Thursday, August 06, 2009

Evidence on whether teams "own" other teams

Here's some new evidence on momentum, as well as on whether teams "own" certain other teams.

For instance: as of right now, the Yankees are 0-8 against the Red Sox this season. Does that mean they should be expected to continue losing to Boston, at least more than you'd expect from the teams' relative talent levels?

Apparently not. "Numbers Guy" Carl Bialik crunched the numbers (disclosure: I produced the raw data for him from Retrosheet game logs), and found that, when one team starts out the season 8-0 against the other:

-- the team with the 8 wins went about .530 that season against other teams.

-- the team with the 0 wins went about .450 that season against other teams.

-- in all remaining games that season between the two teams (of which there were 545 total), the "8" team went about .600 against the "0" team.

What does that mean? Well, the .530 team probably played about .560 for the season when you include the missing games (the eight consecutive games it won, plus the additional games against that team where it went .600). The .450 team probably went around .420.

Regressing to the mean a bit, the .560 team is probably truly .545 or something. The .420 team is probably around .440.

How often will a .545 team beat a .440 team? I'm guessing about 61% of the time -- pretty close to the 60% observed.

Again, my calculations are only as good as my estimates, and don't take all factors into account (for instance: you'd expect the 8-0 team to have had an above-average number of home games out of those eight, since they wound up winning them all. That means you'd expect more road games in the remaining head-to-head matchups, which should reduce the .610 estimate a bit). Still, I'm confident it's all close enough that if you studied the issue in more detail, the results wouldn't be much different.

Conclusion: no evidence of streakiness, momentum, or "owning" another team.

Labels: ,


At Thursday, August 06, 2009 5:37:00 PM, Blogger Phil Birnbaum said...

An interesting tidbit: for all games between pairs of teams since 1956 where one team started out 8-0 against the other, the team first in alphabetical order by city name won 695 out of 1281 games (.543).

The games aren't random and independent -- teams have some tendency to stay good or bad from year to year. But if they *were* random, that would be 3 standard deviations from .500.

So maybe the good teams tend to be the ones early in the alphabet (Baltimore in the 70s, Atlanta in the90s).


Post a Comment

<< Home