Fantasy Fights

Intelligent, unique MMA analysis

Applying Simple Rating System to UFC Fighters

In 2006 Doug Drinen wrote about a system he developed for ranking college football teams. He called this the “simple rating system.” I recommend reading about it here but I’ll provide a very brief explanation of what exactly SRS is.

SRS takes each team’s average margin of victory or defeat, then adjusts that by the average margin of victory of that team’s opponents. For example, if a team has an average margin of victory of +5.0 points, but their opponents have an average margin of -2.0 points, this team’s new rating is (5.0 – 2.0) = +3.0.

This process is done with every team, so every team has a new rating. Then the process is repeated with the new ratings. And repeated again. And again. And the process keeps being repeated until the ratings no longer move. Eventually, it gets to the point where every team’s rating stays the same. When this happens, the resulting numbers are the final team ratings.

In practice, this team that started at +5.0 might have their rating move as follows: from +5.0 to +3.0 to +4.4 to +3.6 to +4.2 to +3.8 to +4.0 to +3.9 to +4.0 to +3.9 to +3.9. The moment the rating no longer moves, we know that is the team’s overall rating in this system.

I know what you’re thinking: this is all well and good, but how could this be applied to MMA? MMA is not like football. In football, each team has a definitive score. In MMA, the result is not score-based, it’s result-based: win, loss, or draw.

This is where Fight Score and FPR come in. By adjusting how I calculate Fight Score, I can give each fighter a rating that estimates his performance every time he fights. This can be used in place of an actual score to get to the bottom of how good each fighter is.

To test this methodology I decided to take every UFC heavyweight fight between UFC 28 and UFC 36. This adds up to just 18 fights total featuring 17 heavyweights:

  • Josh Barnett: 5 fights
  • Randy Couture: 4 fights
  • Pedro Rizzo: 4 fights
  • Andrei Arlovski: 3 fights
  • Ricco Rodriguez: 3 fights
  • Pete Williams: 3 fights
  • Bobby Hoffman: 2 fights
  • Frank Mir: 2 fights
  • Semmy Schilt: 2 fights
  • Aaron Brink: 1 fight
  • Gan McGee: 1 fight
  • Jeff Monson: 1 fight
  • Kevin Randleman: 1 fight
  • Mark Robinson: 1 fight
  • Maurice Smith: 1 fight
  • Renato Sobral: 1 fight
  • Roberto Traven: 1 fight

Based on this very limited record of fights, the simple rating system method using Fight Score yielded these results:

  1. Josh Barnett +5.01
  2. Randy Couture +4.32
  3. Pedro Rizzo +3.38
  4. Ricco Rodriguez +2.97
  5. Renato Sobral +2.30
  6. Kevin Randleman +0.69
  7. Gan McGee +0.40
  8. Frank Mir +0.36
  9. Bobby Hoffman -1.08
  10. Semmy Schilt -1.15
  11. Maurice Smith -2.30
  12. Andrei Arlovski -2.51
  13. Jeff Monson -3.86
  14. Pete Williams -7.32
  15. Roberto Traven -7.58
  16. Mark Robinson -10.13
  17. Aaron Brink -11.00

Now, let’s look at where the UFC heavyweight division was after UFC 36. Josh Barnett had just defeated Randy Couture to win the UFC heavyweight title, then subsequently tested positive for steroid abuse and was stripped of the belt. Couture was coming off two wins over Pedro Rizzo, who had previously knocked Barnett out.

Rodriguez was 3-0 in the UFC but hadn’t fought any of the top three of Barnett, Couture, and Rizzo. Sobral and Smith had only fought each other and therefore SRS doesn’t work properly with them. Randleman had a competitive loss to Couture. Mir was 2-0 but had only fought Traven and Williams. Arlovski’s ranking here might seem low but he was just 1-2 in the UFC at this time.

Overall, despite the limited amount of data present, I think SRS does an excellent job of rating the UFC heavyweights at this point. If these heavyweights were to be ranked at this point in the UFC’s history, the top four could very well have been Barnett, Couture, Rizzo, and Rodriguez, in that order.

The question is: why do this when I already have FPR? The reason is that I’m not thrilled with the “strength of schedule” component of FPR. It’s based on Fight Matrix ratings, which I generally like but it’s difficult to translate them well to the FPR formula. I think SRS could potentially be a very good solution and given the success of this experiment, I’m curious to see how it would work with MMA fighters today.

The problem is that if I’m going to transition to a new FPR based on the SRS method, I’ll need to basically go through Fight Metric’s whole database, which covers over 500 MMA events all time. That will be a lengthy project but it’s one I’m going to begin. The current version of FPR is good enough to use for the time being, until the new version is ready to go.

