Fiu, Cirminiello, Mitchell on TV - Campus Insiders | Buy College Football Tickets

Compu-Picks Analysis: Home Field

Mr Pac Ten
Posted Apr 12, 2011


2011 Compu-Picks Analysis Part 2: Home-Field as a function of Distance

As many of you know, home field advantage is a substantial part of compu-picks, both in terms of how it picks games and how it rates teams (those who have an unusually high number of home games take a ratings hit, and those with an unusually low number of home games get a ratings boost).

The first article about home-field advantage presented a substantial amount of information that clearly demonstrated that home-field is in fact a very big deal in college football. This second article looks at the effect of distance on home-field advantage

Take a look at the following table, consisting of the W/L record of home teams in league play across 1-A football for the last six seasons (2005 - 2010):

Home W/L Record - League Games Only

Same League Record Within 500 km 500 - 1000 km 1000+ km Total
Total BCS 41 - 25 62.1% 41 - 22 65.1% 42 - 19 68.9% 124 - 66 65.3%
Total 61 - 42 59.2% 66 - 34 66.0% 82 - 31 72.6% 209 - 107 66.1%
Home 1 Game better Within 500 km 500 - 1000 km 1000+ km Total
Total BCS 50 - 18 73.5% 43 - 18 70.5% 33 - 16 67.3% 126 - 52 70.8%
Total 80 - 27 74.8% 73 - 26 73.7% 78 - 28 73.6% 231 - 81 74.0%
Home 1 Game worse Within 500 km 500 - 1000 km 1000+ km Total
Total BCS 21 - 29 42.0% 28 - 41 40.6% 36 - 29 55.4% 85 - 99 46.2%
Total 34 - 51 40.0% 47 - 63 42.7% 47 - 56 45.6% 128 - 170 43.0%
Records Within One Game Within 500 km 500 - 1000 km 1000+ km Total
Total BCS 112 - 72 60.9% 112 - 81 58.0% 111 - 64 63.4% 335 - 217 60.7%
Total 175 - 120 59.3% 186 - 123 60.2% 207 - 115 64.3% 568 - 358 61.3%
Home 2+ Games better Within 500 km 500 - 1000 km 1000+ km Total
Total BCS 169 - 7 96.0% 168 - 15 91.8% 134 - 9 93.7% 471 - 31 93.8%
Total 274 - 25 91.6% 307 - 21 93.6% 289 - 16 94.8% 870 - 62 93.3%
Home 2+ Games worse Within 500 km 500 - 1000 km 1000+ km Total
Total BCS 19 - 169 10.1% 15 - 149 9.1% 24 - 114 17.4% 58 - 432 11.8%
Total 33 - 277 10.6% 29 - 288 9.1% 38 - 248 13.3% 100 - 813 11.0%

There are a few interesting takeaways from this table:

1) Home field is impacted by distance, especially if the distance is large
If you're familiar with scorecasting (and you should be), then you know that one of their arguments is that not only is referee bias the main determinant of home-field advantage, but that based on their data it was the ONLY determinant of home-field advantage.

At least in 1-A college football, this does not appear to be the case. As you can see in the above table, there is a nearly across the board trend towards the home team doing better the further the distance between the two schools.

While it is true that this effect is not huge, it is overall noticeable. Across 1-A, the difference between the win % for home teams within 500km and those 1000+km away is about 5% (59.3% vs 64.3%) when the league records are within 1 game of each other, and about 3% (91.6% vs 94.8% and 10.6% vs 13.3%) when they are not.

Interestingly, this effect is minor when you compare distances of 0-500km to distances of 500-1000 km. It is only when the distances are larger than 1000km that there is a large impact. I have not broken down the 1000+ km games yet to come up with a stronger rule, though I may look into this in the future.

I also have not looked into this as a predictive factor in the NFL, though I suspect that it is either much less of a factor or not a factor at all in the NFL, given how NFL teams will travel well in advance of games, enjoy better hotel arrangements, and generally not have to deal with the same scale of negative impact of long-distance travel as college teams have to deal with.

