.004  0.244  0.012  0.217 


3B 
49  0.80  6.24  0.77  6.08  0.023  0.289  0.015  0.279 


SS 
45  0.54  4.90  0.43  4.61  0.039  0.237  0.036  0.171 


LF 
30  0.46  6.82  0.21  6.36  0.003  0.342  0.004  0.281  0.008  0.306 
CF 
49  0.52  6.29  0.27  6.25  0.037  0.367  0.014  0.254  0.031  0.349 
RF 
38  0.53  6.38  0.23  5.78  0.046  0.405  0.018  0.299  0.050  0.392 
The first thing that we have to do before we can compare UZR to Player wins is to put them on the same scale. UZR is expressed in runs while Player wins are, of course, expressed in wins. Traditionally, in sabermetric measures, one win is equal to approximately 10 runs. Looking at the standard deviations in the above table, however, the ratio of UZR to Player wins is closer to 20. In other words, even if you converted UZR to wins, using a conventional runtowin translation, the spread of players' UZR is roughly double the spread of players' net fielding wins.
Why is the Spread on Player Fielding Wins lower than Defensive Runs?
I believe that the spread on my net fielding wins is less than the spread of UZR (and other fielding measures) because I assign more credit on ballsinplay to pitchers, whereas standalone fielding measures implicitly assign all of the credit on ballsinplay to fielders, since that's all that they are measuring.
Specifically, looking at my
Components 4 (excluding home runs),
5,
6,
7,
8, and
9, I assign 46.5% of the (defensive) credit for these to pitchers and only 53.5% of the (defensive) credit to fielders (since 2003).
Is this reasonable on my part?
I believe that it is. Econometric research following up on DIPS theory has
consistently found that pitchers have some effect on batting average on ballsinplay (BABIP). The extent to which I allocate such credit to pitchers is based on the extent to which Player winning percentages
persist for pitchers in these
components.
With specific regard to UZR, Mitchel Lichtman, the creator of UZR, looked at
how UZR differs for specific pitchers on the same team (specifically, the
2012 Detroit Tigers) and found significant differences across pitchers. While this was a quickanddirty analysis that really doesn't even rise to the level of a "study", its results are consistent with the likelihood that there is some pitcher "ability" being captured within UZR.
Putting Things on the Same Scale
In order to really compare UZR and what I'll start calling NFW (net fielding wins), it is necessary to put them on the same scale. To do this, I created
"zscores" associated with both statistics. The basic formula for a zscore of variable x is (x  m) / s, where m is the mean of the statistic and s is the standard deviation. I calculated zscores for each player for UZR and net fielding wins using a value of m equal to zero (since both of these statistics are constructed to be relative to league average by construction) and the standard deviations from the above table.
For example,
Andre Ethier scores at 4.67 total UZR (per 1,000 innings in RF), 2.84 Range+Error UZR (reUZR), 0.431 total NFW, 0.255 Component 5 NFW (NFW5), and 0.527 NFW, excluding Component 6 (NFW589) in right field over the time period being analyzed here. From the previous table, the standard deviations associated with these five numbers
are 6.38, 5.78, 0.405, 0.299, and 0.392, respectively. This translates, therefore, into zscores for Andre Ethier in right field of 0.73 for total UZR, 0.49 for reUZR, 1.06 for NFW, 0.85 for NFW5, and 1.34 for NFW589.
I did this for every player referenced in the earlier table. I then calculated simple
correlations between UZR and NFW by position (UZR to NFW, reUZR to NFW5, and, for the outfield positions, UZR to NFW589).

