Writers of Pro Football Prospectus 2008

31 Mar 2011

ESPN: College Football HFA

Previous efforts to calculate home field advantage have focused on the difference between a team's home and road performance. For this piece, I calculated retrodictive expected home and away game efficiency performances based on end-of-year FEI ratings. I then compared the actual game efficiency against those expectations in order to isolate home field advantage (and separately, road field disadvantage) for each team since 2003. In order to eliminate the noise of outliers, only games between teams within one standard deviation of FEI power (roughly 25-40 ranking places) were included.

Only the top-10 and bottom-10 in home field advantage and road field advantage were included in the piece. Let me know if you're curious about other specific teams.

Underrated home field advantage: Baylor Bears. Overrated home field advantage: Texas Longhorns. There are sample-size issues with this study, so I'll need to add a bit more historical depth. And we'll need to examine this much more closely to find out how to use it in forecasting, if at all.

Posted by: Brian Fremeau on 31 Mar 2011

13 comments, Last at 10 Apr 2011, 12:43pm by Brian Fremeau


by Aaron Brooks' Good Twin (not verified) :: Thu, 03/31/2011 - 5:33pm

Between 1949 and 1956, Oklahoma went 77-5-2.

That included 1 home loss, 2 road losses, and 2 neutral losses. Both ties were on the road.

According to a home/road analysis, there wasn't much different in result by location. Is that because the Sooners didn't have much HFA in the 1950s, or because they were so dominant than it's hard to discriminate between utterly dominant versus merely mostly dominant?

by Brian Fremeau :: Thu, 03/31/2011 - 8:30pm

As a Notre Dame fan, I'm tempted to think that Holtz' teams were much better home teams than current Irish teams are. But in reality, those teams were simply better teams -- home/away may have been average or below average.

by Joseph :: Fri, 04/01/2011 - 1:24pm

Real surprising that LSU is the 12th worst BCS school. I wonder if (after maybe adding in other seasons to increase the sample size) night vs. day makes any difference? Because from what I have heard, Death Valley at night is one of the HARDEST places in the nation to play. The main reason discussed is that the fans have all afternoon & evening to get drunk, hyped-up, etc.--making it EXTREMELY loud.
Brian, one thing that you prob. are factoring in, but would be interesting: I know that you explained that the games compared are teams within 1 SD of each other to avoid outliers (which is smart). But, I wonder if, when the games are on the edge of the SD, if teams do slightly worse against teams that are 1 SD better, but run up the margin of victory against the the teams almost 1 SD worse. In other words, if the games of team X being measured come against teams that are consistently better, even if they're expected to lose by 7, then they lose by 14, they're measured as 7 below. Or, conversely the opposite.
So, Brian, to ask a favor concerning this data:
1. Lump all games the 1 SD games together.
2. Separate by: <1/3 SD, no matter if home or away is favored by FEI; between 1/3 & 2/3 SD--home team favored by FEI, and away favored; then between 2/3 & 1 SD, home favored, then away favored. In other words, 5 categories (you could even do six if you wanted to separate the "equal" teams).
3. Take the aggregate results of all home teams in each category. How do they do, compared to expected?
If certain teams have a big HFA like you posit in the article (and I don't doubt your math or methods), then we shouldn't see much difference in the categories, because the over-performing home favorites should "cancel out" the under-performers. However, if all teams tend to over-perform when they are 7-point favorites by FEI (or whatever the 1 SD advantage is on the scoreboard), and under-perform when they're home underdogs, then I would just say that you're still dealing with "outliers"--that is, the bigger the differential in SD, the larger the over- or under-performance.
I realize that by only including games with 1 SD or less in your study, you tried to control for that. I also wonder if for some outlier results games, there was an injury/suspension that made team X weaker than season-ending FEI perceives it, thus polluting your sample for that team.
In short--good work, and I wonder if you could test some other controlled hypotheses with the data to see if there could be other factors influencing your data.

