OPS and the Ecological Fallacy

[calexpat]

Last month, there was some discussion on this board and elsewhere about whether Barry Bonds's walks were really as valuable as his gaudy OPS made them look, given the fact that they were in many cases semi-intentional. This got me to thinking a bit more generally about OPS and OBA, and how we know that they are, in fact, "life."

As I understand it, the major evidence for OPS--and for other sabermetric measures such as EquA and RC--is that team OPS (and EquA, and RC) are highly correlated with team runs scored. It is then inferred that the OPS, EquA, or RC of an individual player is highly correlated with the runs he is responsibile for.

If that is the logic, it is a clear case of the Ecological Fallacy, which "consists in thinking that relationships observed for groups necessarily hold for individuals" (see http://www.stanford.edu/class/ed260/freedman549.pdf for a longer explanation of the Ecological Fallacy). We already recognize the Ecological Fallacy for stats such as RBIs and Rs, both of which correlate very highly with runs on a team level, but not nec. with "runs created" on an individual level. Less appreciated, it seems, is that OPS etc. are subject to the same logical problem.

Just because the inference is a logical fallacy doesn't mean it's not true; but it does mean it's not necessarily true. OPS, RC, etc. are not necessarily the best measure of true "runs created". Whether they are may very well vary by individual.

Concretely, I think the value of walks (and therefore of OPS, EquA, etc.) is exaggerated for sluggers such as Bonds, because I think their walks systematically occur in situations when they are less valuable than average. If that is the case, those guys' walks (OPS, EquA, etc.) create fewer runs than the aggregate correlation of team OPS etc. to team runs would suggest.

I'm not saying that walks and OPS are not important. I'm just saying they may not be quite as important as we think.

I'm interested in y'all's thoughts on this. Am I overlooking evidence for OPS etc. other than aggregate correlations? Does anyone know of research on this that's been done before? If not, anyone have any ideas on how to study it?

[Duque]

This is an interesting point - walks do not occur in a 'vacuum' - so to speak. Obviously, if you are Barry Bonds and you have JT Snow hitting behind you you'll get walked a lot more than if you have Jason Giambi behind you. They're also impacted by the inning, the closeness of a game, and the runners on base.

[SmedIndy]

Actually, the best measure of true OPS is OBP*1.4 + SLG, or something close to that.

The premise is getting on base is the most important skill you bring to the baseball field.

There are other measures, for sure, like Runs Created, RCAA, but if you only have one to bring, bring OPS, IMHO. It's a lot better than BA, RBI, or whatever else and easily calculatable.

We had a long, long discussion on this before as well. However, if the Giants hitters after Bonds were truly doing their job, he would have scored 150 runs or more last year.

And if you walk Barry everytime, it will burn you in the end.

[calexpat]

Originally posted by SmedIndy
Actually, the best measure of true OPS is OBP*1.4 + SLG, or something close to that.

Based on correlations between team OBPs and SLG, and team runs scored.

Can someone explain what walking Barry Bonds every time has to do with whether or not OPS is the best way to measure offense?

if he walked every time his OPS would be only 1.000, lower than it is now, by the by.

[SmedIndy]

Using the true measure of OPS (the 1.4 figure) and a discussion of a similar phenomenon in the new Historical Baseball Abstract regarding Ruth, it would be quite foolish to walk Barry every time up.

The On-base + Slugging version of OPS is used more widely because it's a decent approximation and easy to compute.

i still don't get why should you walk barry bonds every time up is relevant to the "ecological fallacy"

isn't the point of walking him to keep him from hitting 3-run homers?

[Ytown Tribe fan]

We've done this before, but the short version is this:

If Barry walked EVERY time up, the Giants would score at least 80 more runs than if he were allowed to hit his normals.

You only get 3 outs each inning. A man who gets on base and NEVER makes an out is more valuable than a guy who hits .400 with 80 homers. Any day.

Just think about Barry drawing 650-700 walks a year and making NO outs. There is no way he could possibly be mor valuable even hitting .400 with 500 total bases, numbers that no one has ever achieved.

Excuse me for being silly but I don't see why you folks are carrying on such.

No one here suggested that Bonds should be walked every time up, and he is never going to go a whole season without making an out. i could be wrong but i don't think anyone has ever gone the whole season without making an out.

