The Butterfly Effect is not just a bad Ashton Kutcher movie (rotten tomatoes critics score = 33%). It is also not just part of a pseudo-intellectual sequence of Jurassic Park (rotten tomatoes critics score = 91%). The Butterfly Effect is concept in the field of Chaos Theory, most associated with American mathematician and meteorologist Edward Lorenz (rotten tomatoes: no opinion). Lorenz used the analogy of how a butterfly flapping its wings in the southern hemisphere may affect the path of a tornado in the Midwest. The underlying concept being that in very complex and dynamic systems (like the earth’s atmosphere), small changes in one place can have a very dramatic effect somewhere else.
In many ways, the methods that I use to model college football are quite similar. One lesson that I have learned over the years is that, in general, we just don’t have enough data to work with. I think that I can make some predictions with a known level of mathematical probability, but the variance is really high. Furthermore, the way that I have chosen to model the game relies heavily on the interconnected nature of individual teams and games. A blowout in the Sunbelt can affect my projected point spread for an ACC game two weeks later. That is just how the math works.
This week was a classic case in point. First of all, as I will show later, there was a LOT of chaos. There were 3 upsets where the spread opened over 18 points and 6 upsets in games where it opened over 9 points. The total number of upsets was 17, which is the highest of any week so far this year (although this value is well within the range that my simulation predicted for the week).
MSU didn’t play this week, but the flapping of butterfly wings around the country somehow seemed to do some damage to MSU’s resume. In order to see the details, let’s just move right into:
Big Ten Metrics Update:
Despite now playing a single down of football this week, MSU dropped from #19 in my power rankings to #26. Furthermore, MSU’s expected win total has now dropped from 7.66 to 7.34. This means that according to my calculations, MSU now has a better chance to finish at 7-5 than any other record.
Some of the game-by-game projected spreads have also changed fairly dramatically. For the rest of the season, I now have:
- +11.0 vs. Penn State
- -17.1 vs. Illinois
- +5.9 at Michigan
- -36.8 at Rutgers
- -18.0 vs. Maryland
Just last week, the Penn State line was around 7 (where it actually opened) and the Illinois spread was closer to 30 points. As for the odds to win a certain number of games total, the new numbers are:
- 9 wins: 6%
- 8 wins: 34%
- 7 wins: 49%
- 6 wins or less: 11%
This is not nearly as optimistic as just last week. In this case the flapping of the butterfly wings occurred when the majority of MSU’s previous opponents lost (Tulsa, Western, Arizona State, Northwestern, and Wisconsin). Two of those teams were upset (WMU and Wisconsin) and two others failed to cover pretty badly (Arizona State and Northwestern). In addition, the power ranking of 4 of MSU’s 5 remaining opponents all increased (everyone expect Rutgers). So, it doesn’t exactly take a long series of coupled differential equations to figure out why the numbers changed the way that they did.
As stated above, the actual data that we have with which to make decisions about whether Team A is better than Team B are quite sparse. Every week is a chance to collect more data to give us an idea of how good a team’s “average self” really is. Up until this week, MSU still looked like a Top 20 team, on balance. This week, the balance tilted down a bit. But the question still remains as to how good is MSU’s “average self?”
Has MSU underachieved so far, through a combination of some bad luck, bad officiating, correctable errors, and bad timing of their opponents? If so, MSU might just regress to the mean enough to win 9 regular season games with a shot at 10 over the holidays. Or, is MSU simply who they are now? If that is the case, a 7-5 record and a possible trip to sunny New York City might be the plan for December. Either way, MSU’s next chance to collect data is coming up, and that will go a long way towards determining the team’s ultimate fate.
As for the rest of the Big Ten, the race in the East is largely unchanged. OSU (89%) still maintains a healthy lead in the odds race over Penn State (11%) and those two teams check in at #1 and #4 in my current power rankings. No other team in the East has greater than a 1 in 2000 chance to win the division. MSU, UofM, and Indiana are currently all converging towards an apparent 5-4 league record, based on expected wins, but overall Michigan is still likely to be favored in more games going forward.
In the West, Wisconsin’s upset loss definitely put a dent in their odds, but I still have them at slightly over 80%. The reason there is that even though their raw power index took a hit. I still have them ranked #2 in the country. That is simply a measure of how far ahead of the pack OSU and Wisconsin were coming into the week. The interesting move here is Minnesota’s record and power ranking continues to rise (now at 35) such that the Gophers now have the 2nd best odds in the West (13%) with Iowa down at 6%.
Overall, I still give the Buckeyes the best overall shot (70%) to win the Big Ten, with Wisconsin (24%) still in 2nd and Penn State still with an outside shot (5%).
Betting and Upset Pick Results
As for the performance of my model for the week, the overall performance against the spread (ATS) was less than stellar. It went 27-34 (44%) which brings the model’s performance year-to-date to a dead even 200-200 (50%). Meanwhile, the FPI did a bit better for the week at 32-29 (53%), but overall my model is still ahead as the YTD total for the FPI is now 195-205 (49%).
As for my suggested bets, the summary table is shown below:
This week is pretty simple. I made 10 picks and 5 of them were correct for an even 50%. That brings my YTD total to 37-29 (56%) while the FPI held steady at 16-12 (57%). The combined strategy’s stats are now 50-39 (56%).
As for upsets the busy week in summarized in the following table:
As I mentioned above, it was a chaotic week in general, with a total of 17 upsets in 61 games. As for my picks, I went 3-4 to bring my YTD tally to 18-25 (42%). The FPI did a bit better at 3-2 (60%), bringing its total for the year to 11-13 (46%).
But the real story is the number of large upsets that occurred. Wisconsin’s loss as a 28.5-point favorite is the biggest upset of the year and is the 6th biggest upset on record since 2001 with respect to the opening line. Vanderbilt’s upset of Missouri as a 21-point dog is the 5th biggest upset this year, while GA Tech’s upset of Miami as an 18.5-point favorite is tied for 8th this year.
As for the big picture, here is my plot of the final point differentials versus the opening line
Once again, the teams above the dotted line all over-achieved in victory (or, I suppose their opponents under achieved in defeat). Those teams include Ohio State, App State, Auburn, Navy, Washington State, USC, Memphis, SMU, Air Force, and Virginia. Meanwhile, Houston, UCF, and Texas all underachieved, yet still won.
MSU is back on the horse this weekend, so keep an eye out for my Week 9 Betting picks and other analyses in a few days. Until then, enjoy, and Go Green.