2) There is a sliding scale at work here, just like for home-field advantage as a whole
The bigger the talent gap, the smaller the absolute difference in upset rates between home and away, but the larger the relative difference (home upset rate / road upset rate). The same thing applies when factoring in distance.

When the teams are 1 game different, the road team gets an upset 36% of the time, and the home team 43% of the time (a 7% gap). Within 500km, it's 35.2% vs 40.0% (a 5% gap), but for 1000+km, it's 36.4% vs 45.6% (a 9% gap).

And when the teams are 2+ games different, the road team gets an upset 6.7% of the time, and the home team 11.0% of the time (a 4.3% gap). Within 500km, it's 8.4% vs 10.6% (a 2.2% gap), while for 1000+km, it's 5.2% vs 13.3% (a 8.1% gap, but almost TRIPLE the upset rate).

3) The impact of distance appears to be slightly higher for non-BCS leagues.
There is a lot of noise in these numbers, but overall the BCS splits are nonetheless smaller than the nationwide splits (which themselves would therefore have to be smaller than the non-BCS splits). Records within one game show a 2.5% difference for BCS schools and a 5.0% difference nationwide; home 2+ better shows a -2.3% difference for BCS schools and +3.2% nationwide; though pointing the other way, home 2+ worse shows a +7.3% difference for BCS schools and +2.7% nationwide.

Admittedly, it's possible that this is just noise, but it at least appears to be a real if minor effect.

.

In case any of you are curious, I'm going to attach the above table broken out by league. Please keep in mind that the numbers are sliced finely enough that it's REALLY difficult to draw any meaningful conclusions from the below numbers, but some of you may find the results interesting.

