Let me start this post by saying that statistics-based projections like these are never going to substitute for a detailed look at the actual matchups. On the other hand, I don’t know anyone (including most national columnists) who has enough knowledge of all 65 teams to make a detailed matchup-based projection that covers all 63 games in the tournament. I see these projections as a supplement of sorts to aid in more detailed looks ate the weaknesses and strengths of teams that could matchup with one another.
Here’s a quick outline of what I’m doing. Every year I take the field and using a Markov-like process I start with the first round and go through all the games; picking winners based on the probability of a given team winning the game. Calculating the probability is probably the hardest part about this, and there are many ways to do it. The methods for this are a little dry for a blog post, but I’m more than willing to discuss the process with anyone who’s interested, just drop a comment or we can work something out via email. At any rate, after I get the probability of winning for two teams playing each other, I simply generate a random number. This number determines the winning team according to the probabilities (for example, if you think Clemson has a 70% chance of beating Michigan and the number you generate is 56, Clemson wins. If you generate 79, Michigan wins). Then I continue with the winners and move up round by round until finishing the tournament, keeping tally of which teams finished where. Then I rerun the process a million times; this gives the percent likelihood of each team advancing through to a round.
The chart below shows how well Clemson fares in the simulations, I’m showing it alongside Clemson’s numbers from the 2007 tournament for reference (initially I thought of showing the 2006 numbers as well, but yeah…)
The percentages in each column indicate the likelihood of Clemson making it through to that round. First thing to keep in mind is that all teams see substantial drops as they move through the bracket (well, except the 1 and 2 seeds in the first round) because with every round the pool of teams that can reach a particular spot increases. For example, there are eight teams vying for any particular sweet sixteen spot, so the percentages have to be divided up among all eight teams.
By my projections, Clemson is looking at about a 16-17% chance at reaching the elite eight. Oklahoma has the best chance in our group of eight, clocking in at about 32%, with Syracuse and Arizona State around 25% each and the other four teams at fractional percentages. Doesn’t sound like anything to write home about, yeah? But these are really fascinating results—you don’t usually see such an equal distribution for a sweet sixteen spot and that’s a pretty low percentage for a two seed. Moreover, Clemson comes out way in front of other 7/10 seeds, no one else clears 5% in the simulations. Clemson even equals or performs better than exactly half of the 4/5/6 seeds. I think we can take this as pretty strong evidence that Clemson is grossly under-seeded. If I’m an Oklahoma fan, I’m pretty upset about Clemson being in the neighboring game.
Well, I don’t want to dwell too long on this, because like I said, it’s a poor substitute for looking directly at the matchups. Tomorrow I’ll try to add to what ClemBen said about Michigan in his last post. If anybody has any specific questions about the simulations, drop a question in the comments. I’m probably going to tweak a few things in the next day and see how they come out. I don’t think it’s wise to post the entire projection before Thursday because I don’t want anyone taking this too seriously (just look at Clemson’s chances last year—we all know how that turned out). Plus, I can’t have ClemBen stealing any of my picks…
No comments:
Post a Comment