THE BOOK--Playing The Percentages In Baseball

I haven’t read it yet, but couldn’t the familiarity come from actually living and being in a city you are more used to?  Not the park.  I’m not sure how you would do a study on living situation.  Maybe the first month a player is with a new team, which would mean new living quarters/new city.

Anyways look forward to reading it once I get a chance.

Thanks again for posting this Tango!

Sean that’s a good point, and reminds me of the rumor we hear that rookies perform better at home because they’re up all night partying in new cities at their away games.

To examine this apart from stadium familiarity, I guess you’d want to look at teams that switched stadiums mid-season.  That way, the city would be the same but the field, locker rooms, etc, would be different.  Presumably, if one’s own bed is what matters, you’d see no difference in homefield advantage after the midseason switch.

These midseason, same-city stadium changes were excluded from my study because I wanted a full season of data in both the “before” and “after” seasons.  And, back in 2002 when I began this thesis, we didn’t have nearly the kind of data available that we have today.

Someone did a study, although I can’t seem to find it now, that did show teams in their first year in a new stadium had a slightly lower home field advantage than expected.

I did my own study looking at component factors, and found that almost all of the home team’s advantage was in triples made and prevented.  As triples have decreased, home field advantage in baseball has decreased.

My guess would be that familiarity plays a part in making and preventing triples, as the home outfielders are more experience playing the ‘unusual’ bounces, etc.

My guess would be that familiarity plays a part in making and preventing triples, as the home outfielders are more experience playing the ‘unusual’ bounces, etc.

There is no question that that is the case. There is also no question that teams with unusual home field advantages are more familiar with a quirk of the park (Monster in Fenway, roof in Metrodome, the
bad lighting in the Astrodome, etc.).

Ergo, there is no question that familiarity with the field is part of a home field advantage.

I don’t know why the author got the result he did. For one thing, looking at w/l records is a rather coarse way to measure HFA.  Unless you have a huge sample size, you are going to have problems finding any small yet significant differences.  I would definitely look at RS and RA or even components.  For example, the triples home/road ratio for the two groups (last year in stadium/first year in stadium) would be illustrative.

For another thing, it may be that familiarity does not take very long to develop such that if you look at one whole year in a new stadium you aren’t going to find anything.  It might be that the home team is plenty familiar with the new stadium after the first couple of weeks or even while working out in the stadium before the season starts!

There is some evidence that that might be the case (that becoming familiar with a stadium does not take that long).  The first game in a series has the most significant HFA and the HFA decreases after that as the series goes on.  It is something like an extra .006 for the home team in the first game and an extra .004 for the road team in the third and 4th games.

When I do component HFA, it is the walk and K numbers, in addition to triples.  The walk and K probably relates to visibility.

KJOK, I’d love to see that if you can find it.  If I remember correctly, I tried cutting the data a bunch of ways just for kicks.  Just messing around with the inclusion/exclusion criteria.  And there was one way in which I found Football, I think, to show some loss in HFA in the first year in a new stadium.  But, when you do 20 t-tests at .05, you’re going to get one that appears significant but isn’t.

MGL… Notice the paper says “submitted to the Wesleyan Psych Dept"… not “submitted to SABR”, or “submitted to the Boston Red Sox”.  I was just thrilled that I found a professor cool enough to work with me on a fun study, rather than another 200-page thesis on depression or anorexia.

It was 2002-2003 when I wrote it.  Moneyball wasn’t out yet, and I knew NOTHING about baseball analysis.  I don’t know if at that time I’d have been able to get my hands on game-by-game or play-by-play data (for free… remember I’m a 21 year old college student).  And I sure as heck didn’t know any way to parse the data if I did manage to get it.  I think the database I found had midseason splits for each sport, and that was about as much granularity as I could get.

As for sample size… my understanding of stat testing is that you can test for significance with any sample size.  You just have to find a much larger effect with a small sample size to find it statistically significant.

The triples and first game of a series issues are real interesting.  Thanks for sharing.  I’d guess with a new stadium generally comes a new payroll, so teams may have gotten objectively better in the transitions too.  There’s definitely a lot of confounds.

I wonder how much you could break it down for other sports besides baseball.  There’s no real equivalent to triples, unfortunately.  I think basketball only plays 1-game “series’” at a time, right?  Likewise maybe with hockey?  Maybe you’d have to look at scoring differential by quarter or period, to see if HFA grows or wanes over time?  If that’s even an issue, because basketball and hockey surfaces are much more uniform across teams.

I wonder if the higher-HFA-in-the-first-game has to do with time zones and travel fatigue more than becoming familiar with the ballpark itself…

The first game disadvantage could be due to travel, but at least one person looked at travel versus performance and didn’t find much.  My guess is that the “series” thing is about familiarity.

Wasn’t trying to criticize the study at all.  Didn’t even read it.  Was just commenting on the issue in general.

The only travel studies I’ve seen have looked at entire series’ at once.  Which I’d thought made sense, but now knowing that the HFA is stronger in the first game makes me question that.

Sounds to me like ideally there’d be one that compares, say, the first game in a series HFA of same-time-zone games to the first game in a series HFA of coast-to-coast games.

And of course, if there really is an issue with rookies living it up in other cities, you’d see a stronger HFA in NY and LA, I’d think, versus Cleveland and Minneapolis.  Which would be hilarious, if there was an effect found.  And perhaps an argument for a team in Las Vegas.  At least they could count on a strong home field advantage!

Related study:
sabermetricresearch.blogspot.com/2008/06/in-last-two-posts-i-reviewed-reports-of.html

Nice!  Exactly what I was envisioning.

The above-referenced work by Phil is interesting.  Despite not being “statistically significant” (the ole, and silly, magic “significant if it is more than 2 SD, but not if it is less than 2 SD"), the results are interesting and definitely suggest that travel may have some effect on the first game of a series, which would certainly make some sense.

Let me update the study (it’s 10 years old now) and add words/article to the data, and I’ll put it up on my website at http://www.seamheads.com in a few days.

Just to follow up, I didn’t forget about the updated study - it’s been updated thru 2007 data.  However, with various playoff and world series posts scheduled, I’m not exactly sure when my article will be posted, but it’s likely to be one of the off-days during the world series.

FYI, the article is up at:

http://www.seamheads.com