I started this little excursion to try and figure out why baseball has such a small home field advantage. I had a theory that I could explain it through baseball’s differential travel behavior. That hypothesis failed pretty miserably. But I think I learned a fair bit nonetheless, and I’ll take a few paragraphs to sum it up.

  1. The home field advantage is about 4 percent. I know, this isn’t exactly startling, but it’s been roughly constant for 50 years, or at least as constant as we can glean from the couple of thousand games played per year. When you think of all the other things that have changed in baseball, that’s at least mildly interesting.
  2. About half of that is the bats-last advantage. At least, that’s the portion that allows a tied team late in a game to win differentially. Of all games, about 29 percent are tied at some point around or after the end of the fifth inning.
  3. While travel doesn’t show up as an effect for teams just showing up in town, the cumulative effect of road trips and home stands does matter. Every extra day of a road trip costs the team about half a percent of winning probability on the road trip, and every extra day of a home stand gains about half a percent. However, though this effect is statistically significant, it’s really quite small. If a .500 team had 10 six game road trips a year, then there is only a 2 percent difference between ten 3-3 road trips and nine 3-3 road trips with one 2-4 road trip. So this extra road trip effect is probably not much more than one game a year.
  4. The park advantage is clearly there in that some teams have much higher home-road splits than others, particularly Colorado. That said, even in “odd” parks, it is clearly not easy to leverage that advantage, since almost every team has years of negative splits, and the variance of home-road splits over time is huge. To some extent this might be a management choice: one might sacrifice some measure of fitness at home to make the club better overall and do so optimally. Maximizing the home-road split might make a team so dysfunctional on the road that it’s not worth it. The park advantage (or at least home-road splits) is completely unrelated to how good or bad a team is.

So why does baseball have a 54 percent home advantage and professional basketball have a 63 percent home advantage? Since there is no strategic advantage to being home in basketball, this means that there is at least a 10 percentage point non-strategic gap.

The obvious possibilities are referee bias and its possibly correlated variable, crowd effects. I have always been suspicious of pure crowd effects for the simple reason that professional athletes are exactly those set of people who have inured themselves to such effects. I have always felt that the reason crowd effects are often felt to be so important are: (a) fans know how much they’d be affected by crowds and project that onto the players; and (b) management of teams do nothing to disabuse people of that notion because it helps ticket sales.

Note, by the way, that home field advantages were much higher in the early days of baseball, when crowds were much smaller than they are today. I think there are too many other changes to make much of this but I’m just being honest about where my prejudices lie.

That leaves referee bias — which, I hasten to add, is almost surely subconscious. Many commenters (and a big thanks to all of you — it made this more fun than scary) point to this effect, particularly in soccer and basketball. We should take this as a sign that, Eric Gregg notwithstanding, baseball ought to be proud of the objectivity that the umpires seem to bring to their work. Even incompetent umps like CB Bucknor are just random idiocy generators.

So that’s what I think I learned, but YMMV. I want to thank Alex for agreeing to fill the seasonal interregnum with this experiment and thank all of the commenters here on Braves Journal for such excellent words of encouragement and advice. You guys really helped.

In season I hope to be able to use the Retrosheet database to answer a relevant question here and there, but studies like this are strictly hot stove stuff. Thanks for bearing with me.
Oh, and for mravery, here’s the categorical regression he asked for: