clock menu more-arrow no yes

Filed under:

HOMEFIELD ADVANTAGE

New, 14 comments

 

            For many years, it has been well documented that home teams win baseball games more often than road teams.  In fact, from 1977-2008, home teams have won 54.0% of all baseball games—a .080 percentage point advantage.  Some teams have larger advantages than others, as documented by Nate Silver of Baseball Prospectus in a recent article (http://baseballprospectus.com/article.php?articleid=7732), and looking through this data, the variance between teams’ winning percentages is large (by my calculations, the standard deviation of .046).  There are a number of reasons that home teams could have an advantage in games, and it may be that a larger than usual discrepancy between winning percentage at home and winning percentage on the road could be an indicator of either a team’s skill at taking advantage of homefield advantage or eliminating their opponents’ skill at taking advantage of their own.  It could also be a random correlation that is not based on any particularly characteristic of a team, but instead on other factors altogether.  This area of baseball research is a significant one-- being able to understand what is behind homefield advantage has the potential to help teams in both team building and individual game preparation.

            The first question that we need to check is if certain teams are more prone to homefield advantage by chance or by luck.  To check this, I computed the correlation between a team’s home field advantage in different pairs of years (as measured by the difference in winning percentage at home and on the road).  I found the following correlations.

 

YRS.

’07&’06

’07&’05

’07&’04

’06&’05

’06&’04

’05&’04

Corr.

.196

.072

.162

-.060

-.190

-.060

 

            It appears as though homefield advantage is not a persistent thing—the Phillies have not squandered homefield advantage and the Rays have not capitalized on it for any specific reason and there is no reason to expect that the Phillies will have a weaker homefield advantage next year than the Rays.

            So what does cause homefield advantage and can teams learn to exploit it—or keep other teams from exploiting their own?  For one thing, the correlation in between winning percentage and homefield advantage is strongly negative, -.190, and significant at the 97% confidence level!  That means that there is only a 3% chance of getting a value so far away from 0 by random chance.  In other words, homefield advantage probably entails a disability on the road more than an advantage at home.

 

WHEN DOES HOMEFIELD ADVANTAGE SET IN?

 

            The first thing that I checked was the odds of the home team winning a tie game at the beginning of each inning.  This helped establish some ideas about the timing of when homefield advantage takes place.

 

Tie Game—inning #

1

2

3

4

5

6

7

8

9

Extras

Home Win

Pct.

.540

.533

.533

.525

.529

.522

.520

.521

.522

.520

 

I have a few observations when looking at this table.  Firstly, homefield advantage declines as the game goes on and it remains tied.  That means that the homefield advantage clearly takes place during the game from the start.  It helps teams establish leads early and acquire them late.  Another thing that is clear is that tie games in the 9th inning are still won by the home team over 52% of the time.  If being able to bat last at the end of the game were not an advantage at all, that number would be much closer to 50%-- instead it is halfway between that and the 54% homefield winning percentage at the beginning of the game.  Therefore, I have to conclude that the opportunity to bat last at the end of the game is an advantage.  Another thing that is worth noting is the huge drop off between the 1st and 2nd inning as compared with any other consecutive innings.  This leads me to believe that there is a first inning advantage.  In Silver’s article, he wonders if National League teams might have larger homefield advantages since they are able to pinch-hit for their pitcher in the bottom of the inning and get one more inning out of him in advance.  However, this table indicates that there is no large drop off in the 6th-8th innings when the pitcher is usually pinch hit for, so this probably does not have much of an effect.

 

WHAT DOES THE HOME TEAM ACTUALLY DO BETTER?

 

            In my opinion, the most important and interesting question is what home teams actually do better at home.  Do they field better?  Do they hit for more power?  As it turns out, the answer is pretty much all of the above and much more.  Using data from 2003 through June 29, 2008, I checked a variety of ratios and found significant differences in many.

 

Statistic

Home

Away

%Difference

Statistical

Significance?