Advertisements

6 responses to “Applying Simple Rating System to UFC Fighters

  1. Thomas February 8, 2014 at 12:41 pm

    I have read and read this post over and over again to try and figure out what it is you are trying to create here.

    I take it that you want to create a number on each fighter telling you how good he is, based on his so far performance ? and then what? do you then plan on just picking the guy with the highest number without taking nothing else into consideration?

    Try and explain to me how this system could have helped regarding the Darren Elkins vs Jeremy Stephens fight? how would this system have been able to make you choose Jeremy instead?

    regards

    Thomas

    • David Williams February 8, 2014 at 3:36 pm

      Thomas, here’s the summary of what I’m doing.

      Right now, FPR (Fighter Performance Rating) has two components. There’s the actual fighter rating, which is based on Fight Metric statistics. Then there’s the “strength of schedule” rating, which is based on historical Fight Matrix rankings.

      I like Fight Matrix as a ranking system but I think using their historical rankings as a strength of schedule rating is a poor fit.

      What I’m looking to do is replace the Fight Matrix rankings with the SRS method. That means instead of strength of schedule being based on Fight Matrix rankings, it will be based on the Fight Metric performance of each of a fighter’s opponents.

      So to repeat myself, all I’m doing is changing the way strength of schedule is calculated.

      With that said, it doesn’t matter how good my numbers get, I’ll never be able to pick every fight based on that number alone. Each fight still requires an analysis of the way each fighter matches up.

      Regarding Elkins-Stephens in particular… sometimes I’m going to get a fight wrong. The people who are the very best at fight prediction get 67% of the fights correct, maybe 68%. Last year I got 65% correct which means there’s not a lot of room for improvement. That means I’m just going to get some fights wrong, and Elkins-Stephens was one of them. I don’t regret my pick.

      I hope that answers your questions. Thanks for the support!

      -David

      • Thomas Algiz February 9, 2014 at 10:20 am

        Ahhh! i think i follow you now! So SRS tells us in a way how good the opponents of a fighter have been? but is that only the oppenents the figher has won over or those he lost to aswell?
        If i got that right then i must admit that is pretty clever! such a number could tell us alot and if you need help doing those 500 MMA Events feel free to ask.

        Regarding The Elkins fight then it was not to be picking on you that i brought that up. i didn’t lose regarding that fight cause i put my money on Stephens BUT! you could have told me that Nikita Krylov would win! i didn’t see that comin 😀 Overall i had 9 out of 11 correct to the last event. I got burned on Nikita Krylov and Sergio Pettis at the last event. The Sergio Pettis fight really surprised me but when i look over the fight i think my biggest mistake regarding that fight was that i underestimated Alex Caceres. Perhaps SRS would be able to tell me that i was underestimating Alex.

        I dont really keep a book on it but i guess im between 60-70% like you.

        Im really looking forward to be following the SRS.

        keep up the good work David!

      • David Williams February 10, 2014 at 3:17 am

        Thanks Thomas, and trust me when I say… I thought I was being foolish for betting on Krylov. I only bet on him because I didn’t think any UFC heavyweight should be +310 against Walt Harris.

        As far as the fighter SRS goes, it will probably be quite a while before it’s finished, so don’t hold your breath waiting for it. I greatly appreciate the offer to help but I need to keep all the data in the same spreadsheet, so it’s something only I can work on.

  2. kuboa February 9, 2014 at 2:22 pm

    I have just started betting on fights (via Kountermove), largely because I enjoy the research and also because I won a massive freeroll that makes me feel pretty studly. I will be using this site as a reference, to be sure. One question I hope you can speak to is how to know when an underdog is viable. Obviously if the odds are close one might feel OK going out on a limb of sorts, and I’ve also felt good betting on a guy different websites can’t agree on. What can you add to that? Thanks!

    • David Williams February 10, 2014 at 3:21 am

      I appreciate that, kuboa! How I know when an underdog has value… honestly, if I could answer that, I’d quit writing this blog and move to Las Vegas. My track record picking underdogs has been OK, but not so good that I would recommend you start betting the farm on whatever underdogs I’m picking.

      With that said, when I pick an underdog to win, it’s usually for one of three reasons.

      1) The underdog has a much higher FPR rating than the favorite
      2) The favorite has a lot of hype behind him, usually from a highlight reel knockout or something similar
      3) I think it’s a particularly bad style matchup for the favorite

      Congrats on winning the freeroll by the way. I just started playing on Kountermove myself, so here’s hoping for success for the both of us!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: