Nate Silver is a nerd’s nerd. A number-cruncher extraordinaire.
This year, on his blog fivethirtyeight.com, he began an Elo analysis of the teams, and I began to look for values in the spread difference between his projections and the Vegas odds. I started this in the first week, and though I have slightly changed my picks during this time, my method is 30-18-2 for the year as of today. Yes, you will have to take my word for it – I did not anticipate that I would publicly post this info until recently, but if you go back, you can roughly verify my numbers.
I base the picks favoring Elo. In other words, I assume that Elo is correct and I look for value on the Vegas side, so if Elo indicates the Bears +3 and Vegas says +9, I pick the Bears, or if the Bears are -3 and Elo says they should be -9, I pick the Bears.
What I’ve discovered is that point differences of 2.5 – 3.5, favoring the Elo side produces the best results. Nate (can I call you Nate?) will tell you these results are within “luck”, but I’m not sure. Its working now. Interestingly, higher differences – +5 or more – seem to taper off on win percentage, but I still go with them. For the past few weeks, I have dropped anything less than 3 point differences, but I think I am going to start adding the 2.5 differences back in.
Bear (no pun intended) in mind that the ratings DO NOT take into account injured quarterbacks or other extenuating circumstances. For one or two weeks, I pulled some picks where the quarterback was out. However, it appears to depend on the team, and the value. So while Arizona can seemingly overcome key player outages, other teams cannot. Also, it is possible that the quarterback could be out on the “other” side. Statistically, it doesn’t seem to matter much, so I have decided to ignore injuries to key players to keep it simple.
FiveThirtyEight.com publishes their Elo ratings no later than Thursday evening. Here is where their Elo ratings are explained.