HR/100PA

3.18

3.06

5.75

Strong

K/100PA

16.28

17.26

-5.69

Strong

BB/100PA

8.76

8.20

6.90

Strong

BABIP

.304

.298

2.19

Strong

CG SHO/100 BIP

1.52

0.95

59.84

Strong

ROE/100 BIP

1.43

1.35

5.86

Strong

SB%

72.2

70.8

2.01

Strong

HBP/100PA

0.974

0.934

4.30

Strong

3B/100BIP

0.759

0.640

18.61

Strong

2B/100BIP

6.919

6.853

3.807

Not

Picked off/100 baserunners

0.487

0.518

-6.06

Strong

 

The fact that homeruns, batting average on balls in play, reaching on error, and complete shutouts are significant indicates that hitters, fielders, and pitchers probably receive an advantage from homefield advantage. 

The statistical significance of walks and strikeouts were particularly strong.  This means that homefield advantage works its way into the strike zone, and not simply onto the field.  While it is not particularly surprising that hitters both get more hits on balls in the field of play, and reach on errors more often, it is important to notice that the difference takes place even in the strike zone where the homefield itself is not necessarily part of the homefield advantage.  Some of this could be familiarity with the mound or batter’s box.  However, it may be that umpires are even biased.  The fact that there is an improvement in stolen base percentage at home would lend some credibility to this theory.  However, it may be that even the basepaths are more familiar to baserunners and that could explain some of this as well. The lack of pickoffs by home teams can probably be explained by some of the same logic.

            The statistical significance of triples but not doubles is particularly interesting to me.  A few weeks ago, the Phillies played a home game against the Red Sox in which slowfooted Pat Burrell and Ryan Howard each had triples—initially, the odds of these each happening in the same game given their triples/game for their recent careers seemed to be one in 40,000.  My observation watching the game was that both balls were misplayed by Jacob Ellsbury due to the fact that he did not know where the wall was and how the ball would bounce off of it.  I believe that this anecdote explains a lot of triples—many are misplayed doubles—and it indicates why home teams have 18.61% more triples than road teams.  I believe that the statistical insignificance of the difference in doubles is a result of many balls that would have been doubles in some circumstances turning into triples due to road teams’ unfamiliarity with the field.

            I’m particularly surprised that home teams are more likely to be hit by pitches than road teams.  If anybody has any ideas why that may be, I would love to know what they are. 

            There are a few pieces of data that I would love to check that I was unable to check related to this.  One is groundball BABIP vs. flyball BABIP.  Do both of these exhibit homefield advantage?  This could help establish what types of things are occurring on the field that cause homefield advantage.  Another thing that would be very interesting is to check if umpires make more mistakes on balls and strikes in favor of home teams.  If they do not, it is unlikely they make biased calls about whether a player is safe or out.  If it seems that pitchers have a homefield advantage, then it would be interesting to compare their velocity at home and on the road.

 

INTERLEAGUE AND INTERDIVISION HOMEFIELD ADVANTAGE

 

            Looking through the data above, it’s hard to doubt that homefield advantage affects nearly all aspects of a player’s game.  Whether they are trying to hit for power, average, steal, field, or pitch, baseball players do it better at home.  The question about why still remains somewhat unanswered.  I have split games into a variety of categories and checked homefield advantage in each of them.   Keep in mind that home teams win 54% of games—a winning percentage of approximately .080 more than away teams.  If homefield advantage were exclusively related to location and playing field, then one would expect there to be little difference between homefield advantage in interleague games and non-interleague games.  This is not the case.  From 1998-2008, winning percentages of home teams in interleague games are .101 higher than road teams in interleague games; winning percentages of home teams in non-interleague games are only .077 higher than road teams in non-interleague games.  This is a statistically significant difference.  Part of the reason that interleague games may exhibit larger homefield advantage is the existence of the DH in American League parks, but not in National League parks.  Teams in each league construct their benches differently according to whether there is a DH in their league’s games, and National League pitchers may hit better as well.

