Saturday, March 21, 2009

Why Dayton Was A Bigger Upset Than Cleveland State

I realize that I didn't really put any statistical backing when I made this statement earlier today. My regular readers know about which computer ratings to use when predicting games, and which to ignore. But I get a lot of non-regular readers around Tournament time, and they might be surprised to read somebody who isn't just regurgitating the crap from morons like Digger Phelps and Greg Anthony. Dayton was indeed the biggest upset yesterday, but let me give you some reasons.

The best predictive tool, of course, is the Sagarin PREDICTOR. Unfortunately I can't pull up the numbers from yesterday, so I can only give you the Sagarin numbers as they are now, but with the sample sizes as large as they are nobody moves all too much because of one game. But given the caveat that the numbers have changed a bit, Wake is currently an 86.31, West Virginia is 89.25, Dayton 80.43, and Cleveland State 81.56. In other words, on a neutral field Wake would be favored over Cleveland State by 4.75 points, while West Virginia would be favored over Dayton by 8.82. Considering that both Dayton and Cleveland State won yesterday by fairly similar margins, I would guarantee that the gulf between those predicted margins of victory hasn't changed much.

Now, let's move over to Pomeroy. Pomeroy has also updated with yesterday's results, but again, I'm certain that the relative numbers for Wake-vs-Cleveland State and Dayton-vs-WVU haven't moved much. Pomeroy has Wake at .9135, WVU at .9500, Cleveland State at .8367 and Dayton at .7973. If we use the classic Bill James method of calculating winning percentages with Pythagorean rankings (which is what Pomeroy is) we find an estimation that Wake would have a 67.3% chance of beating Cleveland State on a neutral floor right now. West Virginia, on the other hand, would have an 82.8% chance of winning. Again, not only is West Virginia the bigger surprise, but it's not even close.


Before the Tournament, Basketball Prospectus put some numbers together with statisticians like Ken Pomeroy, and gave the percentage chance that each team had of reaching each round of the Tournament. The East and Midwest are here, and the South and West are here. I actually wasn't aware of these previews before the Tournament started, but I don't think it would have hampered your bracket too much if you only read this blog. They use the same computer ratings that I do, so they ended up with similar results. The reason that I think my previews are better is because they take into account even more factors, like psychology. For example, while they did nail Wisconsin (they gave the Badgers a 53.0% chance of winning over Florida State, and made them the sixth most likely team to win the whole region at 2.8%), they erred on two games that the computers would not have noticed. They gave Arizona only a 42.3% chance of beating Utah, which doesn't take into account the fact that I always talk about, which is that BCS conference 12 seeds actually have a winning record since the bracket moved to 64/65 teams, and non-BCS 5 seeds have historically been much more likely to lose than BCS 5 seeds. The computers also gave Marquette a far-too-large 71.68% chance of winning over Utah State, because they can't take into account the fact that Marquette's best player wouldn't be on the floor. Marquette did win by 1 point, but if you played that game ten times Utah State would win a majority of the time.

Getting back on topic, the computers agreed with my sense that Wake Forest was a heck of a team, but that West Virginia was better. They gave West Virginia an 87.3% chance of winning in round one, compared to only a 75.1% chance for Wake Forest. These are very similar to the numbers that I came up with. Also, they gave West Virginia a 54.0% chance of making the Sweet 16 (I also put them through to the Sweet 16), compared to 45.2% chance for Wake Forest. Of course, that 45.2% chance is far higher than either Utah (28.6%) or any other team in that pod, so Wake was still the right play to the Sweet 16, which is also as far as I had them going.


I'm sorry for throwing so many numbers at people. If you have an educational background that is heavy on numbers and/or have a job that is numbers-oriented, I think all of that math will be really interesting to you. It sure is very interesting to me. If you aren't a big numbers person (and don't view that as a putdown, it's just a different style of thinking and many of the smartest people I know are not numbers people), you might glaze over some of those numbers. But the moral of the story is that the computers agree and the numbers are overwhelming: the Dayton result was the biggest upset of yesterday by far. The Cleveland State over Wake Forest result was a surprise, but not a huge surprise. You won't hear anybody point that out on ESPN or CBS, but nobody gets an analyst position on ESPN or CBS because they make good, educated points. Most of the time they're told what argument to make by their producers, so they don't even have to think. But thankfully we can all do the math ourselves. And the results are enlightening, don't you think?

2 comments:

DMoore said...

I don't know if this is a comment or a question. I agree that the Wake Forest loss was not too surprising, because according to the eyeball test, they were not as good at the end of the season as they were in January. Do any of the computer predictors really account for that? I just don't see the slip in the numbers that I would have expected for Wake or UConn, compared with what I saw on the court. Likewise, I would have expected Maryland and Florida State to get a boost, and Clemson to dive more than they did.

Jeff said...

Different computer ratings have different policies. Some weight more recent games more, and others weight all games equally. Depends on the rankings. Personally, I don't think those teams were ranked too badly based on how they finished, honestly.

Even though UConn isn't what they were before Dyson, they're still definitely one of the top five or six teams in the nation. And Wake is still a very good team - if they'd gotten past Cleveland State they'd have had a great shot of going to the Elite 8 or Final 4. If you look at Pomeroy's "Consistency" rating, you can see that Wake rates as extremely inconsistent (272nd in the nation), which is evidence of what we saw with our own eyes.