Home W/L Record - League Games Only

Same League Record
League Within 500 km 500 - 1000 km 1000+ km Total
ACC 11 - 2 84.6% 2 - 5 28.6% 13 - 5 72.2% 26 - 12 68.4%
Big 12 6 - 3 66.7% 8 - 4 66.7% 8 - 4 66.7% 22 - 11 66.7%
Big East 3 - 2 60.0% 7 - 1 87.5% 10 - 2 83.3% 20 - 5 80.0%
Big Ten 11 - 9 55.0% 7 - 4 63.6% 2 - 2 50.0% 20 - 15 57.1%
Pac-10 1 - 4 20.0% 3 - 3 50.0% 8 - 6 57.1% 12 - 13 48.0%
SEC 9 - 5 64.3% 14 - 5 73.7% 1 - 0 100.0% 24 - 10 70.6%
Indep 0 - 0 0 - 0 1 - 1 50.0% 1 - 1 50.0%
Mountain West 1 - 0 100.0% 7 - 2 77.8% 5 - 1 83.3% 13 - 3 81.3%
C-USA 3 - 2 60.0% 7 - 3 70.0% 7 - 6 53.8% 17 - 11 60.7%
MAC 13 - 10 56.5% 7 - 3 70.0% 0 - 0 20 - 13 60.6%
Sun Belt 3 - 4 42.9% 3 - 4 42.9% 9 - 2 81.8% 15 - 10 60.0%
WAC 0 - 1 0.0% 1 - 0 100.0% 18 - 2 90.0% 19 - 3 86.4%
Total BCS 41 - 25 62.1% 41 - 22 65.1% 42 - 19 68.9% 124 - 66 65.3%
Total 61 - 42 59.2% 66 - 34 66.0% 82 - 31 72.6% 209 - 107 66.1%
Home 1 Game better
League Within 500 km 500 - 1000 km 1000+ km Total
ACC 8 - 2 80.0% 6 - 2 75.0% 14 - 3 82.4% 28 - 7 80.0%
Big 12 8 - 1 88.9% 9 - 2 81.8% 7 - 2 77.8% 24 - 5 82.8%
Big East 2 - 3 40.0% 2 - 1 66.7% 5 - 2 71.4% 9 - 6 60.0%
Big Ten 12 - 4 75.0% 10 - 5 66.7% 0 - 0 22 - 9 71.0%
Pac-10 3 - 1 75.0% 9 - 2 81.8% 6 - 8 42.9% 18 - 11 62.1%
SEC 17 - 7 70.8% 7 - 6 53.8% 1 - 1 50.0% 25 - 14 64.1%
Indep 0 - 0 0 - 0 0 - 0 0 - 0
Mountain West 4 - 2 66.7% 7 - 1 87.5% 12 - 1 92.3% 23 - 4 85.2%
C-USA 3 - 0 100.0% 10 - 2 83.3% 11 - 5 68.8% 24 - 7 77.4%
MAC 16 - 6 72.7% 6 - 1 85.7% 0 - 0 22 - 7 75.9%
Sun Belt 6 - 1 85.7% 7 - 3 70.0% 4 - 5 44.4% 17 - 9 65.4%
WAC 1 - 0 100.0% 0 - 1 0.0% 18 - 1 94.7% 19 - 2 90.5%
Total BCS 50 - 18 73.5% 43 - 18 70.5% 33 - 16 67.3% 126 - 52 70.8%
Total 80 - 27 74.8% 73 - 26 73.7% 78 - 28 73.6% 231 - 81 74.0%
Home 1 Game worse
League Within 500 km 500 - 1000 km 1000+ km Total
ACC 4 - 4 50.0% 5 - 11 31.3% 8 - 15 34.8% 17 - 30 36.2%
Big 12 8 - 1 88.9% 4 - 3 57.1% 6 - 4 60.0% 18 - 8 69.2%
Big East 0 - 4 0.0% 3 - 2 60.0% 6 - 5 54.5% 9 - 11 45.0%
Big Ten 5 - 11 31.3% 5 - 6 45.5% 1 - 0 100.0% 11 - 17 39.3%
Pac-10 0 - 1 0.0% 5 - 8 38.5% 15 - 3 83.3% 20 - 12 62.5%
SEC 4 - 8 33.3% 6 - 11 35.3% 0 - 2 0.0% 10 - 21 32.3%
Indep 0 - 0 0 - 0 0 - 2 0.0% 0 - 2 0.0%
Mountain West 2 - 0 100.0% 5 - 5 50.0% 2 - 6 25.0% 9 - 11 45.0%
C-USA 3 - 2 60.0% 9 - 7 56.3% 4 - 5 44.4% 16 - 14 53.3%
MAC 6 - 18 25.0% 2 - 5 28.6% 1 - 0 100.0% 9 - 23 28.1%
Sun Belt 2 - 2 50.0% 2 - 2 50.0% 3 - 7 30.0% 7 - 11 38.9%
WAC 0 - 0 1 - 3 25.0% 1 - 7 12.5% 2 - 10 16.7%
Total BCS 21 - 29 42.0% 28 - 41 40.6% 36 - 29 55.4% 85 - 99 46.2%
Total 34 - 51 40.0% 47 - 63 42.7% 47 - 56 45.6% 128 - 170 43.0%
Records Within One Game
