| Boyd's World-> Breadcrumbs Back to Omaha-> Getting a Grip on Home Field Advantage | About the author, Boyd Nation |
Publication Date: July 18, 2000
The Search for Evidence
The ISR's, in their current form, are extremely simple -- they take into account who you played and who won, and that's all. Because they're trying to solve a difficult problem, ordering all 280+ Division I teams, based on limited evidence, 50-60 games per team, the ordering that they produce tends to be as accurate as any other that can be produced, as far as we can tell, but there's a lot of uncertainty to that. That is, there may be more accurate orderings, but we just don't have enough evidence to tell.
Because of that uncertainty, I'm always hesitant to add complications to the formula. There are several factors that logically should be included, and they might produce a better ordering, but since it's extremely difficult to prove that those orderings are better, and common sense and logic are so frequently wrong, I have resisted adding them in without some evidence that they are beneficial.
Given all of that, you'd think I'd just leave things alone and do something else, but, well, I'm a guy, and that's just not how we do things. We tinker, we putter, we tweak. Some times that's a good thing -- da Vinci kept touching up the Mona Lisa until the day he died, for example -- but some times it leads to disaster, as anyone who's ever tried to debug code that directly violated the original specification because the programmer thought that some feature didn't make sense at the last minute can tell you.
The most notable of these factors, and the one that is most frequently suggested (usually in terms like, "You idiot, it's obvious that there's a severe home team advantage"), is game location. While I've always granted the possibility that there was a home field advantage (there is a small one that's provable in the major leagues, for example), I could never come up with a good way to quantify it -- to prove that it existed and get some feel for how big it was -- with the unbalanced scheduling that the college ranks play with. Virtually everyone plays a tougher schedule on the road than at home, and that makes it hard to get a handle on the problem. I think I've finally found a way that I'm comfortable with, though.
I finally realized that teams do not have complete control over their schedules, and that I could use that fact. While each team sets their own non-conference schedule, their conference schedules are largely consistent from year to year, and are constrained by the conferences to be fair over time. By looking at two years worth of conference results, I could get a data set where any two teams, with some degree of consistency of quality, most likely played each other in home-and-home series. Looking at these results shows some fairly strong evidence for the existence of a weak home field advantage.
Obviously, there are some flaws in this study. Team quality is not a constant from year to year, so there will be cases where, for example, the home team is much stronger in both years. Rainouts, opponent rotation in large conferences, and other factors can keep the number of games from being perfectly symmetrical. Still, any flaws that I can think of are things that should tend to even out over the whole data set, and I've tested for possible effects where I could.
The Results
The first question, of course, is whether or not there is a home field advantage. It looks like there is:
W L %
All Games 3716 3015 0.552
Home Team Favored 2536 868 0.745
Visiting Team Favored 2134 1174 0.645
Teams Even 6 13 0.316
In other words, the home team won 55.2% of the conference games over the last two years. All of the numbers in this column are based on the 1999 and 2000 seasons, with the WAC and Mountain West games removed to eliminate symmetry problems. I ran the same tests on the numbers for the 1998 and 1999 seasons and got essentially the same results; obviously that's not an independent data set, but it can be taken as a small bit of confirmation of the results.
In an attempt to guard against a possible odd effect that could have occurred if, through coincidence, the better team wound up playing at home more often, I checked to see what the records were for home favorites (defined as the team with the higher ISR) and home underdogs; that test also shows a 10% higher winning percentage for the home team. I'm willing to write off the 6-13 record for home teams with even ISR's as a sample size anomaly.
Fridays?
I've always heard that most of the home advantage effect in the pros took place in the first game of the series, so I took a look to see if that was true in the college ranks. While the home record is better in the first game, I wouldn't describe the effect as overpowering:
W L %
Home team -- First game of series 1141 883 0.564
Home team -- Subsequent games 2575 2132 0.547
Obviously, the actual stress of travel does have an effect on the first game of the series, but it's not a large effect compared to the overall home field advantage.
Individual Teams
When we get down to the level of looking at individual teams, we're really stretching the limits of the viability of our data, given that we're only talking about sixty games or so per team, but that's probably enough to give a feel for how teams have done over the two seasons in question, anyway. Whether or not these factors hold up over time is a question that I don't have sufficient data to determine; that will be interesting to see in a few years.