UZR v. NFW 
Position 
Total 
Comp. 5 only 
excl. Comp. 6 
1B 
0.864 
0.879 

2B 
0.810 
0.827 

3B 
0.869 
0.880 

SS 
0.757 
0.788 

LF 
0.730 
0.720 
0.716 
CF 
0.559 
0.493 
0.516 
RF 
0.724 
0.698 
0.731 
Keep in mind that correlations do not tell us which of two measures is more accurate, merely how similar they are to each other.
The correlations associated with the infield positions here are exceptionally high. To the extent that the correlations are somewhat higher for Component 5 only, I believe this is indicative of the extent to which UZR is missing information that I am capturing, particularly via
Components 6,
8, and
9. But, to the extent that the difference in correlations is very slight, this is indicative of the fact that this additional information is fairly minimal (and/or that player fielding value is fairly highly correlated across
components).
The correlations associated with the corner outfield positions, while not as high as those for the infield, are nevertheless extremely high. The relative correlations with and without Components 6, 8, and 9 differ between leftfield and rightfield such that I'm not inclined to really draw any conclusions in that regard. But overall, the level of correlation between these two systems is very encouraging to me. If UZR is capturing something that I am missing, it does not appear to be a very major factor in the infield or the corner outfield positions.
The lowest correlations are for centerfield. Even here, however, the correlation between overall UZR and total Net Fielding wins, 0.559, is fairly high. I look more closely at the centerfield numbers and what they might mean below.
The next section looks more closely at how UZR and Net Fielding Wins compare on a positionbyposition basis.
PositionbyPosition Analysis
First Base
For the 36 players evaluated here, the average difference in zscores between UZR and NFW (NFW minus UZR) is 0.066. The average absolute difference in zscores between UZR and NFW is 0.436. For Component 5 only, the average difference is 0.090 and the average absolute difference is 0.399.
There are two players for whom the difference in zscores is greater than one (in absolute value). Looking only at
Component 5, there is only one player for whom the (absolute) difference in zscores is greater than one. These players are shown in the next table.


Fielding ZScore 


UZR (Fangraphs) 
NFW (Thress) 
Player 
Innings 
Total 
Range+Err 
Total 
Comp. 5 
Mike Jacobs 
3,236.1 
2.099 
2.083 
1.020 
1.439 
Scott Hatteberg 
4,777.0 
0.413 
0.400 
0.604 
0.871 
The only first baseman whose Component 5 zscores differ by more than one was minor
Moneyball star
Scott "Picking Machine" Hatteberg. My system thinks that Ron Washington did a pretty good job of teaching Hatteberg how to play first base.
Second Base
For the 41 players evaluated here, the average difference in zscores between UZR and NFW (NFW minus UZR) is 0.010. The average absolute difference in zscores between UZR and NFW is 0.493. For Component 5 only, the corresponding numbers are 0.043 and 0.466, respectively.
There are a total of three players for whom the difference in zscores is greater than one (in absolute value) for total UZR v. total Net Fielding Wins. For Component 5 vs. Range+Error UZR runs, there are four such players. These players are shown in the next table.
The difference in zscores for
Roberts and
Weeks exceed one for total fielding, but are within one (albeit not by a lot) when only Component 5 is considered. On the other hand, the overall zscores are within one for
Freddy Sanchez but UZR and Player wonlost records disagree more strongly about Sanchez's basic ability to turn batted balls into outs.
Third Base
For the 49 players evaluated here, the average difference in zscores between UZR and NFW (NFW minus UZR) is 0.049. The average absolute difference in zscores between UZR and NFW is 0.403. Looking only at Component 5, the differences are 0.072 and 0.407, respectively.
There are a total of 4 players for whom the difference in zscores is greater than one (in absolute value). Looking at only Component 5, however, there is only one player with zscores that differ by more than one. These players are shown in the next table.


Fielding ZScore 


UZR (Fangraphs) 
NFW (Thress) 
Player 
Innings 
Total 
Range+Err 
Total 
Comp. 5 
David Bell 
4,456.0 
0.960 
0.938 
0.199 
0.013 
Eric Chavez 
6,700.1 
0.624 
0.633 
1.667 
1.510 
Geoff Blum 
3,256.0 
1.157 
1.223 
0.133 
0.331 
Mike Lowell 
8,048.0 
0.163 
0.121 
1.360 
0.939 
Vinny Castilla 
4,308.1 
0.275 
0.206 
0.990 
1.234 
Shortstop
For the 45 players evaluated here, the average difference in zscores between UZR and NFW (NFW minus UZR) is 0.053. The average absolute difference in zscores between UZR and NFW is 0.597. For Component 5, the corresponding numbers are 0.118 and 0.542.
There are a total of five players for whom the difference in zscores is greater than one (in absolute value) for total UZR and six players for whom the (absolute) difference exceeds one for (Range+Error) UZR. These players are shown in the next table.
Left Field
For the 30 players evaluated here, the average difference in zscores between UZR and NFW (NFW minus UZR) is 0.060. The average absolute difference in zscores between UZR and NFW is 0.577. For Component 5, the corresponding numbers are 0.040 and 0.601, respectively. For total Player wonlost records, excluding Component 6, the numbers are 0.040 and 0.579.
There are three players for whom the difference in zscores is greater than one (in absolute value) for total Net Fielding wins and four players when Component 6 is excluded. Looking only at Component 5, there are seven players for whom the (absolute) difference in zscores is greater than one. All of the players for whom zscores differ by one or more in at least one of these comparisons are shown in the next table.


