Home/Away Splits In March

The comments of the "Original Bracketology" post are really excellent. David pointed out what looks like a trend in the data from teamrankings.com. They have separate home and road ratings for each team, dating back to 98-99. He noticed that teams with good home rankings, but poor road marks, tend to fail pretty spectacularly in March:

"I looked for all teams in that span that were:
A) ranked 20th or better at home
B) ranked 51st or worse on the road
C) seeded 1 through 6

Year #Seed Team (Home/Away) - Result
----------------------------------------------------------
2006 #3 Iowa (2/53) - 1st round upset
2005 #6 LSU (15/51) - 1st round upset
2003 #6 Missouri (17/57) - 2nd round loss to #3 seed
2002 #6 Texas Tech (17/51) - 1st round upset
2001 #5 Virginia (4/56) - 1st round upset"

This is obviously not a very large sample, but it does look like there's something here.

I collected data on the top six seeds in each region from the '99-'06 tournaments (excluding '01, since the teamrankings.com data is incomplete). On the right are the expected wins for each seed. This gives us a baseline to fairly assess the wide range of teams that we are looking at.

Using this, we can see how each team did compared to their expectation. For example, if a #1 seed wins two games, they're -1.04. But if a 6 seed reaches the Sweet Sixteen, they're +0.75.

Next, taking David's idea, I took the teams with the biggest differentials between their home and away rankings. There are 168 teams in the sample, so let's look at the top and bottom 10%. Here are the teams with great home ranks that were awful on the road:


This is pretty amazing. 15 of these 17 teams won fewer games than expected*. The average "vs Exp W" is -0.50, which is remarkable. Let's look at the other end of the spectrum; teams with weak home ratings, that were good on the road:


It's not as extreme, but we do see the opposite trend. These teams win an average of 0.24 more games than we would expect.

This trend is both anecdotally and statistically significant. I did a regression using these 168 teams. The independent variables were their home and away ranks, and with the dependent variable (what we're trying to predict) being tournament wins. Since a lower rank should lead to more wins, we would expect both estimates to be negative. Here are the results:

The minuscule P-Values show that both variables are statistically significant.

The absolute value of "Away" is 58% larger than that of "Home". Clearly, a team's road performance is a much better predictor of tournament success than what it does on it's home floor.

What does this all mean? Well, let's apply it to this year. Since the data only considers the top six seeds, we'll stick to that- conveniently, that's what I covered in the Bracketology post. So, of those 24 teams, who has the most dramatic home/road splits?

I think sufficient attention has been paid to the first team on this list. I'm not at all surprised that Notre Dame has a significant split- they were 9-0 at home in the Big East, but only 5-4 away from South Bend.

Xavier was a little surprising, but understandable. They are 16-1 at home, with their only loss coming to Tennessee. Their road record is 7-4- none of these opponents have been particularly good, and they lost to the best one, Arizona St., by 22 back in December.

It should be noted that these two teams have the second and third largest differentials, but they're not huge. So I'd say it's a slightly negative indicator, but nothing to be too worried about. This does not apply to Vanderbilt, obviously.

The second half of the list is interesting as well. Washington St. has been tremendous on the road, with wins over Baylor, Gonzaga, USC, Arizona St., Oregon, Cal, and Washington. To this point, they've been held back by their disappointing 5-4 home record in the Pac-10. This is a dangerous team.

I am surprised that Drake appears here. I suppose road wins against Butler, Illinois St., Creighton, and Bradley help. I am still not sold on them, but maybe my skepticism is misplaced.

Update: I forgot to include Michigan St., because I didn't have them as a top 6 seed. In reality, they're probably a 6 right now. They are 3rd in the home rankings, and 59th away; they'd be right behind Vandy.

*Because of the weird distribution of wins (each year, one team wins six, one wins five, two win four, four win three, and so on), 65% of teams are below their expected wins. In our example, 88% of the teams in the first table (bad road teams) won less games than expected, while for the second table (good road teams) only 47% won less than expected. This confirms what is displayed in the average (road success is a better indicator than home games), but it's important to have a relevant baseline.