THE BOOK--Playing The Percentages In Baseball

can one of you guys correlate OBP to the time of the game?

Quick look at RetroSheet data and a duration model for 2000-2009 returns the following effects (all significant but with 24,000 observations, some are small):

RUNS: -0.3 (minutes per additional run)
PITCHER: 4.3 (minutes per additional pitching change)
PA: 1.6 (minutes per additional plate appearance)
AL: 0.4 (minutes for home team = AL)
BOS: 6.0 (minutes for a game with Boston)
NYY: 11.6 (minutes for a game with NY Yankees)
BOTH: 4.3 (minutes for a game with both Boston & NYY)

I also control for year effects. I can’t find data in the game logs on #pitches, and I wasn’t interested in messing with play-by-play data.

But it looks like games played by the Sox are about 6 minutes longer than they “should” be, games played by the Yankees are about 11.6 minutes longer than they “should” be, and games with both are almost 22 minutes longer (6 + 11.6 + 4.3) than they “should” be.

JP: fantastic work!

I don’t think you have to go to the number of pitches, if you have number of batters.  You’re not going to find much difference there. 

So, if each extra batter is 1.6 minutes (96 seconds), then each extra pitch is 26 seconds.  Let’s round it off to 30 seconds.

There’s almost 300 pitches per game, so that’s 150 minutes based on number of pitches thrown.

Add to that the time between innings (you’d want to add that in as a parameter as well).

The runs thing looks wrong, if only because of the interdependency of that to PA.  I’d include HR, because a HR clears the bases without throws, relays, slides, etc.

But, yeah, 22 minutes simply because BOS faces NYY, after controlling for number of batters faced?  That’s bad.

Two thoughts on the effect of runs. First, the Runs effect might actually be a margin of victory effect. Games with more runs will, on average, have a larger margin of victory, which will reduce the number of pitching changes late in the game.

Second, plate appearances with runners on base take longer. Controlling for plate appearances, games with more runs will have fewer plate appearances with runners on base, and therefore shorter games.

Comment #7 on Posnanski’s site:

“How much time do Fox and ESPN games add? Isn’t it about 30 second per break, or at least 18 minutes? And therefore happens more often in Yankees-Red Sox matchups.

As a fan, I never understood complaints about the length of a game since you’re getting more for your money. Terrible games and great games can be any length, but I’d prefer to be at the ballpark as long as I can.”

#Runs -0.17
#Pitchers 3.87
#PAs 1.97
#HRs -1.00
#Innings -3.03
Margin of Victory -0.89
AL Dummy 0.11
BOS Dummy 6.01
NYY Dummy 11.86
BOS*NYY Dummy 4.20

I think it is difficult to look at things like Runs, PAs, HRs, Innings, etc., all in the same model and draw too many conclusions about them. For example, the -3.03 coefficient on Innings just seems absurd, but the reality is that an extra full inning adds at least 6 PAs. So the six PAs add almost 12 minutes and the extra inning subtracts 3 for a net of 9 minutes. Let’s be clear that, omitting other variables, Runs, #Pitchers, #PAs, HRs, Innings all increase duration. But the effects are confused with one another. A study attempting to gauge the actual effect of a home run (versus another type of hit) on duration would need to be more careful with this. But as a model that attempts to clear up the “Boston-NYY games take longer because those teams just score more runs” complaint, I think throwing in the kitchen sink is a fine approach.

To me the dominant takeaway is that the number of pitching changes and the number of plate appearances are the core drivers of duration, or at least they appear to be. Scoring faster (i.e. HRs) or winning by more (i.e. Margin) both drive down time, which makes sense. To Ken’s point, PAs with runners on base always take longer (throws to first, more likely to have coaches/catchers come out).

Again, though, the only REAL point of this model was to test the BOS and NYY dummies plus the interaction. The same results are still there—a Sox v. Yankees game in the 2000s was 22 minutes longer than we would have expected based on the score, the number of plate appearances and innings, the count of pitching changes, and the margin of victory.

Oh, and for what it is worth, here’s the breakdown by 2000-04 and 2005-09:

BOS: 6.01 overall, 1.4 00-04, 10.6 05-09
NYY: 11.86 overall, 10.4 00-04, 13.2 05-09
BOS-NYY: 4.20 overall, 6.7 00-04, 1.5 05-09

So it looks like the Yankees drew things out much more than Boston in the first half of the decade (especially once you chalk up the entirety of the 1.4 extra minutes for Boston to Nomar). But their games together were extra long in 00-04. In the second half of the decade, Boston was as “bad” as the Yankees were in the first half of the decade, and the Yankees got “worse”. But their joint games were only slightly longer than expected (again, though, recognize that the games were already 10.6+13.2 minutes longer because they were BOS and NYY games).

To stretch the data way beyond where I should…
1) Boston might have learned the whole “slow the game down to a crawl” thing from the Yankees.
2) Both teams have largely implemented this slow down in all of their games (driving up their main effects and diminishing the interaction effect).

ok, since it’s the Yanks and Red Sox are doing this (and I say this as a Yanks fan), and since they are the two dominant clubs of the present moment, it is an ABSOLUTE NO-BRAINER for MLB to issue a memo to their umps about enforcing certain rules.

what we have is two clubs bending the rules, getting away with it, and, at least arguably, gaining an advantage from doing so--at the expense of the fans and every other team. 

the gamesmanship gets dull; all dull aspects of baseball do not merit traditionalist defenses.  MLB can easily make it go away, and gain an easy PR triumph as well.  so JUST DO IT.  limit time outs, step offs, and mound visits.  award penalty balls/strikes.  do it.  please.

I posted some numbers here comparing the Yankees-Red Sox and Mariners-Rangers games from 2008-2009, based on the detailed pitch time data.

http://www.hardballtimes.com/main/blog_article/red-sox-yankees-slow-games/

Mike, if you have time, please try this:
take fast pitcher, Buehrle comes to mind, and check differences in his starts against Bos/NY vs against rest of AL, or, maybe, 2-3 slowest teams outside AL East. I wonder if Bos/NY hitters slow down all the pitchers, or if they only slow down those pitchers who don’t pitch fast (like a game between hitter and pitcher, where hitter will call time and step outside the box if pitcher is leaving him standing there for a long time).