August 25, 2009
Ahead in the Count
Home-Field Advantages, Part Three
In trying to understand home-field advantage, we have asked what home-field advantage actually makes a team do better, and we have asked who has the biggest home-field advantage. The first article of this series answered the question of what home-field advantage actually makes you do better-everything, as home teams do better on walks, strikeouts, balls in play, and errors. They are better at pitching, hitting, baserunning, and defense, and all aspects of their games seem to improve. The second article of this series showed that most teams have pretty much the same size home-field advantage, with the exception of the Rockies. Even though natural luck can make a team look like they are particularly good or bad at home, the 29 non-Rockies teams are pretty much right around eight percent home-field advantage, plus or minus a little statistical noise.
However, that does not mean that every game has about an eight percent home-field advantage. In other words, if we know that the Yankees would beat the Dodgers in 52 percent of games they played against each other, we cannot infer that the Yankees would beat the Dodgers in 56 percent of the games at Yankee Stadium and the Dodgers would beat the Yankees in 48 percent of the games at Dodger Stadium. Certain types of games have a different degree of home-field advantage.
I divided the 27,613 games that did not result in a tie from 1998-2008 into a variety of subsets to evaluate the different magnitudes of home-field advantage. This is important to study because if we can understand where home-field advantage exists, we can better understand why home-field advantage exists and how it comes to be.
Home-field advantage is commonly attributed to the rigors of travel and familiarity with the ballpark. Therefore, I compared games played among teams within the same division to games played among teams in different divisions, since intradivision matchups occur more frequently and often require less travel. Home teams win just over 53.3 percent of intradivision games, which means that they have a 6.7 percent home-field advantage. In contrast, interdivision games showed an 8.7 percent home-field advantage. Although this seems like a large gap, it remains a little shy of statistical significance over the 1998-2008 period. The problem with statistical significance tests for this type of analysis is that to find anything conclusive among minor differences, we would need to look back over many years to be able to pinch the variance enough that home-field advantage differences were statistically significant. Doing so would risk losing meaning, by dipping into different eras where effects may be different. Instead, it seems safe to say that since the confidence interval of intradivision games (52.4-54.3 percent) and the confidence interval for interdivision games (53.6-55.1 percent) barely overlap, intradivision games probably involve smaller home-field advantages. This may be either because road teams are more familiar with the stadiums of their most frequent opponents, or because road teams suffer less when they travel shorter distances. There will be some evidence of both of these factors in what follows.
I also separated the home-field advantage for interleague games and intraleague games as well, in addition to interdivision games within the same league. Interleague games demonstrate a 10.1 percent home-field advantage, far larger than intra-league games, which show a 7.4 percent home-field advantage, although still not quite statistically significant. Specifically, interdivision games within the same league have a home-field advantage of 8.4 percent. This indicates that both distance effects and familiarity effects may be in play.
In summary, this means that we have the following home-field advantages:
Same Division, Same League: 6.68%
Of course, interleague games may also have an extra home-field advantage, since the DH is only in place in AL stadiums (giving an added advantage to AL teams who construct their roster with this rule in mind), and no DH rule in NL stadiums (giving an added disadvantage to AL teams who construct their roster with a DH). Therefore, it is worth looking at different kinds of interleague games to determine the magnitude of this effect.
The first thing that I did to parse the set of interleague games was to compare the home-field advantage in interleague games among teams in the "equivalent" division in the opposite league (i.e. NL Central vs. AL Central) to interleague games among teams in different divisions. The results came out rather clean-home-field advantage in the equivalent division interleague games was only 7.95 percent, and home-field advantage among teams in different divisions was a whopping 13.42 percent. It certainly seems that familiarity may be in play, since the home-field advantage is smaller for same-division games within the same league (6.7 percent) than it is in equivalent division games in different leagues (7.9 percent), and it is also smaller for interdivision games within the same league (8.4 percent) than it is interleague, different-division games (13.4 percent).
Justification of the theory that familiarity is playing a role in home-field advantage comes from the fact that when I looked at "interleague rivals," the home-field advantage came out very small. Since interleague rivalries have changed over time, it was not easy to define who was encompassed by this term. To determine who these rivalries were, I simply found the 14 NL teams that had played the 14 AL teams the most frequently (Angels/Dodgers, White Sox/Cubs, Rays/Marlins, Indians/Reds, Rangers/Astros, Mets/Yankees, A's/Giants, Royals/Cardinals, Mariners/Padres, Twins/Brewers, Blue Jays/Nationals, Red Sox/Braves, Tigers/Pirates, and Orioles/Phillies). In those games, the home-field advantage is only 4.4 percent, even less than among teams in the exact same division. It certainly appears that familiarity is playing a role, since these teams play each other more frequently than other interleague games, even those within the same division, although it could also be attributed to distance.
Strengthening the argument that this might be the effect of familiarity rather than distance is that if you look at the three pairs of teams who play in the same metropolitan area (Yankees/Mets, Dodgers/Angels, and Cubs/White Sox), those matchups have a 15.8 percent home-field advantage, though in a sample of just 190 games. This also highlights another important fact, which is that crowd support is unlikely to be the primary cause of home-field advantage, since one would expect home-field advantage would smaller in a mixed-fan crowd. The "same city" argument for familiarity over distance is not very conclusive, as the confidence interval for those types of games is really large, (50.9-64.9 percent), but it does seem to point to distance or familiarity being a factor rather than crowd effects, since even this large interval rules out the possibility that there is no home-field advantage in the same metropolitan area.
As these "interleague rival" teams do not play each other any more frequently than teams in different divisions in the same league, it seems that distance may be playing a role too, since the home-field advantage is almost twice as large (8.4 percent) in those games. Another argument for the effect of distance on home-field advantage is that games between teams in the same division who travel further distances show a larger home-field advantage than teams who do not travel as far. To define "further" I had to resort to a small trick-rather than determine the mileage between each of the 450 pairs of major league cities, I gathered the degrees latitude and longitude for all 30 teams, and then used the Pythagorean Theorem (the one by the guy from ancient Greece, not the guy from Kansas) on the difference in latitude degrees and longitude degrees to figure out how far apart two cities were in degrees. The average distance in degrees for teams within the same division was 10, which is the distance between Seattle and San Francisco. Games among teams within the same division who were less than 10 degrees apart showed a home-field advantage of 5.5 percent, and games within teams who traveled further than 10 degrees was 8.1 percent. That is nearly as large as the home-field advantage among teams in different divisions of the same league (8.4 percent).
So, we have learned from this look at where home-field advantage exists that both travel distance and stadium familiarity are probably significant causes of home-field advantage. We found clear effects of familiarity since interleague rivals do better (even when geographically distant) than interleague teams in equivalent divisions, and since same city interleague rival games show a larger home-field advantage than other interleague rival games. We also found clear effects of distance, such as the fact that home teams within the same division win more frequently when the away teams have traveled further.
However, the differences in home-field advantage remain rather small. The difference between home-field advantage among intraleague games within the same division and between different divisions is less than two percent. Ignoring this effect and assuming that all games have an eight percent home-field advantage will only lead you to guess wrong on a couple of games more each year than you would if you had known this fact. Nonetheless, it gives us yet another insight into home-field advantage. Since travel seems to be playing a large part in home-field advantage, next week we will look at home-field advantage in different games within a series, as well as different lengths of series. We will also consider whether offdays and consecutive series at home or away play a role in home-field advantage.