I keep going back to the original post and don't see how any of these responses are related. Cal's talking about "walks systematically occur in situations when they are less valuable than average."

[calexpat]

Good question. The basic answer is that this is not about a hitter walking every time, which has never happened in the history of baseball. It's about a hitter walking disproportionately often in high impact situations. All else equal, (ie, BA and SLG), this will result in a higher OPS for the slugger than if he were not walked disproportionately often in those situations, but it may result in fewer runs for the team, if the slugger is sufficiently better than the guy hitting after him. If that is the case, then OPS overstates the slugger's offensive value to the team.

In valuing OPS and other similar stats as proxies for "runs created" to the point of ignoring direct measures of runs created such as RBIs and runs, we implicitly assume that hitting outcomes are randomly distributed among at-bats and situations--or at least, distributed the same for individuals as they are for teams. Research has shown this assumption to be true, in general, for hitting outcomes other than walks. It is not true for walks. Good hitters such as Bonds are walked disproportionately often with runners on base, with runners in scoring position, with two outs, etc. If we don't allow for this in interpreting the OPS of sluggers such as Bonds, we are guilty of the ecological fallacy.

Check out Bonds's splits by situation for 1999-2001. (I haven't figured out how to insert tables, so bear with me.)

With nobody on, he hit .282 with a .410 OBP, a .745 SLG, 96 home runs in 901 PAs for a home run rate of .107 HR/PA, 151 BB for a walk rate of .168 BB/PA, and an OPS of 1.155. With runners in scoring position, Bonds hit .307 with with a .530 OBP, a .686 SLG, 27 HR in 426 PAs for a home run rate of .063, 143 BB for a walk rate of .336 BB/PA, and an OPS of 1.216. You get similar results if you look at runners on base.

I don't have the numbers of the guys hitting after Bonds, but it seems pretty clear to me that if Bonds had hit, hit home runs and been walked at the same rates with runners on scoring position as he was with nobody on, the Giants would have scored more runs. This is despite the fact that Bonds had a higher OPS with RISP.

I am not arguing that BA, RBIs, or Rs are better single measures of offensive value than OPS. I am arguing that they are not meaningless, and that OPS is not perfect.

[tenkevcardinal]

I wish they had some sort of statistics like total bases advanced which would be the total bases you advanced runners. Like if a guy is at first and you hit a double and the runner scores you get three TBA and 2 TB. It wouldn't be that hard to keep track of and it would be of great service. Then we wouldn't have to say whether or not his walks were effective, we could just see something like TBA divided by PA w\ risp.

[calexpat]

Originally posted by Sheila
Excuse me for being silly but I don't see why you folks are carrying on such.

Welcome to netshrine, Sheila. We carry on a lot here. Good questions.

[SmedIndy]

Pat -

If OPS was perfect, we wouldn't need any other measure.

It is not, and that's why we while away the hours debating things like this

The beauty of baseball is that there is no one single accepted measure you can provide to comprehensively get the true value of a player.

One that has promise is Win Shares, but the formulas are so complicated you couldn't do them in your head if you wanted to.

[VNV Nation]

Smed: Well said (that rhymes).

The new runs created method is incredibly complicated. The old one was easy, but suffered seriously from the Ecological Fallacy, as calexpat would put it. So much so that, if you don't use the new formula, you might as well use something else. Linear Weights or one of the BP formulas. and they aren't easy either.

[KCBOOMER]

I think we should all run out and find a statistical that has been created a hundred years worth of data, split a hair, and then bring it back for discussion.

OPS (or more accurately OBP*1.4+ SLG) has shown itself to be an effective tool. Not an omnipotent but effective.

You keep saying "nobody walks everytime" and you keep missing the argument. You want to take a single case of "I walked him and got away with it" into a general case. Take the situation in which you want to walk Bonds and put a decent hitter behind him and run the numbers a couple thousands times and see how it plays out.

The intentional base on balls has it's place in the game, but notice when it is used. It's either the National League with two outs, second and third occupied (first base open) and the eighth place hitter coming up or late in the game.

The late in the game scenario usually is a "game" situation where it doesn't matter how many runs the opposition scores so let's face the weaker hitter.