I’ve just finished work on a new method for rating competitors on The Ultimate Fighter. To do so, I gathered all the data I possibly could on TUF competitors of the past, and tested each and every data point to find out what had predictive value.
Out of everything I put into my SPSS program, it returned a grand total of two statistics that were decent predictors of a fighter’s performance.
The first was a variant of an early version of SILVA. This version of SILVA worked by simply giving each fight on a fighter’s record a rating based on the record and strength of schedule of the opponent. Wins were given a positive score, losses a negative score, and draws could be either positive or negative, depending on the strength of the opponent. The resulting SILVA score was the average of every individual fight rating.
The other decent predictor was a fighter’s ELO rating, as displayed in my top 25 rankings here on this blog.
I already published the ELO ratings of all 14 fighters who have made it into the house. That list is as follows:
TUF 17 ELO RATINGS
- Tor Troeng – 2403
- Robert McDaniel – 2268
- Uriah Hall – 2160
- Zak Cummings – 2110
- Josh Samman – 2102
- Luke Barnatt – 2086
- Dylan Andrews – 2062
- Kevin Casey – 2021
- Clint Hester – 1931
- Gilbert Smith – 1917
- Kelvin Gastelum – 1833
- Jimmy Quinlan – 1796
- Adam Cella – 1783
- Collin Hart – 1732
One good way to think of ELO ratings is a way of measuring a fighter’s divisional relevance – fighters with high ELO ratings tend to have had more recent success. This is exemplified by Tor Troeng, who is 11-1-1 in his last 13 fights, and “Bubba” McDaniel, who is on a six-fight winning streak and is 8-1 in his last nine.
Now, here is the list of fighters, with their ratings according to the SILVA variant:
- Zak Cummings – 37.62
- Uriah Hall – 30.77
- Luke Barnatt – 29.68
- Josh Samman – 28.30
- Tor Troeng – 26.14
- Dylan Andrews – 24.55
- Jimmy Quinlan – 23.28
- Kevin Casey – 22.49
- Clint Hester – 21.85
- Gilbert Smith – 21.37
- Kelvin Gastelum – 20.50
- Robert McDaniel – 19.86
- Adam Cella – 19.17
- Collin Hart – 17.51
Another way to think of these SILVA scores is as an estimate of how successful a fighter will be based on his first 12 fights. Fighters who are SILVA success stories – Daniel Cormier, Chris Weidman, Johny Hendricks, and Chad Mendes, to name a few – all rated at 45.00 or higher in this metric.
I believe that the best predictor would be to combine SILVA with ELO. In other words, to combine the estimate of a fighter’s potential with the estimate of that fighter’s real-world accomplishments.
Here is the resulting list:
OVERALL TUF 17 RATINGS
- Zak Cummings – 8.022 (Team Sonnen)
- Uriah Hall – 7.383 (Sonnen)
- Tor Troeng – 7.348 (Sonnen)
- Luke Barnatt – 7.127 (Sonnen)
- Josh Samman – 7.009 (Jones)
- Dylan Andrews – 6.532 (Jones)
- Robert McDaniel – 6.421 (Jones)
- Kevin Casey – 6.234 (Sonnen)
- Clint Hester – 5.995 (Jones)
- Gilbert Smith – 5.917 (Jones)
- Jimmy Quinlan – 5.893 (Sonnen)
- Kelvin Gastelum – 5.655 (Sonnen)
- Adam Cella – 5.428 (Jones)
- Collin Hart – 5.154 (Jones)
Despite all the work I’ve put in to this, these rankings still have to be taken with a grain of salt (or maybe even a shaker full of salt). The first reason is that there just isn’t a long enough history of TUF fights to draw precise conclusions about the accuracy of this model. The second is that there will always be a lot of variance – even if these ratings are a perfect measure of how good these fighters are (not even close), stuff happens. Upsets happen. Like I’ve said before, there was no information anywhere that would have suggested that Colton Smith would win the last season of TUF. At some point, I have to acknowledge that my attempts to rate these fighters and predict these fights will only get so good.
By performing a binary logistic regression on past TUF fights (you have to be a true stats nerd to know what that is), I’m able to translate these ratings into a probability of one fighter winning against another. Again, take it with a lot of salt.
With the disclaimers out of the way, my estimate is that Luke Barnatt has a 77 percent chance to beat Gilbert Smith in the first announced fight. At first glance, it appears to be a poor first fight choice for Team Jones. But I’m saying that with a completely unproven model. We’ll find out what happens in the second episode.