Fielding ZScore 


UZR (Fangraphs) 
NFW (Thress) 
Player 
Innings 
Total 
Range+Err 
Total 
(excl Comp. 6) 
Comp. 5 only 
Alfonso Soriano 
7,680.2 
1.375 
1.060 
0.507 
0.486 
0.048 
Carlos Lee 
10,568.1 
0.222 
0.134 
1.420 
1.332 
1.147 
Cliff Floyd 
3,757.0 
0.402 
0.690 
0.580 
0.817 
0.593 
Hideki Matsui 
5,220.1 
1.871 
2.206 
0.961 
0.955 
1.202 
Jay Payton 
3,572.2 
0.094 
0.026 
1.127 
0.883 
0.983 
Luis Gonzalez 
6,392.2 
0.690 
0.263 
1.377 
1.696 
1.135 
Moises Alou 
4,007.2 
1.003 
1.235 
1.211 
1.115 
0.857 
Randy Winn 
3,084.0 
0.528 
0.713 
1.385 
1.200 
1.730 
Two names that surprised me when I saw them on the above table were
Carlos Lee and
Moises Alou, not because my ratings surprised me, but because I thought it was a widelyaccepted fact that Lee and Alou were belowaverage, and likely well belowaverage, defensive leftfielders.
The next table shows their yearbyyear ratings in UZR and Net Fielding wins, expressed as zscores.

Carlos Lee 
Moises Alou 
Season 
Innings 
UZR 
NFW 
Innings 
UZR 
NFW 
2003 
1,328.2 
0.86 
0.51 
1,219.0 
1.43 
0.63 
2004 
1,277.2 
1.68 
1.16 
1,338.1 
2.00 
0.98 
2005 
1,404.0 
0.25 
0.45 
576.0 
0.61 
0.11 
2006 
1,259.1 
1.69 
2.31 
79.0 
3.71 
0.03 
2007 
1,369.1 
0.32 
1.54 
703.0 
0.65 
2.98 
2008 
915.1 
0.02 
3.06 
92.1 
0.48 
8.03 
2009 
1,272.1 
1.08 
2.50 



2010 
1,096.1 
2.37 
4.62 



2011 
645.1 
1.95 
1.39 



Right Field
For the 38 players evaluated here, the average difference in zscores between UZR and NFW (NFW minus UZR) is 0.030. The average absolute difference in zscores between UZR and NFW is 0.583. For
Component 5, the corresponding numbers are 0.019 and 0.636, respectively. For total Player wonlost records, excluding
Component 6, the numbers are 0.043 and 0.583.
There are a total of seven players for whom the difference in zscores is greater than one (in absolute value) for total fielding, and eight players for whom the difference in zscores is greater than one (in absolute value) in the other two comparisons (
Component 5 and excluding
Component 6). All of the players for whom zscores differ by one or more in at least one of these comparisons are shown in the next table.