            National League home teams have a different environment than American League home teams in that the existence of pitchers (who rarely finish games) in the NL lineups may affect homefield advantage.  Since the inception of the DH in 1973, National League home teams’ winning percentage is .081 higher than that of road teams and American League home teams exceed road teams by .076.  That is notable but not statistically significant.  (For most of this, I am using 1998-2008 data.  The same test with those numbers is similar: .078 homefield advantage in NL games, and .072 homefield advantage in AL games.)

            If we want divide non-interleague games into intra-division and inter-division games, a statistically significant difference emerges: homefield advantage is smaller in division rivalries.  Home winning percentages are only .067 higher than road winning percentages within a division, but are .082 higher in games between teams in different divisions.  This is not a small finding—whatever it is that makes players hit, field, and pitch poorer away from their home turf does not hurt them as severely when they play teams in their own division.  This could be a factor of at least two (and probably more) causes—travel distance and familiarity.  In Silver’s article, he notes that homefield advantage was stronger before World War II when transportation was worse, and especially stronger for teams who played further from the rest of the league.  The discrepancy between homefield advantage in and out of the division hints that some of this effect may be present.  However, players who play in their division rival’s stadium nine teams a year will eventually get very used to the field.  I have little doubt that AL East leftfielders know how to play a ball off the Green Monster better than NL West leftfielders.

            The next thing I did was divide interleague games into those against teams in the same division (i.e. AL West vs. NL West).  Despite the overall .101 homefield advantage since 1998 in interleague games, interleague same-division games had a smaller homefield advantage of only .079.  When I divided this group into teams that had played each other 20 or more times and teams that had played each other less than 20 or more times, the difference was staggering—teams that played each other more frequently had only a .059 homefield advantage and teams that played each other less frequently had only a .092 homefield advantage.  There are two notable comparisons here—one is that the sometimes-manufactured interleague rivalries (Mets/Yankees, Padres/Mariners, Reds/Indians) have smaller homefield advantage than teams that actually play each other more often.  These teams play each other less often than typical division rivalries within the same league, which exhibit a .067 homefield advantage during the same time period.  Certainly, this could be attributed to both shorter travel time (Queens is further from Miami than the Bronx) for the teams, and shorter travel time for the fans.  The closer teams are, the more likely it is that fans will travel to root for the away team and this may have some effect on players.   The other thing that’s worth noting is that the .092 homefield advantage of interleague games of non-"rivals" teams in matching divisions is larger than the .082 homefield advantage of interdivision games in the same league.  This gives strength to the argument that homefield advantage is also based on familiarity of stadiums or the existence of the DH in AL parks.

 

HOMEFIELD ADVANTAGE AT A NEW HOME

 

            If stadium familiarity is so essential in causing homefield advantages, then one might expect to observe a difference when teams change stadiums.  This effect could go in either direction—a team may have less familiarity with its own park and not have as strong an advantage, or opponents may have never played in the park before and this lack of experience could add to the homefield advantage, as the home nine have relatively more experience with their new parks.  Three teams have always played in the same stadium (Rays, Marlins, Diamondbacks), and three teams have played in the same stadium for far longer than any other teams (Yankees, Red Sox, and Cubs).  Due to lack of information in the first case, and biased data in the second case, I removed those from the analysis (though including them had little effect on the results).

 

Year

Homefield Advantage

Average League Homefield Advantage in that year

Last year in old stadium

.087

.073

First year in new stadium

.081

.070

Second year in new stadium

.085

.073

Third year in new stadium

.050

.077

 

These differences are notable, but none of them are statistically significant.  I have to assume that despite the bizarrely small homefield advantage in the third year of a stadium’s existence, there is not much of an effect the third year in.  There was not enough data to establish statistically significance here, so the only cause I can think of is that the lure of a new stadium is smaller after a couple years and fans may have less affect on outcomes after a while, but I find that doubtful.  It is worth nothing that in the year after switching stadiums, homefield advantage is smaller.  If the main reason behind homefield advantage was fan enthusiasm, one might expect homefield advantage to grow significantly upon switching stadiums.  Keep in mind that the last year of a stadium’s existence is not a random sample though—teams frequently move because their stadiums are falling apart and home teams may be able to play funny bounces in crumbling stadiums particularly better than their opponents.

