THE BOOK--Playing The Percentages In Baseball

Might using Pythag instead of W-L record change the results?  It seems like that might knock the contenders down a bit, and get the non-contenders closer to .500, since teams with good W-L records are more likely to have outperformed their Pythag than teams with bad ones (I think?).

Don’t ever use a non .500 w-l record from one time period and expect it to be the same in another time period, as comment 1 is suggesting!  Both w-l records will regress toward .500, so that the expected w-l record when thez plaz one another is going to be much closer to .500!  Using pythag w-l record will help, but will not solve the problem!

Assuming the basis of studying such a question is that noncontending teams will dog it & throw in the towel in September games while contending teams will be motivated to give their best performance, I would suggest the August time period might be a better time to study such a question.  “Motivation” by non-contending teams may actually be higher in September when the young calls-up come up & the journeymen who have most likely underperformed up to this point may start feeling a little heat from the youngsters and realize that their jobs for next year are not locked in stone.  Also, if motivation is a problem, I would suggest it will be lower in the summer heat of August than the cooler temperatures of September.

A.  Why is it that you hear all the time from the TV commentators that, “It is dangerous for a contending team to play a non-contending team because the latter have nothing to lose?”

Which is it?  They have nothing to lose so they play loose and better, or they have nothing to gain so they play with no motivation and therefore worse?  The answer of course is that it is neither as far as we know and we couldn’t tell anyway even if it were one or the other, because as usual these are professional athletes playing for a current or potential millions of dollars and therefore by and large play as hard as they can more or less most if not all of the time.  And if some tend to play a little harder under these circumstances and some tend to play a little less hard, and others play around the same (which is most likely the case), then the whole net is going to be more or less a wash.

B.  Motivation or no motivation or anything in between, you are NOT going to find an “answer” in 200 or even 1000 games, especially if you don’t control for anything and everything under the sun, such as exact schedules, exact personnel, etc.

And again, if you don’t regress each team’s prior w/l record to some mean, your results will be meaningless.

A team that is .554 before September or August is a true .530 team or whatever, and a .394 team is a true .425 team or so.  So if you want to predict their expected wl% when playing each other for the remainder of the season, you have to do a log5 using those true wl% and not the sample ones.

And whatever you get after 257 games, even with a confidence interval of 1 sigma, you are plus or minus .035 or something like that and even more when you are comparing or combining a bunch of samples.  So, even a .661 wp is .661 plus or minus .070 at 2 sigma, which is .590 to .730 which is way too large an interval to get any meaningful results whatsoever.  What if the expected wl% is .630? .700?  .595?  .710?  You are forced to conclude the same thing which is that there is not enough evidence to conclude that any class of teams play more or less motivated than expected.

Sorry tango, but this is a bad post.  To say that the effect is “almost certainly not true” based on this study is ludicrous.

Even a layman like me could immediately see the selection bias and the small sample size problems with the study.  Also it looks like it includes games where a team has their playoff seeding locked as long as they’re within 5 games of another team.  Those teams have the lowest incentive of all to try and win and would skew the results

So, the idea that the contenders have it better against non-contenders because non-contenders are putting out “trial” lineups is almost certainly not true.  Even if they are trial lineups, those trial lineups might have been the better option to begin with!

I have to agree with #5 above.

Why did you say that?  First of all, the contending teams outperformed their expected win rate .661 to .665.  I think maybe you misread something in the “article.”

Most of all, there is not a shred of evidence one way or another as has been explained in #2 and #4 above.

That should be .655 and not .665 of course.

"Most of all, there is not a shred of evidence one way or another as has been explained in #2 and #4 above. “

How can you say that?

We know that contending teams have incentive to leverage their future for their present, and we know that crappy teams have incentive to leverage their present for their future.  We know that occasionally Carlos Beltran goes from a crappy team to a contender mid-way through a season.  It also seems extremely likely to me that crappy teams tend to shut down their best pitchers more often than contending teams do in September, and work the ones left standing less hard.

Brute forcing your way through the data will tell you how big the effect is, or if there is an unintuitive counter-effect, but to me it doesn’t seem like a necessary step in order to assume that the effect is there at all