THE BOOK--Playing The Percentages In Baseball

http://www.fangraphs.com/blogs/index.php/historical-position-adjustments/

I posted this at BPro:

Nice article. Position adjustments are a hairy issue, but they’re so critical to everything.

...

Do you think the “Mays Problem” as you describe it is really an issue of outliers? Or is it just an overall talent gap between positions? Teams often start most players out of high school and college at the far left of the defensive spectrum (SS if righty, CF if lefty) and then move them to the right when you decide they can’t play that fielding position adequately.

In other words, doesn’t your method still assume, in the end, that total position talent (offense + fielding) is constant across positions? I’ve always felt that assumption was the biggest problem with offense-based position adjustments, because I don’t think that’s likely to be the case. 2B, in particular, seems to be a position to which players are moved to when they can’t field well enough to be a SS, but you’re also not a particularly good hitter. I’ll accept that they should be better fielders than 3B’s (skills should be different, of course), but I do tend to think that 2B is a below-average position in terms of talent.

Here are data from Tango’s fan scouting report last year. No position-specific weights, just overall averages of skills (rated 1-5) across positions:

CF 3.7
SS 3.6
2B 3.4
3B 3.3
RF 3.3
C 3.0 (weird position, though, so not apples to apples)
LF 3.0
1B 2.98

This seems about right to me in order and size of gap between players, maybe with the exception of CF’s over SS. I wonder if this kind of data, though, is the better solution to all of the various issues you’ve raised with position switching data than going back to offense-based adjustments. Obviously it only applies to modern baseball, but it’s a start.

Cheers,
Justin

"No matter what UZR says about the position switchers of LF-RF, it seems pretty clear that MLB teams, these days anyway, put their worst fielder, usually, in LF, not RF.”

There has been a bit of a shift in this area though as teams have started to play guys who can cover a lot of ground in LF: Carl Crawford, Brett Gardner, Juan Pierre, Jose Tabata, Nyjer Morgan (as a Pirate), et alii.

I agree with Colin in the BPro comment section that the issue is mostly about 2B v 3B.  Here are threads that discuss that:

http://www.insidethebook.com/ee/index.php/site/comments/2b_v_3b/

http://www.insidethebook.com/ee/index.php/site/comments/2b_v_3b1/

http://www.insidethebook.com/ee/index.php/site/comments/2b_v_3b_v_cf/

If you’re using hitting values, would it make sense to only look at the bottom half of all players at a position, and then dole out responsibility based on that? Say, calculate Super Linear Base Runs Created (the exact runs generated formula won’t matter a whit) for Atlanta catchers, and for Baltimore catchers, and so on and so forth, then compare the worst fifteen team/catchers to the worst fifteen team/first basemen and other positions, and then dole out fielding responsibility based on that. It might get out of the Willie Mays problem.

A quick follow-up:

http://www.baseballprospectus.com/article.php?articleid=11944

I don’t have positional adjustments figured the way you and Rally have presented them, but you should be able to derive them by taking the positional RPA and subtracting the league RPA, then multiplying by whatever PA constant you were using. (Rather than making you derive the RPA values from the chart, I published the actual values in that link above.)

jar75/#3: I wonder if those speedy/good-fielding LFers are put there because of a weak arm?  I.e. teams really really want a good arm in right, and while these are better fielders overall, it’s a better distribution of talent to put the better arm in right.

(Of course, you could have the two guys switch for each batter based on flyball tendencies/handedness...)

Nice summary graph of Rally’s research, if I can be so bold:

http://www.beyondtheboxscore.com/2009/1/27/737974/position-adjustments-acros

Sky/7,

Yes, I do think arm is a factor. I don’t know how close this is to reality, but I’d assume the main rationalities are:

-More RH batters in baseball means more flyballs to LF and the value of a rangy player there increases.

-A strong throwing arm is best utilized in a non-LF position.

-Managers are reluctant to change a position player’s position based solely on a temporary match up.

In Gardner’s case, at least, it may have to do with the Park.  LF in Yankee Stadium is big.  RF is small.  Fenway is the opposite, so it paid to have Manny in LF and a superior defender in RF.  Ellsbury was supposed to play LF this season, but a good case could still be made that he would have been the weakest of the three starting OFers (Cameron, Drew, Ellsbury). 

Carl Crawford has been playing a mean LF for the (Devil) Rays for many years.  I don’t recall Tropicana having an especially difficult LF.  But, if RF and LF are basically equal from the standpoint of required fielder range, then you put the guy with the weak arm in LF.  That may explain Crawford.

Heh, crosspost with jar75… batted ball distribution may explain it.  Good thought.