The best measure that I can find for an individual team's home field advantage is the difference in their home winning percentage and their road winning percentage. There are some flaws with this; most notably, it flattens out for teams at either end of the winning percentage spectrum -- Tennessee-Martin was 4-24 at home and 4-22 on the road, which doesn't give much room for an effect either way. However, it's the best measure I can come up with, and most teams play far enough from .000 or 1.000 that a home field advantage is possible.
Looking at the shape of these numbers is somewhat interesting: The numbers form a fairly-smooth bell curve around a difference of 9.5%. There are extremes on both ends, and one thing worthy of note is that around one third of the teams were worse at home than on the road. Some of the teams at the extremes are interesting, so here are the top and bottom twenty:
Home Road
Diff W L W L Team
0.524 16 5 5 16 North Carolina-Wilmington
0.463 17 10 5 25 Georgia
0.434 22 5 8 13 Oklahoma State
0.431 26 7 10 18 Miami, Ohio
0.414 23 7 12 22 New Orleans
0.400 16 8 8 22 Troy State
0.381 12 9 4 17 George Mason
0.379 29 4 15 15 Louisiana-Lafayette
0.375 17 7 8 16 North Carolina
0.373 12 5 7 14 Florida A&M
0.369 14 3 10 12 Maryland-Baltimore County
0.356 15 9 7 19 Pittsburgh
0.346 21 5 12 14 Southern Mississippi
0.322 23 7 12 15 UCLA
0.317 26 10 15 22 Texas Southern
0.313 19 5 11 12 Clemson
0.306 8 16 1 35 Iowa State
0.306 16 6 8 11 Wagner
0.305 13 7 10 19 Youngstown State
0.300 21 9 12 18 East Tennessee State
-0.102 6 5 11 6 Southern Utah
-0.111 18 9 21 6 Central Florida
-0.111 13 14 16 11 South Florida
-0.119 11 13 15 11 Murray State
-0.125 7 17 10 14 Morehead State
-0.126 11 10 13 7 Massachusetts
-0.128 21 9 24 5 Cal State Fullerton
-0.132 10 10 12 7 Charleston Southern
-0.132 11 9 15 7 Dartmouth
-0.138 10 15 14 12 Niagara
-0.143 10 18 15 15 Kentucky
-0.143 23 12 16 4 Minnesota
-0.145 10 23 13 16 Akron
-0.151 16 15 20 10 Evansville
-0.163 9 10 14 8 Old Dominion
-0.167 12 12 18 9 Fairfield
-0.190 11 10 15 6 George Washington
-0.247 7 11 14 8 Army
-0.298 3 19 10 13 St. Joseph's
-0.300 9 11 15 5 Brown
It's interesting to me that several of the teams that are frequently criticized as inconsistent -- Oklahoma State, Georgia, UCLA, and North Carolina -- are on the top list; their problem may be less a problem with consistency than a problem with playing on the road.
Conferences
Finally, are some conferences tougher on the road than others? Yes, although maybe not by enough to be outside the range of randomness:
Home Team % W L Conference 0.606 188 122 Big 12 0.596 106 72 MEAC 0.586 99 70 CAA 0.585 145 103 Sun Belt 0.582 124 89 ACC 0.580 112 81 NEC 0.578 203 148 SEC 0.564 185 143 SWAC 0.564 137 106 WCC 0.562 104 81 Mid-Continent 0.562 153 119 Big Ten 0.559 181 143 Southern 0.556 65 52 MCC 0.553 142 115 Pac 10 0.553 136 110 Big East 0.551 157 128 MVC 0.550 88 72 Big South 0.549 158 130 Southland 0.539 146 125 TAAC 0.536 142 123 C-USA 0.536 112 97 America East 0.531 191 169 MAC 0.527 135 121 MAAC 0.519 124 115 Big West 0.512 131 125 Atlantic 10 0.509 84 81 Ivy 0.493 110 113 OVC 0.483 58 62 Patriot
So What Do We Do with This?
Having shown all this, so far I'm at a loss as to what to do with it. Game location obviously does matter. On the other hand, how much it matters appears to vary with the team, although whether that's consistent or just random is a subject for future study. It's not really possible to include a varying home field advantage for each team, but it's not necessarily fair to just use a constant adjustment factor when one third of the teams actually play better on the road. I'll probably end up ignoring it for another couple of years, but suggestions are always welcome.
| Boyd's World-> Breadcrumbs Back to Omaha-> Getting a Grip on Home Field Advantage | About the author, Boyd Nation |