Can't wait to see the data on conferences, I think that could be very interesting.
The Over and Underperforming part 2
So this is a follow up to The Over and Underrated, however I have changed the title because I decided "overrated" maybe wasn't the best description of what I was trying to do. "Overrated" is a very strong word that involves hype, the media, and giving talking heads subjects to talk about. What I am trying to do is compare team performance to what is expected of them.
This time around, I will be comparing all teams in Division I-A over the past 10 years. I've also made another few important changes. First, I wanted to weigh team upsets; I didn't think it was fair a team would be knocked the same for losing a game where they were a 1 point favorite compared to losing a game as a 20 point favorite. A convenient method of weighing this is the actual point spread. So a team that pulls off a win when they were slated to lose by 20 would get 20 points, a team that lost when they were suppoed to win by 1 would lose 1 point, and a team that did what they were supposed to go would get 0. The total over 10 years is the "W Upset" column.
Another important change I made was giving the scores some context. In my previous analysis, basically we found that the more successful teams were overrated, while the crummy teams were underrated. Just because Indiana wins 4 games in a year they were only supposed to win 2, does not make them "underrated". So we need to compare the teams to their peers. I did this by plotting teams' total "Win %" vs. total "W Upset". Turns out there is a very decent linear correlation between the two (EDIT: chart now shown below). Teams that deviate the most above the trendline are overperforming, those that deviate below are underperforming.
Note that the trendline crosses the X axis at almost exactly 50%.
So finally, your most underperforming Division I-A football teams of the past 10 years:
W Upset: upset x line
# of Games: # of games played w/spread
Win %: actual win % in games w/spread
Pred Win%: percentage of games supposed to win
Plot: W Upset score based on linear trendline
Variance: variance from trendline
| # of
|Win %|| Pred
|New Mexico State||-5||113||31%||30%||54.8||-59.8|
So while some of these teams are likely expected (cough, Florida State, cough), others might be more of a surprise. But keep in mind that these variances show how these teams compare to others in their "class". For example, Fresno State's weighted Upset score of -107 would be very respectable in Florida's class. But compared to other teams with a Win % of 54%, it is terrible.
The most overperforming:
| # of
|Win %|| Pred
Boise State is far and away the most overperforming team. Boise is that badboy circled in red on the chart above. Just ... wow. I think this is good evidence as any that they need to exit the WAC, while I don't think Fresno State is ready to spread its wings yet.
And finally, here is how the Big 10 shakes out with the new strategy:
Purdue still sucks diddlyuck, and Northwestern is still the best. But now Michigan isn't looking so bad; when compared against other teams in its class, it actually performs just about where it should be. And MSU is finally near the top, where it belongs. I think this is undeniable proof that this system works. Ohio State has improved as well, which OK, fine, this is probably accurate.
- Single most overperforming team of the past 10 years: Stanford 2007. Huge upsets against Cal, Arizona, and oh yeah, USC. Honorable Mention: Notre Dame 2006... ugh.
- Single most underperforming team of the past 10 years: Louisville 2007. Upsets by juggernauts Kentucky, Syracuse, Utah, and Connecticut. Ouch. Honorable Mention: Bowling Green 2005.
Next up, we will look at conferences as a whole. Thoughts?
One correction though, you're going to want to change "underrated" to "underperforming" before your first table.
I was initially confused because I thought that the chart was supposed to be showing me underrated teams, which was inconsistent with the numbers.
I stared at the first chart for a few minutes trying to reconcile my interpretation of underrated with what I saw. Then seeing over performing made it make sense.
Thank you for pointing that out - I was sure I would mix the two up at some point.
This is great information for my gaming habit. Great work overall.
Noticed a trend that you probably did as well - most of the overperforming teams played in power conferences where they had quite a few opportunities to pull the upset. In particular, schools like NW and Ok St. seem to be perpetually underrated even though they consistently field decent-to-good teams. Boise, of course, blows everyone away, but that kind of makes sense - no matter how many giants they take down, people just cannot believe this little school in the middle of Idaho could be this good.
At the same time, everyone seems to want Fresno State to succeed; hopes are high for them every year, and they are consistently upset by "inferior" teams.
Maybe you are on to something with the location; Fresno is certainly "cooler" than Boise. The blue turf doesn't help either.
Fresno St. has been living off nearly beating USC a couple years ago the same way Charlie Weiss and ND has. I saw that game, and while it was certainly exciting, you had a sense that USC just forgot to turn it on until the second half. Fresno is a nice program, but you put them in the Pac-10 and they would be no more consistent than a UCLA or Oregon St. - occasionally competitive, but clearly in that second tier. Boise St., if it was in the Big 12 or the Big 10, would probably be in the top 1/4 of each conference most years.
Florida as an underperforming school is a real surprise. I assume that's mostly due to the Zook years.
Florida's problem in recent years is the number of games they are expected to win. Florida was only predicted to lose one game in the past 2 years, and instead got upset by Auburn, Georgia, MICHIGAN, and Ole Miss... all with large margins in Florida's favor. I think the problem here is that folks expect a little too much out of them.
Have you looked at the data if you add or subtract points based on how teams fair against the spread? For example a team that is supposed to lose by 20 only loses by 5 so they earn 15 points or a team favored by 20 wins by 25 so they earn 5 points. Obviously, you would lose points for not covering.
Thanks for the good work.
That's a good idea, and I am going to check it out. I initially wanted to focus on W's and L's since at the end of the game, that's all that really matters. But performance ATS might also provide some interesting results.
So this is basically all against the spread, right? So an upset is not winning a game outright, it's covering the spread? I may have missed something fundamental here.
I explain my reasoning a little more in my previous post (link above), but I only considered games where the game's (Hail! to the) victor was different than predicted. I didn't see much of a trend regarding performance against the spread, but I plan on looking at this more closely in the future.
So if Team M is supposed to beat Team O by 10 points and loses, Team M gets -10, Team O gets 10. If Team M wins by 1 pt, 10 pts, 29 pts, etc... both teams get a 0.
This might sound a little goofy, but it actually seems to work out very well when grading teams.