One of the ways I was able to look smart over the past six years, during which time I spent a lot of effort on political forecasting, was by betting on the favorite. I wasn’t literally placing bets, mind you (unless you want to count my proposed bet with Joe Scarborough). But for some reason, in political prognostication, you can be regarded as a savant just by pointing out that the favorite is probably going to win.
The standard in sports prediction is higher. And this year’s NCAA basketball tournament is designed to make me look dumb. There aren’t any favorites. Sure, some teams are better bets than others. (I wouldn’t advise staking your fortune on Cal Poly.) But the team that our statistical model regards as the favorite to win it all, Louisville (more on the Cardinals in a moment), has just a 15 percent chance of doing so. In other words, there’s an 85 percent chance that Louisville won’t cut down the nets again and that I’ll be wrong.
Likewise, the team we regard as most likely to reach the Final Four, Arizona, has only a 42 percent chance of doing so. The Wildcats are clearly the best team in the West, and they have some other factors (like a convenient travel schedule) working in their favor. However, it’s more likely than not that they won’t make it to Arlington.
All this ass-covering aside, we still think we can help you perform reasonably well in your office pool. Our interactive bracket will provide you with the odds of every team advancing to every stage of the tournament. What you elect to do with this information is up to you: Many office pools award bonus points for picking lower-ranked teams, and even when they don’t, there is some incentive to go against the grain by picking schools that everyone else ignores.
The model, which is in its fourth year, is principally based on a composite of five computer power ratings:
- Ken Pomeroy’s ratings
- Jeff Sagarin’s “predictor” ratings
- Sonny Moore’s ratings
- Joel Sokol’s LRMC ratings
- ESPN’s Basketball Power Index, a new addition this year
Each of these ratings has a strong track record in picking tournament games. We shouldn’t make too much of the differences among them: They are all based on the same basic information (wins and losses, strength of schedule, margin of victory) computed in slightly different ways (for example, the Moore ratings place a lot of emphasis on recent performance whereas the other systems do not). However, the reason to use five systems instead of one is because even small differences can compound over the course of a tournament that requires six or seven games to win. For instance, a team that is an 80 percent favorite to win each of its games has a 26 percent chance of going on a six-game winning streak, whereas one with a 70 percent chance has only a 12 percent chance of doing so. Each system has different features and bugs, and blending them helps to smooth out any rough edges.
These computer ratings are combined with a couple of human rankings:
- The NCAA selection committee’s 68-team “S-curve”, and
- Preseason rankings from the Associated Press and the coaches.
These rankings have some predictive power — if used in moderation. They make up two-sevenths of the rating for each team, as compared with five-sevenths for the computer systems.
It’s not a typo, by the way, to say that we look at preseason rankings. The reason is that a 30- to 35-game regular season isn’t all that large a sample (consider how much the consensus on the strongest teams shifted between midseason and the conference tournaments). Preseason rankings provide some estimate of each team’s underlying player and coaching talent. It’s a subjective estimate, but it nevertheless adds some value, based on our research. This benefits Louisville, Kentucky, Michigan State, Ohio State and other teams that slightly underachieved expectations but that have the skill to make a deep run.
The ratings are further adjusted for three factors. The first is injuries and player suspensions. We review injury reports and deduct points from teams that have key players out of the lineup: for instance, Joel Embiid from Kansas or Kyle Collinsworth of BYU. (This process might sound arbitrary, but it isn’t: The adjustment is based on Sports-Reference.com’s Win Shares, which estimate the contribution of each player to his team’s record.)
The injury adjustment also works in reverse: We’ve reviewed each team to see which are healthier now than during the regular season. This helps Michigan State, for example, which finally has its starting lineup together.
The next adjustment is for travel distance. Are you not at your best when you fly in from LAX to take an 8 a.m. meeting in Boston? The same is true of college basketball players. In extreme cases (a team playing very near to its campus, or having to travel across the country to play a game), the effect of travel can be tantamount to playing a home or road game, despite being on an ostensibly neutral court.
The final adjustment takes place only once the tournament is underway. The FiveThirtyEight model will continually update the odds at the conclusion of each game, and it will adjust each team’s odds based on the score of each tournament game so far. A No. 12 seed that waltzes through its play-in game and then crushes a No. 5 seed may be much more dangerous than it initially appeared; our model accounts for this. On the flip side, a highly rated team that wins but looks wobbly against a lower seed (like Indiana last year) often struggles in the next round, we’ve found.
On this basis, the model accounts for all possible opponents that a team might face as it moves forward in the tournament. Obviously, some teams face much tougher draws. The computer and human rankings aren’t extraordinarily high on Wichita State this year, despite its 34-0 record. But the Shockers are a legitimate No. 1 seed. What really kills them is an absolutely loaded Midwest region. Even if the team is a favorite to win each game it plays, the odds are it will stumble along the line.
So let’s take a whirlwind tour of the four regions. In each one, I’ve listed the five teams that are most likely to advance to the Final Four, according to our forecasts.