Fielding ZScore 


UZR (Fangraphs) 
NFW (Thress) 
Player 
Innings 
Total 
Range+Err 
Total 
(excl Comp. 6) 
Comp. 5 only 
Gary Sheffield 
3,925 
1.478 
1.566 
0.202 
0.352 
0.078 
J.D. Drew 
7,805.1 
0.922 
1.393 
0.009 
0.100 
0.335 
Jason Heyward 
3,524.1 
1.708 
2.166 
0.195 
0.132 
0.839 
Jeremy Hermida 
3,745 
0.360 
0.180 
1.018 
0.556 
0.474 
Jose Bautista 
3,450.2 
0.354 
1.280 
0.984 
1.145 
0.646 
Kosuke Fukudome 
3,273.2 
0.043 
0.138 
1.067 
1.085 
1.146 
Mike Cuddyer 
6,064 
0.724 
1.222 
1.191 
1.164 
2.482 
Randy Winn 
3,622.1 
1.874 
2.027 
0.625 
0.607 
0.944 
Trot Nixon 
3,924 
1.019 
1.129 
0.703 
0.655 
0.639 
Xavier Nady 
3,579.1 
0.473 
0.145 
1.411 
1.486 
1.176 
Randy Winn shows up on the lists for both corner outfield spots. Interestingly, UZR rates Winn as an aboveaverage left fielder and an outstanding right fielder, while Player wonlost records rate Winn as an outstanding left fielder but merely an aboveaverage right fielder.
Jose Bautista just misses showing up on two lists. He shows up here in right field; at third base, the difference in zscores for UZR and Net Fielding Wins was 0.995. In both cases, Bautista scores slightly better in terms of converting ballsinplay to outs in net Fielding wins, but with a difference in zscore of less than one. But in both cases, Bautista scores especially well in the extra components of Fielding Player wonlost records. Specifically, Jose Bautista scores extremely well at preventing baserunner advancement,
Component 9, at all of the positions which he has played over the years; in fact, he's in the
top 5 in career net Component 9 fielding wins among all players for whom I have calculated Player wonlost records. The next table shows Bautista's career record in Component 9 by position.
Position 
eWins 
eLoss 
Win Pct 
Net Wins 
3B  1.9  1.3  0.586  0.5 
LF  0.4  0.3  0.572  0.1 
CF  1.2  0.9  0.576  0.3 
RF  10.6  7.7  0.579  2.9 
Total  14.1  10.2  0.580  3.9 
Center Field
For the 49 players evaluated here, the average difference in zscores between UZR and NFW (NFW minus UZR) is 0.018. The average absolute difference in zscores between UZR and NFW is 0.748. For Component 5, the corresponding numbers are 0.014 and 0.775, respectively. For total Player wonlost records, excluding Component 6, the numbers are 0.006 and 0.771.
There are a total of 18 players for whom the difference in zscores is greater than one (in absolute value) in at least one of the three comparisons made here. This is 36.7% of all of the centerfielders that I evaluated. All of the players for whom zscores differ by one or more in at least one of these comparisons are shown in the next table.


