Examining J.T. Snow's Defense

I know it's 2011 and all, but I like to look back on the old days every once in a while.  This has absolutely nothing to do with the 2011 Giants, but I figure this may make some good food for thought.

Growing up, J.T. Snow was my favorite player.  He and I were alike in a lot of ways-we shared the same nickname, took a good number of walks, and hit for modest power at a position that stereotypically consisted of hulking sluggers with iron mitts.  We also shared another thing in common: we were both good defensive first basemen.  I even wore the same glove-the black Wilson A2000-and wore a similar forearm band on my right wrist.  His swing was a thing of beauty to me; his play in the field divine. 

The man's reflexes at first were admirable, and in my opinion, Snow was one of the slickest fielding first basemen I've ever seen.  When people think of "elite defensive first basemen," they usually think of Keith Hernandez and Don Mattingly.  I think of Snow.  He was a fan favorite in San Francisco and for good reason.  He was a fine ballplayer and he was a steady presence for the Giants for many years.  His peers believed him to be a phenomenal glove man as well, awarding him with six Gold Gloves over the course of his career.

But...maybe he wasn't as good as most thought him to be.

While looking at Snow's career numbers over at Baseball-Reference, I was surprised to find Sean Smith's Total Zone(TZ) estimates that Snow was approximately 30 runs below a league average first baseman for the duration of his career.  TZ, for those unfamiliar with it, is a defensive metric that incorporates play-by-play data from Retrosheet.  It uses the batted ball out tendencies of each individual hitter as a proxy for estimating where the ball in play was hit in order to estimate each fielder's individual chances.

From 1989 through 1999, however, Total Zone uses more detailed information-batted ball locations from Project Scoresheet, which you can find here.  This rendition of TZ uses the out conversion rates from each "bucket" (i.e. 3 versus 3/4, or 6 versus 6M for shortstops) to determine how the player fared relative to average.  Think of it as a crude rendition of UZR, without adjustments made for the hardness of the batted ball hit.  I assume Smith makes adjustments for the batter's handedness.

In any case, despite the "more precise" estimate of chances, I decided to cross check Smith's estimates with STATS Inc.'s Zone Ratings (ZR), which date back to 1987. ZR, which cuts the baseball diamond into 22 major slices (and then cuts it down even further), and assigns "zones of responsibilities" to different fielders based on the out conversion rate (responsibility is given for zones with >50% of batted balls turned into outs).  First basemen, for example, are responsible for zones V-X.  This is a similar process to TZ, but it's a different approach-rather than using smaller buckets, ZR uses one large zone.  I'm uncertain as to the data source from Project Scoresheet/Retrosheet-I don't know who is recording the batted ball locations-so a different source, one that implements multiple stringers and reviews video, should lend a nice, alternative insight.

STATS ZR lists Snow as a -49; about twenty runs worse than TZ.  This is a huge discrepancy, as twenty runs are a bit over two wins of value.  The problem might be with the stringers (the individuals recording the data), so perhaps an alternative source might provide us with a different answer.

Michael Humphreys' Defensive Regression Analysis (DRA) uses regression analysis (surprise!) to estimate a player's runs saved or cost based on variables such as left-handed and right-handed balls in play and their ground and fly ball outs.  It's a relatively obscure metric, at least in comparison to TZ, but it will likely receive quite a bit more attention now that Humphreys' book, Wizardry, is out.  There is no issue with stringer bias here.  DRA lists Snow as a -15.  That's substantially better than ZR, which seems to hate him, and better than TZ, which merely dislikes him.  DRA just isn't much of a fan, either.

Tom Tango was kind enough to share with me his With Or Without You (WOWY) figures for Snow, which looks at the rate of balls in play converted into outs with J.T. on the field, compared to all other first basemen (with the same pitchers on the mound).*  It doesn't get much simpler than that.  Snow rates as a +44, or about +3 runs per year, over the course of his career.  Hallelujah!  Some evidence that Snow was a good defender.  But, as Humphreys was so kind to point out to me, WOWY counts all outs on batted balls, which includes pop ups and the like.  This is certainly a factor-catching infield flies account for 7.7% of Snow's total putouts afield, as compared to an observed league rate of 6.6%.  Over 12,930 putouts, that infers he caught about 145 more pop flies than the league (this is without accounting for actual infield fly opportunities, however).  Humphreys suggests that Snow was perhaps a "ball hog" on balls in the air.

That leaves us with two batted ball location sources that strongly dislike him, one that estimates opportunities based on purely mathematical formulae that isn't much of a fan, and one metric that thinks he's decent, but has a bias towards easy outs.

At this point, I'm beginning to wonder if Snow wasn't as good of a defensive first baseman as I (and most others) originally thought.  Yes, there are a multitude of issues with defensive metrics-much has been made of the unreliability of batted ball data, and there are so many intricacies left unaccounted for, they are certain to be missing a number of things-but I find it a bit alarming that he's consistently being seen as less than spectacular, which makes me wonder if Snow's defense has been somewhat overrated.

Then again, Snow's range hasn't been lauded nearly as much as his "sure hands" have.  Sure enough, Snow has committed 38 less errors (catching, throwing and fielding) than average over his 14-year career, dropping about 12 less catches, six less errant throws, and committing 21 less fielding errors than the typical first basemen.  When Snow got to the ball, it wasn't getting past him-and he very rarely made the mistake of dropping a throw or throwing a ball away.

In addition to his sure-handedness, Snow was also consistently praised for preventing errant throws from the other infielders.  Due to the complete lack of detailed data, we don't know how many scoops J.T. made, or how many chances he had-but if we apply the same WOWY method-by looking at infielder's rates of throwing errors with Snow and without him-we find that he was pretty darn good in this category.  Sean Smith, the developer of TZ and rWAR, estimates that Snow saved approximately 22 runs, or close to two runs per season,** by turning some of those errant throws into outs.  This would change his TZ rating to a -8, his ZR to -27, and his DRA to a +7.  Better, but still unspectacular.

All in all, ZR estimates Snow to be a -3 defender, TZ a -1, and DRA a +1.  Pretty much a league average first baseman; nothing particularly special.

This makes me think:

1. Something that immediately pops out is the huge difference we see when we actually estimate the impact of errant throws saved.  Analysts generally say that, due to the relatively low spread in scooping talent (+/- 5 runs), it isn't a big deal to ignore scoops in player value analysis.  In the case of Snow, I'd say it's pretty important.  On a micro scale, say, a season, it probably won't make a big difference.  But on a macro scale, it's something that we need to pay attention to.  Especially if we attempt to do historical player assessments and comparisons.

2. Perhaps the numbers really are missing a lot, so much so that the numbers are completely unreliable and that Snow really did have exceptional range in the field.

3. Or, perhaps our perceptions of Snow have been clouded by the fact that he was so smooth in the field.  Call it the "Derek Jeter Effect."  A player makes a bunch of diving plays not because he's a tremendous fielder, but because he lacks the range to make a play that an otherwise average fielder would.  They make it look much harder than it really is; but damn, they sure do it with style.

I'm hoping the reason behind those low range totals is #2, not #3.  In fact, I'd probably put good money on it.  The question is, how far off are those estimates?

*For a wonderful read on WOWY, I highly suggest anyone read Tango's article on Derek Jeter in The Hardball Times Annual, 2008.  Luckily, it's freely available for all to read here.

**I know two runs may seem like a low total, but some detailed work from Baseball Info Solutions, which uses "video scouts" to track opportunities and scoops, suggest that the best first basemen in the game will save somewhere around five runs per season.