Baseball team and player performance examined realistically and accurately.Search this site, or just roll your cursor over the colored boxes below the pictures.
"Thou hast set thine house of defence very high."
Quick page jumps:
(This discussion assumes you have already read the material available on this site about general baseball statistical-analysis theory.)
Baseball is traditionally held to comprise three elements: pitching, hitting, and defense. In that tradition, the three are further held to be roughly equal in importance. Modern baseball analysis, discussed at length all over this site, has had much to say about batting and pitching performance; but many feel, as analysis pioneer Branch Rickey put it in 1952, that there isn't much that you can (analytically) say about baseball defense. But that is not so at all.
First off, we know (from the successful games-won formulae) that runs scored and runs allowed are of equal value*, so since runs allowed are a function of pitching and of fielding, clearly batting, pitching, and fielding cannot be of even roughly equal importance: offense must come first, then pitching and fielding can vie for second and third place. And, as we will see, there are profound problems in trying to disentangle pitching from fielding (or vice-versa).
* In truth, offensive runs and defensive runs are not dead exactly equal in value, because improvements in offense (more runs scored) raise the "run environment", while improvements in defense (fewer runs allowed) decrease the "run environment" (the "run environment" determines the relative importance of runs, because one run tends to mean less when many are being scored and more when few are being scored). But in the realistic range of run scoring in major-league baseball, the difference is essentially negligible. A .500 team that both scores and allows 750 runs but then improves its defense by 50 runs (to allowing only 700) will likely win 86 games; if it instead raises its offense by 50 runs (to score 800), it would likely win…86 games. The exact projected numbers are 86.25 wins from improved defense and 85.92 wins from offense, a "difference" of 0.33 wins that washes out in the rounding off (since one cannot win or lose a fraction of a game). And a 50-run change is a pretty big one. So none of that materially impacts the order of importance.
As to pure fielding, one major problem (as always with analysis) has been a lack of good data. The traditional (that word again) baseball measure—Fielding Percentage—may have been (but one doubts even this) useful in an era when even "major-league" baseball fields were so terrible that errors were common in the best of cases: today it's flat-out silly.
Only yesterday, or so it seems, the concept of "Range Factor" was introduced in an attempt to make better measure of individual fielders' abilities. Range Factor in its earliest incarnation was total successful fielding chances (by player by position) per game played; later versions reasonably changed the measure to total successful chances per 9 innings played at a given position (rather than "games", which could mean anything from 15 innings to 1/3 of an inning).
More recently, we have seen a slew of player-fielding measures, such as "UZR" (Universal Zone Rating"), many derived from "secret-sauce" undisclosed proprietary techniques (some of which are actually subjective, or at least rely on certain subjective assessments). We will not here explore the problems with such numbers, though they are manifold; rather, we will focus on fielding at the team level.
A team-fielding metric is not hard to invent, though there are some tricky details. We start by asking ourselves what is the basic job of fielders? Answer: To make an out on every possible play. And what is a "possible" play? We start with every man who comes to the plate, then subtract out those men whom the fielders can do nothing about: walks ("you can't defense a walk"), hit batsmen, strikeouts, and home runs (in theory some minute fraction of home runs could be caught at the fence by a leap, but the number is very small, and the variation in that number from team to team is negligible).
Stats, Inc. reported (in their Baseball Scoreboard for 1996) that the number of home-run-saving catches in all of major-league baseball for the 1995 season—presumably representative—was all of 64, or about 2 a team a season. Even over a three-season period, no one outfielder made more than 6 such catches, or 2 a season a man (and those in parks with low fences). And we may confidently assume that at an average rate of 2 a year a team, the variation from team to team in ability to stop at-the-fence homers is utterly negligible: those that can be made mostly are made and those that can't, aren't.
So the metric would seem to be outs made by fielders divided by "available" outs. Outs made by fielders are all out minus strikeouts. Now let's look at those tricky details (besides the minute "home runs at the fence" issue). One is that some especially important outs are those made on the bases, notably double plays and caught-stealings, plus pickoffs and outfield assists; they are especially important in that they deal with men who already successfully reached base, and so are an imminent threat to score. That's not a problem except if they are men who reached base by one of the not-a-fielding-opportunity means such as a walk: a man who walks and is then thrown out on a double play does not show in the denominator (because he was not "a fielding opportunity"), but he does in the numerator, because he became an out the fielders made. Well, we just don't worry about it because there's little we can do about it. We are chiefly interested here is comparative measures—team vs team—and those things are small parts of the total that almost surely vary in proportion as the team is good-fielding or poor-fielding, so that the metric still expresses team fielding very well.
We call that metric "Fielding Efficiency" (or just FE). We first developed and published it in 1981 or thereabouts. Well after that, Bill James came up with something he called "Defense Efficiency Rating". The DER denominator is the same as the for FE (all batters less home runs, strikeouts, walks, and hit batsmen), but the numerator differs: it is basically all "available" batters not put out by the fielders (that is, it is the denominator minus hits and errors made on batted balls). That formulation sidesteps the issue of outs made on runners already on base, and so is, in a sense, more exact; but we feel that that sidestep ignores some of the more important aspects of fielding as run prevention. Mais chacun à son goût.
The perspicacious will have noticed the resemblance of these fielding metrics to a sort of "reverse BABIP". BABIP is a fairly new and quite interesting stat, being "Batting Average on Balls in Play". Now "Balls in Play" is essentially the same thing as the denominator of FE and DER (BABIP seems to ignore sac bunts, and no one pays any mind to Catcher's Interference because it is so rare), so BABIP is very like 1-DER (but not FE, owing to its recognition of on-base runners put out by the defense). But the link is close: in their essence, FE and DER measure the men put out by the fielders, while BABIP measures the men not thus put out (and also ignores errors).
Elsewhere on this site, in the daily stat tabulations, we show team fielding and team FE numbers. But neither an FE nor a DER directly answers the question of how fielding affects run scoring, and thus wins. Fortunately, with a little thought we can use the FE to analyze with reasonable accuracy the contributions (always at the team level) of fielding to total defense. We do that on the page linked at the head of this paragraph, and so won't here repeat the explanation of the mechanics of how it all is done.
What we will say here is that the data show that fielding is something like 25% to 30% of total defense. Defense being almost exactly (see the note above) 50% of the game, fielding is thus between 10% and 15% of the total game. Something like 35 years ago, we wrote that fielding was about 5% of the game; we now know that that was an underestimate, but it wasn't a grotesque underestimate; pitching still outranks fielding by almost a 3:1 ratio (speaking just of the defensive side of the game).
All content copyright © 2002 - 2017 by The Owlcroft Company.
This page was last modified on Sunday, 9 August 2015, at 8:51 pm Pacific Time.