Fielding ZScore 


UZR (Fangraphs) 
NFW (Thress) 
Player 
Innings 
Total 
Range+Err 
Total 
(excl Comp. 6) 
Comp. 5 only 
Adam Jones 
6,371.1 
0.487 
1.030 
1.121 
1.094 
0.282 
Andruw Jones 
7,327.0 
2.223 
1.894 
0.370 
0.270 
0.050 
B.J. Upton 
7,024.2 
0.421 
0.080 
1.417 
1.659 
1.326 
Chris Young 
7,271.0 
0.416 
0.517 
0.620 
0.579 
0.269 
Denard Span 
3,712.1 
0.617 
0.953 
1.651 
1.645 
1.664 
Endy Chavez 
3,200.0 
0.731 
0.040 
2.238 
2.077 
1.130 
Franklin Gutierrez 
3,875.1 
1.995 
1.925 
2.392 
2.254 
3.133 
Jacoby Ellsbury 
4,030.2 
0.746 
1.207 
1.538 
1.655 
1.584 
Johnny Damon 
5,524.0 
1.287 
0.658 
2.018 
1.975 
1.920 
Josh Hamilton 
3,148.1 
1.132 
1.073 
0.046 
0.032 
0.179 
Juan Pierre 
7,316.1 
0.387 
1.079 
1.385 
1.468 
0.778 
Ken Griffey, Jr. 
3,198.0 
3.278 
3.269 
2.125 
1.957 
2.293 
Mark Kotsay 
5,805.1 
0.225 
0.433 
0.733 
0.796 
0.320 
Matt Kemp 
6,025.0 
1.170 
1.371 
0.195 
0.356 
0.014 
Michael Bourn 
6,280.1 
1.307 
1.096 
0.373 
0.211 
0.167 
Rocco Baldelli 
3,332.0 
0.086 
0.663 
1.033 
1.181 
0.348 
Scott Podsednik 
3,326.1 
0.765 
0.669 
0.968 
1.181 
1.504 
Of the players in the above table, the fewest zscore differences greater than one were actually found comparing total UZR to total Net Fielding wins and, interestingly, the most zscore differences greater than one were found when I removed
Component 6 decisions from Fielding Player wonlost records. The correlation between the two measures is also strongest (0.559) comparing totals and actually slips just below 50% (0.493) when net
Component 5 fielding wins are compared to (Range+Error) UZR.
In other words, Fielding wonlost records correlate most strongly to centerfield UZR when Component 6 is included. It seems to me that my Component 6  which measures whether a fielder gives up singles, doubles, or triples on hitsinplay  acts as an effective proxy for outfield location data.
Focusing on total UZR vs. total Net Fielding Wins, there are 11 players (22.4%) with zscores which differ by more than one:
Adam Jones,
Andruw Jones,
Chris Young,
Denard Span,
Endy Chavez,
Jacoby Ellsbury,
Josh Hamilton,
Juan Pierre,
Ken Griffey, Jr.,
Matt Kemp, and
Scott Podsednik.
In the case of Griffey, while his two zscores differ by 1.15, they basically agree that he was an extremely bad defensive centerfielder over the time period in question here. His UZR zscore is 3.3 versus an NFW zscore of 2.1. In fact, both of these zscores are the lowest among the 49 centerfielders considered here. Really, UZR and Fielding wonlost records agree more than they disagree about Griffey's latecareer fielding.
Let me focus, though, on 10time Gold Glove winner,
Andruw Jones. Jones scores significantly better in UZR than in Net Fielding wins over the sample period considered here. For his career, I actually agree that Andruw Jones was a brilliant defensive centerfielder, among the best of alltime. In fact, he ranks
second alltime in career Net Fielding wins among all centerfielders for whom I have calculated Player wonlost records. He rates as the best defensive centerfielder in the National League for five consecutive seasons from
1998 through
2002, in most cases by a lot (he led 2ndplace
Terry Jones in net wins 1.8  0.4 in 1998). I still think that (Andruw) Jones was pretty good from 2003  2006, although not the best in the league anymore, but had fallen to belowaverage by 2007.
In contrast, the UZR numbers at
Fangraphs show Jones as remaining an excellent defensive centerfielder through 2007 (average UZR from 2003  2007 of 20.8 runs per season).
Except for one little twist. The UZR numbers on Fangraphs aren't the only UZR numbers for Andruw Jones that Mitchel Lichtman has calculated.
Sensitivity of UZR to Data Source
Earlier in this article, I explained how I use location data in calculating Fielding wonlost records and defended my decision not to use this data directly even for those seasons where
Retrosheet provides location data. As part of that discussion, I cited two studies that were reported on the Internet that looked at differences in UZR calculated using different source data.
In August, 2007, Hardball Times published an article by Michael Humphrey (the author of
Wizardry), titled
Ghosts in the Outfield. In this article, Humphreys reported a comparison of what he called "simplified UZR" ratings using two sets of location data: one from BIS (Baseball Information Systems) and one from STATS. For 2003  05, over a sample of 24 outfielders, the two systems  which should have been identical in all respects except for the firm/person recording the location and hittype data  had a correlation of only 0.60.
Humphreys' sample included 9 centerfielders. The results for these nine players are shown below. Humphreys' numbers are for 2003  2005 and are presented as runs saved per 1,450 innings played (~162 games).
So, Humphreys is showing a correlation of UZR
with itself of 0.60 for all outfielders and 0.52 for centerfielders. Suddenly, my correlations of net Fielding wins to UZR of 0.66 for all outfielders and 0.56 for centerfielders look pretty good, don't they?
In a more detailed analysis along the same lines, Mitchel Lichtman, the creator of UZR, calculated UZR data for 2003  2008 using data from BIS (bUZR) and again, using the same UZR system, using data from STATS (sUZR) for 240 players. The results were discussed by Licthman, Tom Tango, and others
here. That discussion did not include a correlation between the two, but it was noted that "5 of the top 9" differences in players were for centerfielders. The top two differences were
Andruw Jones, +112 runs using BIS data vs. 5 using STATS data, and
Carlos Beltran, +9 using BIS data vs. +86 using STATS.
Using the UZR standard deviation (per 1,000 innings) for centerfielders reported above (6.29) and Jones's and Beltran's innings played from 2003  2008, the two numbers quoted in the previous paragraph for Jones translate into zscores of 2.45 (BIS) and 0.11 (STATS), a difference of 2.56. For Beltran, the two zscores are 0.19 vs. 1.78, a difference of 1.59.
Of the 288 players that I looked at for this article, the difference in zscores between UZR and Net Fielding wins exceeded 1.60 in only 10 cases (3.5%): 6 CFs, two corner outfielders, and two infielders. The difference in zscores between UZR and NFW exceeded 2 in only 3 cases:
Moises Alou (LF),
Jacoby Ellsbury (CF), and
Denard Span (CF). And there were
zero players for which the difference in zscores exceeded the difference in zscores for Andruw Jones with BIS vs. STATS UZR.
Tom Tango reported that the standard deviation of the difference between bUZR and sUZR in Lichtman's study was 6.0 runs per 150 games with a median difference of 4.0 runs per 150 games and 10% of players having a difference of at least 10 runs per 150 games. I converted my Net Fielding wins into a UZRlevel number (by multiplying my NFW zscores times the UZR standard deviations by position) and calculated differences between UZR and this UZRlevel Fielding wins number per 150 games (actually, per 1,350 innings). These results, compared to the results reported by Tango are shown in the next table.