League Within 500 km 500 - 1000 km 1000+ km Total
ACC 23 - 8 74.2% 13 - 18 41.9% 35 - 23 60.3% 71 - 49 59.2%
Big 12 22 - 5 81.5% 21 - 9 70.0% 21 - 10 67.7% 64 - 24 72.7%
Big East 5 - 9 35.7% 12 - 4 75.0% 21 - 9 70.0% 38 - 22 63.3%
Big Ten 28 - 24 53.8% 22 - 15 59.5% 3 - 2 60.0% 53 - 41 56.4%
Pac-10 4 - 6 40.0% 17 - 13 56.7% 29 - 17 63.0% 50 - 36 58.1%
SEC 30 - 20 60.0% 27 - 22 55.1% 2 - 3 40.0% 59 - 45 56.7%
Indep 0 - 0 0 - 0 1 - 3 25.0% 1 - 3 25.0%
Mountain West 7 - 2 77.8% 19 - 8 70.4% 19 - 8 70.4% 45 - 18 71.4%
C-USA 9 - 4 69.2% 26 - 12 68.4% 22 - 16 57.9% 57 - 32 64.0%
MAC 35 - 34 50.7% 15 - 9 62.5% 1 - 0 100.0% 51 - 43 54.3%
Sun Belt 11 - 7 61.1% 12 - 9 57.1% 16 - 14 53.3% 39 - 30 56.5%
WAC 1 - 1 50.0% 2 - 4 33.3% 37 - 10 78.7% 40 - 15 72.7%
Total BCS 112 - 72 60.9% 112 - 81 58.0% 111 - 64 63.4% 335 - 217 60.7%
Total 175 - 120 59.3% 186 - 123 60.2% 207 - 115 64.3% 568 - 358 61.3%
Home 2+ Games better
League Within 500 km 500 - 1000 km 1000+ km Total
ACC 30 - 3 90.9% 21 - 3 87.5% 28 - 4 87.5% 79 - 10 88.8%
Big 12 29 - 0 100.0% 29 - 3 90.6% 30 - 3 90.9% 88 - 6 93.6%
Big East 14 - 1 93.3% 18 - 0 100.0% 20 - 1 95.2% 52 - 2 96.3%
Big Ten 41 - 0 100.0% 30 - 2 93.8% 10 - 0 100.0% 81 - 2 97.6%
Pac-10 15 - 0 100.0% 31 - 4 88.6% 39 - 1 97.5% 85 - 5 94.4%
SEC 40 - 3 93.0% 39 - 3 92.9% 7 - 0 100.0% 86 - 6 93.5%
Indep 0 - 0 0 - 0 1 - 0 100.0% 1 - 0 100.0%
Mountain West 11 - 1 91.7% 41 - 0 100.0% 26 - 0 100.0% 78 - 1 98.7%
C-USA 16 - 3 84.2% 41 - 4 91.1% 36 - 2 94.7% 93 - 9 91.2%
MAC 64 - 13 83.1% 24 - 1 96.0% 7 - 0 100.0% 95 - 14 87.2%
Sun Belt 5 - 1 83.3% 23 - 1 95.8% 26 - 4 86.7% 54 - 6 90.0%
WAC 9 - 0 100.0% 10 - 0 100.0% 59 - 1 98.3% 78 - 1 98.7%
Total BCS 169 - 7 96.0% 168 - 15 91.8% 134 - 9 93.7% 471 - 31 93.8%
Total 274 - 25 91.6% 307 - 21 93.6% 289 - 16 94.8% 870 - 62 93.3%
Home 2+ Games worse
League Within 500 km 500 - 1000 km 1000+ km Total
ACC 3 - 29 9.4% 5 - 17 22.7% 6 - 20 23.1% 14 - 66 17.5%
Big 12 4 - 31 11.4% 4 - 25 13.8% 3 - 27 10.0% 11 - 83 11.7%
Big East 2 - 11 15.4% 0 - 20 0.0% 5 - 16 23.8% 7 - 47 13.0%
Big Ten 6 - 43 12.2% 2 - 26 7.1% 1 - 8 11.1% 9 - 77 10.5%
Pac-10 1 - 10 9.1% 3 - 28 9.7% 8 - 39 17.0% 12 - 77 13.5%
SEC 3 - 45 6.3% 1 - 33 2.9% 1 - 4 20.0% 5 - 82 5.7%
Indep 0 - 0 0 - 0 0 - 2 0.0% 0 - 2 0.0%
Mountain West 1 - 8 11.1% 1 - 39 2.5% 2 - 23 8.0% 4 - 70 5.4%
C-USA 1 - 15 6.3% 7 - 44 13.7% 10 - 20 33.3% 18 - 79 18.6%
MAC 10 - 70 12.5% 2 - 24 7.7% 0 - 3 0.0% 12 - 97 11.0%
Sun Belt 2 - 8 20.0% 3 - 25 10.7% 0 - 21 0.0% 5 - 54 8.5%
WAC 0 - 7 0.0% 1 - 7 12.5% 2 - 65 3.0% 3 - 79 3.7%
Total BCS 19 - 169 10.1% 15 - 149 9.1% 24 - 114 17.4% 58 - 432 11.8%
Total 33 - 277 10.6% 29 - 288 9.1% 38 - 248 13.3% 100 - 813 11.0%