by Brian Fremeau :: Fri, 04/01/2011 - 1:34pm

I think this is only the tip of the HFA iceberg for me. We had a good discussion of other tests/questions that this analysis posed over at Football Study Hall last night, and your questions are great additions.

by Alan Darkwater (not verified) :: Fri, 04/01/2011 - 10:14pm

I have some serious issues with this "study". I'm not sure how anybody who follows college football can say that Tiger Stadium is not one of, if not the hardest stadiums to play in on the road. The numbers you looked at are great and props to you for trying to scientifically explain some of the . But when you look at the win-loss record at LSU at home against quality opponents, especially at night, I would think you'd see something different. Even over the last couple years with a below-average offense they have beat some very good teams. I will give some of this magic to the hat but I think more times than not opponents are scared to go into Death Valley. I'm actually curious to find out if you think differently or you stand by the study.

by cfn_ms :: Sat, 04/02/2011 - 2:03pm

FWIW, my own numbers for 2003-2010 showed LSU as having one of the LOWEST HFA's in 1-A (not quite lowest 10, but really close). More than most, they've tended to play about as well on the road as at home. It's possible that if you isolate only night home games the story changes, but in aggregate, their performances at home hasn't been hugely different from their performance on the road.

In fact, there were two years where they were pretty clearly better on the road: in 2005 (when they had that horrible home loss to Tennessee and were 5-0 in true road games) and 2008 (1-3 at home in league play, 2-2 on the road)

by zlionsfan :: Sat, 04/02/2011 - 3:28pm

You have to keep in mind that he's not saying that Tiger Stadium doesn't provide an advantage for LSU. What he's saying is that Tiger Stadium doesn't provide a significant advantage for LSU against peer opposition relative to what other schools get in the same situation.

In the same way, he's not saying that Memorial Stadium in Bloomington is one of the most difficult places to play in the country (which it isn't: I doubt IU has ever drawn serious crowds there - they certainly didn't when I lived there - and more than likely, when they do, it's from teams like Michigan, Notre Dame, Purdue, and Ohio State - and soon, Nebraska - who either travel well, are close by, or both) ... he's saying that at home, Indiana performs better against its peers than you'd expect.

You can see that including only IU's peers removed 20 games from that eight-season period, games in which Indiana was 3-17. (Baylor saw a similar effect: they went from 15-27 to 10-2.) And that seems to take us in a different direction: perhaps for schools like these, their HFA is most significant when the opposing crowd is small. This seems to make sense, assuming Baylor is like Indiana in that they don't sell out frequently and probably have a lot of visiting fans when larger schools come to play.

Why are so few power schools in the top 10? Probably a similar reason. Many of them play in imposing stadiums, but their peers are accustomed to playing in similar places. An average team probably doesn't do well in front of 90,000 or more hostile fans, but then this set of data doesn't (for the most part) include normal teams unless it's measuring a normal team.

I'd imagine Ohio State is near the bottom largely for Tressel's coaching style ... he's more than happy to open up a 10-point lead and sit on it the rest of the game. Boring, but effective. Of course if FEI expects the Buckeyes to win by 17, they'll "underperform" ...

by lionsbob :: Sat, 04/02/2011 - 12:31pm

LSU HFA is overrated. I am pretty sure they have spent more time booing their own team then scaring the opposition....

by mtharper7 (not verified) :: Mon, 04/04/2011 - 4:00pm

I was wondering where Florida ranked. I know that the Swamp is always regarded as one of the most difficult places to play, but I would like to know if there are stats that back that up.


by cfn_ms :: Mon, 04/04/2011 - 11:47pm

I can't comment on the FO numbers, but my own say that Florida is right around average HFA for 2003 - 2010.

by Brian Fremeau :: Fri, 04/08/2011 - 7:22pm

Florida ranked 31st in adjusted HFA, +2.0 points per 24-poss game better than expected.

by cfn_ms :: Sat, 04/09/2011 - 3:12pm

Out of curiosity, what did your model have as Florida's RFA?

by Brian Fremeau :: Sun, 04/10/2011 - 12:43pm

Average (No. 62): +0.1 points per 24-poss game