CageRank analyzes MMA fighter stats and previous match-ups to rate fighters and predict the outcomes of potential fights. Historically, we have predicted the outcome of fights correctly 64% of the time, and we are continuing to improve our model.
How We Rate Fighters
We assign each fighter a CageRank (a numeric rating) using the standard Elo Rating System (with k = 100 for those playing along at home). CageRanks are then used to predict fight outcomes. The fighter with the higher CageRank is given a higher probability of winning a fight. A difference of 400 points in CageRank translates rougly to a 10x better chance of winning. For example, we'd expect a fighter with CageRank 2100 to beat a fighter with CageRank 1700 over 90% of the time. Elo has been used for Tennis and Chess for many years, and appears to be a good fit for MMA, which is also a head-to-head contest.
Offense and Defense Ratings
In addition to an overall fighter rating, we use an Elo-like approach to evaluate a fighter's skill in striking, takedowns, knockouts, submissions, and finishing fights fast (before decision). For each of these categories we maintain an offense and defense rating that changes over time similarly to CageRank. For example, a fighter gains points in knockout offense if he knocks out another fighter, and loses points in knockout defense if he or she is knocked out. At the same time, the winner also gains points in knockout defense, since they weren't knocked out. The number of points exchanged is based on how good his opponent was at delivering and avoiding knockouts. A fighter won't gain much for knocking out a person who often gets knocked out, but will gain a lot from knocking out someone who is rarely knocked out.