Focusing on the "All League Games" and "Records Within 1 Game" sections, these four leagues stand out as having especially strong home-field edges, both overall and in close matchups (I'll get to the WAC later). The Big 12 is especially striking, with the home winning at well over 60% when the records are the same, when the home team is one game better, AND when the home team is one game WORSE in the final standings.

The Mountain West has home teams winning at a similarly impressive clip, over 70% when the final league records are within a game of each other, and just shy of 59% overall. Even more striking is the fact that they did this despite having a very substantial record disparity, with less than 30% of all league matchups being "close" (final record within 1 game). CUSA and the Big East didn't have quite as striking numbers, but they were still solidly above-average.

Interestingly, the WAC had home teams winning at almost a 73% clip when the records were close, but only at a 56% rate overall. I could be mistaken, but it looks like the reason for this is that in the WAC, only 25.5% of all league games were close, far below average and in fact the lowest number anywhere. This suggests that the "true" HFA in the WAC is being dampened by the fact that the bulk majority of the games are simply mismatches. But that's just a guess on my end.

6) The SEC, MAC and Sun Belt have had especially weak home-field advantages.
It's been said in a few places that the SEC has lower home winning rates than other leagues, and the data bears this out (though it's worth noting that the SEC had one of the highest home winning rates in 2010, so it MAY be changing). For whatever reason, the SEC has had an especially high home winning rate in even matchups, but especially low across the rest of the board, most notably most notably for big home underdogs, who pulled off upsets at only half the national rate.

The MAC is especially striking, as the home winning rate is barely above 50%, and the home win rate is below average almost across the board.

I'm actually very curious if any of my readers has a good explanation why the Big 12, Big East, MWC and CUSA have such strong home-field advantages, and why the SEC, MAC and Sun Belt have such weak ones. I'm playing with the theory that distance is a big driver (longer distance = bigger HFA), since the SEC, MAC and Sun Belt cover relatively small geographic footprints, while the aforementioned four cover relatively large ones (this theory would also suggest that the WAC should have an especially large home-field advantage, given how widely spread out it is, which agrees with the idea that its HFA is dampened by the wide talent disparity).

Another reasonable guess is that weather/environment plays a factor, since the SEC, MAC and Sun Belt have similar weather patterns, and lack extreme altitude programs like Colorado (Big 12) or the majority of the Mountain West. Meanwhile the Big 12 is half Texas/Oklahoma and half freezing, the Big East runs from New England to Florida (with Syracuse playing in a dome), and the Mountain West has warm weather TCU, UNM and SD St, with the rest cold weather (though I don't think CUSA fits into this theory very well).

Some notes and caveats:

1) Geographic data from cfbtrivia.com.
Thanks very much to them for providing the geographic data I needed in order to create this article.

2) This article only looked at intra-league games.
I chose to look only at league games because inter-league games introduce a whole slew of biases, from the fact that better teams usually make favorable scheduling arrangements to the fact that some leagues are much better than others to the fact that every team schedules differently, which means that you can't just analyze teams by their simply W/L record.

By choosing to only look at league play and league records, I avoided these issues. In every league, there is long-term parity between home games and road games (though some teams will have one more home game this year and one more road game the next), and in every league the schedules are determined by the league and tend to be reasonably equal.

There's still the issue of division strength (a 4-4 Big 12 South team tends to be better than a 5-3 Big 12 North team, for instance), as well (slightly) unbalanced home-road splits in a given year, but overall, it should be a pretty consistently good measurement of quality.

Of course, it's reasonably possible that by only looking at league games I've missed important information. I may take a look at the non-league games sometime later this offseason and see if I can come up with anything interesting.

3) I've used 2010 league alignments for all seasons.
Other than WKU joining 1-A, I don't recall there being any realignment moves from 2005 - 2010. It's possible I may have overlooked something, though if I have it shouldn't be material to the high-level points.

4) League title game wins were counted as league wins.
This throws a little bit of noise into the numbers, but I don't think it's at all material. If someone wants to run the numbers adjusting for this and prove me wrong, go for it.

5) This article presumes that all home games were played at the school location.
This creates some level of noise given that sometimes Arkansas will have a home game in Little Rock, or Washington St in Seattle (and probably others that escape my immediate recall), but I do not believe that it is material.

2010 Compu-Picks Blog

Questions, comments or suggestions? Email me at cfn_ms@hotmail.com

Follow cfn_ms on Twitter