THE BOOK--Playing The Percentages In Baseball

Wednesday, September 08, 2010

Constructing fielding spectum values

By Tangotiger, 02:51 PM

For those new to this issue, and I highly encourage you to read the many threads we’ve had on this subject these last few years, let me give you some inviolable truths:

1. On a league level, the average fielding CF must be better than both of the average fielding LF and RF. Sure, for some minority of teams, it is possible that the LF or RF is a better fielder than the CF. Ichiro for a couple of his years, same with Crawford, and so on. But, given 16 or 30 teams, that won’t happen. To the extent that it can possibly happen, then chalk that up to a system limitation that will self-correct extremely quickly.

2. On a league level, the average fielding SS must be better than the average fielding 3B and 2B. In this case, it’s not such a hard and fast rule. For some reason, Zimmerman, Chavez, Beltre, Rolen, Longoria, et al are playing 3B instead of SS, presumably because of body type. And, for reasons we accept, Barry Larkin was at SS when Pokey Reese was at 2B. These things will happen to some extent. And perhaps in a 16-team league, it might happen that for a year, maybe two, the average fielder at 3B or 2B will be greater than that of SS. It will be inconceivable that both 3B and 2B are better fielders than SS though. And for 30 teams, I don’t see it possible that 3B or 2B will exceed SS on average. Again, I’d ensure that SS is always better.

3. On a league level, the average fielding 1B must be worse than the average IF and average CF. Sure, you might get into a situation where you have a bunch of Minkys and Erstads around, guys that should be playing 2B or 3B or CF. But, that just won’t happen league-wide.

4. The average IF (2B, SS, 3B) must be better than the average OF. This is true if for no other reason that LH throwers would be abysmal in the IF. But more than that, infielders are involved in more plays than outfielders, and while the OF may have more higher variance plays (extra base hits possible), it won’t make up for the gap. Not to mention that virtually all OF-IF moves are infielders going to outfielders.

Anyway, those are the 4 hard-and-fast rules you should adhere to. If Rally or Colin or me or whoever else does this has any season in which any of those four rules are broken, I’d like to see it. (It may have happened pre-1919 or pre-1901.)

• Sabermetrics • Fielding

#1 Greg Rybarczyk (see all posts) 2010/09/08 (Wed) @ 16:09

If you can tolerate an ignorant question from the peanut gallery…

Why do you have to tell your system how much better one position is, on average, than another? Isn’t that sort of like having a scale that demands to know whether you are fat or thin before it is capable of telling you how much you weigh?

I’m hoping for a paragraph type answer, right now I haven’t got time to scrutinize the other threads. I’ll bet I am not alone in wondering this, though…

#2 Tangotiger (see all posts) 2010/09/08 (Wed) @ 16:18

Because our data is based on samples, and possibly biased samples at that, we don’t have the luxury for a complete set of translations.

In short, our uncertainty range is big, if we rely only on the data.

Not to mention that every year, we have new players, and every year, teams move players around. So, we don’t even have the luxury of stable translations.

On top of which, LH throwers don’t play 4 of the 8 positions, which further changes the values.

And don’t get me started at the tough task of dealing with catchers.

So, given all the uncertainties, we need to ensure that the real-life model is adhered to. We know SS are better than 2B and 3B, and we know CF are better than LF and RF. If your automated system says otherwise, then you’ve made a bad assumption in your system, or your system is based on biased data.

#3 Greg Rybarczyk (see all posts) 2010/09/08 (Wed) @ 16:59

OK, I can work with that. I like the whole “recognizing that the data is incomplete yet still trying to make the best of it” approach, it’s very… pragmatic… you ought to hold this up as an example of flexible thinking when someone comes along with the same old “stat-heads just blidnly follow numbers” trope…

#4 Tangotiger (see all posts) 2010/09/08 (Wed) @ 17:51

This also came about because statheads were blindingly doing this, even for DH. The average hitting DH is worse than the average hitting 1B. So, an average fielding 1B ended up with more value at DH than 1B, which made no sense.

Plus, I never liked the whole everyone is average, because we don’t do that for any other sport. An average center is better than an average winger. An average QB is better than most, if not all, other positions, and so on. An average high school SS is better than an average high school 2B.

This is one of those few things that Pete Palmer got wrong in The Hidden Game.

#5 (see all posts) 2010/09/08 (Wed) @ 18:51

TT@4: Pete also missed the park adjustments (though he fixed them for Total Baseball) and the fielding system, en toto. Indeed, the fielding system is so bad, even in retrospect, that the position adjustments are much better by comparison.

But more germane ... in one of our old BBBAs, someone studied the DH penalty, and found that it also applied to the first at bat of the game, where the you either haven’t played the field or haven’t played the field much. This strongly suggests an injury of some sort for the DHes.

Page 1 of 1 pages

<< Back to main

THE BOOK--Playing The Percentages In Baseball

Wednesday, September 08, 2010

Constructing fielding spectum values

Latest...