UZR vs. Net Fielding wins 
bUZR vs. sUZR 

Std. Deviation of Difference 
5.5 
6.0 
Median Difference 
3.4 
4.0 
Difference > 4 
42% 
50% 
Difference > 10 
7% 
10% 
Or, in words, my results are closer to UZR (using BIS data) than UZR results are to themselves.
Conclusions
Overall, I'm quite pleased with the results here. The total correlation across all seven positions (288 players) investigated here, between UZR and my Fielding wonlost records  expressed in terms of zscores  was 0.742. The zscores associated with these two systems differed by more than 1 in 35 cases (12.2%).
For infielders, the results are even better. The correlation between total UZR and Net Fielding wins for infielders was 0.830 for the 171 infielders that I evaluated here. The zscores associated with these two systems differed by more than 1 in 14 of 171 cases (8.2%).
As close as these two results are, some of the difference between the two systems is because Fielding wonlost records incorporate factors which are not considered in measuring UZR, including fielders' abilities to limit extrabase hits and control baserunner advancement. When only the common factor of simply converting ballsinplay into outs (what I call
Component 5) is compared in the two systems, the correlation across all infielders rises to 0.863 and the number of cases where the zscores differ by more than one falls to 12 (7.0%).
For corner outfielders, the results are not quite as close as for infielders, but are still very similar. The correlation between total UZR and Net Fielding wins for corner outfielders was 0.728 and the zscores differed by more than 1 in 10 of 68 cases (14.7%).
The results for centerfielders show the lowest correlation, 0.559, with zscore differences greater than one in 11 of 49 (22.4%) cases. My results for centerfielders are closer to UZR including
Component 6  the extent to which hitsinplay are singles, doubles, or triples. This suggests to me that taking explicit account of hitsinplay serves as a useful proxy for more detailed location data.
The lack of correlation between Net Fielding wins and UZR for centerfielders is not necessarily an indication of a weakness in Net Fielding wins as a measure of fielding ability. In fact, my results visavis the UZR numbers presented by
Fangraphs are comparable to comparisons of UZR calculated using different data sources.
Overall, I believe that my Fielding Player wonlost records stack up extremely well as measures of player fielding with any other fielding measures out there.
Home
List of Articles
The information used here was obtained free of charge from and is copyrighted by Retrosheet. Interested parties may contact Retrosheet at "www.retrosheet.org". Baseball player wonlost records have been constructed by Tom Thress. Feel free to contact me by email or follow me on Twitter.