I don’t think there really is much difference in the quality of defense in LF and RF, and I’m not convinced that’s a recent development either.  In addition to players who move between LF and RF rating about the same at each position, there are a couple other things that lead me to trust that data as an accurate representation of the difference.

One is that teams tend to use left fielders in center (or center fielders in left) more than right fielders.  They would not do that if there were a dominant tendency to put your worst outfielder in left rather than right.

The other is that, historically, there is very little offensive difference between LF and RF.  While there are problems with using offensive figures to compare defensive ability, you would think that if there were a consistent defensive discrepancy between two positions, and the pool of players who can play each position has a huge overlap, that long-term, the discrepancy would most likely show up in the offensive numbers.  Since it doesn’t, I don’t see much reason to not go with the data Tango uses that says they are pretty much equal.

As jar75 mentioned, there are also anecdotal examples of teams putting their best corner OF in LF, as well as examples of teams putting really bad fielders in RF (i.e. Dye, Abreu, Hawpe).  I don’t think there is a clear trend either way so much as that is the conventional wisdom, so the examples that do fit jump out and look like trends.

I would agree that arm strength is probably the primary factor for a lot of teams (or possibly park configuration in some cases), and if the better fielder has a clearly stronger arm, he’ll likely go in right, while if he has a clearly weaker arm, he’ll likely go in left.  I don’t think that causes much difference in the overall defensive value of the two positions.  I think the end result is better arms in right, a little better range in left, with neither difference having a significant run value.  I could maybe accept them being anywhere from equal up to 1 run apart, but I don’t see any reason to think the difference is more than that.

Your “really bad fielders in RF” is a little misleading. From TZ, Dye and Abreu were both above-average, and Hawpe was below-average, but not horrendous at the start of their careers. I think those examples just show that people are unwilling to shift positions (either players or managers).

"-More RH batters in baseball means more flyballs to LF and the value of a rangy player there increases.”

How sure are you of that?  From what I’ve seen there are about as many fly balls hit to right as to left.  Flyballs seem to be hit as often to the opposite field as they are pulled.  Check out how often outfielders shade hitters the other way.

One of my favoirte Bill James quotes, from the NBJHBA:

“Those who can throw but can’t run are right fielders,

Those who can run but not throw are left fielders,

Those who can do both are center fielders, and

Those who can’t do either are first basemen.”

Rally/14,

Not confident at all. I was just looking for logical reasons why teams could be favoring rangy left fielders. I did not do any pre-post research.

Have you seen any noticeable difference in power to the pull side? Perhaps opposite field fly balls are more often of “lazy fly ball” variety?

According to Retrosheet’s batted ball classifications, there are more fly balls to right but more line drives to left (judging by who fielded the ball).  RF also has a slightly higher zone rating (at least in recent years), and players who switch between right and left see a rise in ZR by about the same amount as the overall difference, which you’d is what you’d expect if the balls hit to left are harder to field.  So, there is probably something to the balls hit to left being a little harder to field on average.  I think RF had more BIZ when I looked at that, though (which you’d expect from more fly balls and fewer line drives).  I don’t know if it makes any difference or how much it might make.

The range thing could be mostly a function of players having weak arms needing more speed or better reactions to make up for it, and players with strong arms being able to play without good range.  I wouldn’t be surprised if some teams put players with better range in left because they think more balls are driven there, though.

The cliche I’ve always heard is that ground outs are pulled, weaker fly balls are sent the other way, and well hit balls (line drives and home runs) are pulled.

(You know, because this group puts a lot of faith in cliches one random guy heard growing up...)

Sky@18: Yes, in fact that is true. The data I’ve seen shows something like for every 8 extra AB-H-SO a team has against right-handed batters than the league proportion, the third baseman will get an extra assist. However, for every 25 or so extra AB-H-SO a team has against right-handed batters, the RIGHT fielder will get an extra putout, and the hits to right field will NOT go up in proportion to putouts. (I forget the exact numbers of outs in play for each. I have it somewhere, and use it for making CAD versions. They’re from like a season of data I got from someone else, like the 1980 NL I think, and Tom could make a better one with more data, but it isn’t going to be much different.)

Colin: ok, the chart didn’t come through at the office, and your data file was also blocked.  Now that I’m here, I’m trying to understand what you are doing.  Are you somehow using the offensive performance to figure out the adjustments?

For example here is your LF numbers for an 8-yr period:
1.229
1.160
1.123
1.069
1.064
1.118
1.150
1.199

Was there that much great fielding LF coming in, and then leaving at that period of time?

To me, this at least begs for a smoothing, so that we are not beholden to annual blips of performance data that doesn’t reflect true talent.