            Overall, the new stadium effect seemed to produce comparatively little in the way of interesting results.

 

MARGIN OF VICTORY AND TOTAL RUNS SCORED

 

            The last thing I checked was which kinds of outcomes are more affected by homefield advantage.  This is tricky to study as there is a huge censoring bias--- as games stop when the home team has scored more in the 9th inning or extra innings, home teams do not have the ability to tack on runs in their final inning.  It is unsurprising that for each of 2003-2008, the Runs Created formula overestimated home team runs scored by more than it did for road team runs scored.  Using games from 1978-2008, the most clearly biased result due to censoring is home teams’ winning percentage in one-run games: .611.  Clearly, if the rules dictated that teams continue to chip away after walk-off hits, presumably some of these one-run home victories would be two-run and three-run victories.  Home winning percentage in two-run, three-run, four-run, five-run, and six-run games were .508, .508, .512, .509, and .510, respectively.

            I then compared home winning percentage in games with larger-than-four-run margin of victories and compared it to four-run-or-less margin victories.  The former was .513 and the latter was .550.  Dividing the data similarly into larger-than-eight run margin victories and eight-or-less produced winning percentages of .519 and .541, respectively.  All of these are statistically significant differences.  It appears clear that homefield advantage has a larger effect on close games than on blowouts, even accounting for the censoring bias that the home team stops scoring when they have won—it is unlikely that home teams would have scored that many more runs in the bottom of the 9th to produce such a discrepancy in winning percentage.

            I also checked whether high-scoring or low-scoring games had a larger homefield advantage. 

 

TOTAL COMBINED RUNS SCORED

1-5

6-10

11-15

16-20

WINNING PERCENTAGE

.567

.543

.525

.504

 

Homefield advantage is seemingly larger in low-scoring games.

 

CONCLUSION AND FUTURE RESEARCH

 

            This is clearly just a starting point for future research.  There is a lot of information here, but there is a lot more to acquire.  If travel negatively affects players’ performance, then it would be interesting to check how home teams fared in the first game of a series as compared with later games in a series—especially for teams who were already at home for previous series.  I was unable to obtain this data, however.  If anybody knows how to acquire it, I would appreciate some guidance.  I also believe that there is a lot of work to be done with Pitch F/X data, but I am not adept at using this, so I would suggest that others research this.  It would be interesting to check umpire accuracy, pitch velocity, and groundball/flyball/line drive data for both hitters and pitchers and compare them at home and away.

           

In summary, I have made the following observations and conclusions.

 

  1. There is no persistence in which teams have larger homefield advantages
  2. Weaker teams have larger homefield advantage—or more likely, disadvantage away from home
  3. Homefield advantage sets in particularly during the first and last innings
  4. Home teams succeed at
    1. Hitting more homeruns
    2. Striking out less
    3. Walking more
    4. Getting hits on balls in play
    5. Pitching complete game shutouts
    6. Allowing fewer runners to reach on error
    7. Stealing bases more successfully
    8. Getting hit by more pitches
    9. Hitting triples
    10. NOT especially at hitting doubles, but at hitting extra base hits on balls in play overall
    11. Picking off opponents’ runners
  5. The largest homefield advantage in order occurs from largest to smallest in:
    1. Interleague cross-division games
    2. Interleague same-division games between non-“rivals”
    3. Non-interleague cross-division games
    4. Non-interleague same-division games
    5. Interleague “rivals” games
  6. There is no strong tendency to gain or lose homefield advantage when moving stadiums—but a very slight tendency to lose homefield advantage (though statistically insignificant)
  7. Home teams have more success in close games than blowouts
  8. Home teams have more success in low-scoring games