South
| Seed | Team | Chance to advance to Final Four |
|---|---|---|
| 1 | Florida | 41.1% |
| 2 | Kansas | 20.8% |
| 3 | Syracuse | 8.6% |
| 6 | Ohio State | 8.6% |
| 4 | UCLA | 8.1% |
| Rest of field | 12.7% |
No surprises here: Florida is the favorite. The Gators, the No. 1 overall seed, benefit in part from the injury to Kansas’ Joel Embiid. (Although Embiid is more a great prospect than a dominant college player: If he were guaranteed to be healthy, Kansas’ probability of reaching the Final Four would rise only to 24.3%, while Florida’s would fall to 39.9%).
Nevertheless, the SEC isn’t the basketball power that it is in football. It wouldn’t be overstating the case to say that the SEC is halfway to being a mid-major: According to Jeff Sagarin’s conference ratings, the difference between the SEC and the Big 12 (the top-rated basketball conference this year) was larger than that between the SEC and the West Coast, Atlantic 10 or Mountain West conferences. Florida could be challenged not only by Kansas, but by whichever of No. 3 Syracuse and No. 6 Ohio State advances to the regional semifinals.
East
| Seed | Team | Chance to advance to Final Four |
|---|---|---|
| 4 | Michigan State | 24.0% |
| 1 | Virginia | 23.3% |
| 2 | Villanova | 21.1% |
| 3 | Iowa State | 8.2% |
| 6 | North Carolina | 6.4% |
| Rest of field | 17.0% |
I grew up in East Lansing, Mich., and am expecting to be accused of being a homer for having Sparty as the nominal favorite in this region. However, there is no subjective component to the FiveThirtyEight model. After winning the Big Ten tournament yesterday, Michigan State should have been a No. 2 or No. 3 seed based on its regular-season resume, according to the computer ratings. Other factors boost its standing further. Its lineup is finally healthy — something our model accounts for — and its No. 2 preseason ranking further speaks to its talent. The model does not account for coach Tom Izzo’s extraordinary postseason record, a statistical outlier that might be another reason to pick the Spartans.
And yet, if we have Michigan State as the favorite in this region, it is only by the slimmest margin. Virginia, the No. 1 seed, has essentially the same chance of advancing. More overlooked, perhaps, is Villanova, which has a plush travel schedule that will take it no farther than Buffalo and Madison Square Garden, where the regional finals will be held.
West
| Seed | Team | Chance to advance to Final Four |
|---|---|---|
| 1 | Arizona | 42.4% |
| 2 | Wisconsin | 15.6% |
| 3 | Creighton | 12.2% |
| 4 | San Diego State | 7.5% |
| 9 | Oklahoma State | 4.8% |
| Rest of field | 17.5% |
In our model, travel often plays a fairly large role in determining the favorites in the NCAA tournament’s West region, with some teams having to cross the country while others are on home terrain. But this season saw something of a revival of college basketball in the western half of the country. All schools in the region but No. 2 seed Wisconsin and No. 15 American are located west of the Mississippi River.
Nonetheless, Arizona has an especially easy travel schedule and can reach the Final Four while crossing only one state border (into California, where it’ll play in San Diego and Anaheim). The Wildcats are the favorites here, although they face a potentially tough second game against Oklahoma State, which the model regards as being more talented than its regular-season record might imply.
Midwest
| Seed | Team | Chance to advance to Final Four |
|---|---|---|
| 4 | Louisville | 38.1% |
| 3 | Duke | 17.9% |
| 2 | Michigan | 14.2% |
| 1 | Wichita State | 13.5% |
| 8 | Kentucky | 7.9% |
| Rest of field | 8.3% |
Louisville is the defending national champion, was ranked third in the preseason, was ranked fifth in the AP poll last week, won its conference championship in dominant fashion, and ranks between first and fourth nationally in each of the five computer systems. The team’s strength of schedule wasn’t great, but its five losses came against ranked opponents by single-digit margins. So how were the Cardinals a No. 4 seed?
The answer, almost certainly, is the selection committee’s attachment to Ratings Percentage Index (RPI), which somehow evaluates Louisville as only the 18th-best team in the country. RPI, as I’ve written previously, was “developed in 1981 in the era of the DOS prompt and the Commodore 64.”
There’s room to debate whether today’s computer systems give Louisville too much credit for blowout wins (such as its 61-point victory against Rutgers on Thursday). But it’s not as though the Cardinals are lacking in other credentials. The debate ought to have been about whether Louisville was a No. 1 seed or a No. 2.
That’s why they play the games, of course. But the problem with so badly mis-seeding a team is that it penalizes those that stand in its way. In this case, that means Wichita State, which will potentially face Louisville in the regional semifinals. Nor did the selection committee do the Shockers any favors: Kentucky is exceptionally talented for a No. 8 seed, as Duke is for a No. 3.
RPI is the Dick Morris of algorithms. Maybe I should challenge it to a bet.
