These are the predicted fight outcomes based on the Glicko-2 rating system. I describe the Glicko-2 system in my previous post about fighter rankings.

Prediction estimates are probabilities ranging from 0 to 1, and reflect the probability of "Fighter 1" winning. The closer to 1, the closer to 100% probability "Fighter 1" will win.

11 views0 comments

If you're like me and you've been following MMA for a long time, especially the UFC, then you know how much the sport has changed in terms of what is valued and how fighters are assessed. In the beginning the UFC was essentially a meritocracy. The best fighters would fight one another and their rankings were supposed to reflect something about their relative skill. The belt was a tangible symbol indicating who among the fighters was proven to be the best. Now however, we are in an era where the meaning of the belt is in flux, and fighter rankings are an opaque and subjective designation voted on by members of the media.

While this modern era offers a lot in the way of excitement and entertainment, it also breeds substantial confusion as many still look to the rankings as a way to measure a fighter's worth. Most of us find much to enjoy in the modern spectacle and character driven narratives we are presented with today, however a minority of us also miss the simplicity and relative objectivity of the meritocratic era of the UFC, and long for a sense of sobriety when it comes to belts and rankings.

Lucky for us, there are a number of statistical approaches to dealing with fighter rankings where our assessment of relative skill can be honest and based solely on fight outcomes. Here, using data on fight outcomes from Wikipedia, I implement the Glicko-2 rating system to determine fighter rankings. I am partial to this system because in addition to calculating a rating, it provides a way to evaluate both the uncertainty in that rating, as well as a measure of how erratic a fighter's performances are.

In the information I'm about to provide, "deviation" is a measure of uncertainty in the rating. In statistical terms, this is the standard deviation, so if a fighter has a "deviation" of 50 and a rating of 1750, their true rating is somewhere between 1650 and 1850. The less often a fighter competes, or the longer it's been since their last fight, the larger the deviation will get. The "volatility" measure tells us how consistent or how erratic a fighter performs. The more consistently they perform, the smaller this number gets. Lag indicates the number of events since the fighter last fought. The "Status" column indicates weather a fighter is an active member of that weight division or not. An "A" indicates that they are active in that division. An "I" indicates they are inactive, either because they have left that division or they are no longer on the UFC roster. Information on currently active fighters was also obtained from Wikipedia. A (c) indicates who the current champion is.

All ratings provided are here are all time ratings to date.

Disclaimer: Any catchweight fight with Cris Cyborg was lumped into Women's Featherweight for simplicity. All of Cyborg's opponents for those catchweight fights are therefore included in this ranking list.

I will try to update this on a regular (probably monthly) basis. If anyone knows of any better (and free) sources of data that would improve these rankings please let me know. Also, if anyone is interested in any more sophisticated statistical analyses of what predicts fighter success (wins-losses) and has access to detailed fight statistics, please let me know.

22 views0 comments
  • Facebook Social Icon
  • Twitter Social Icon
  • LinkedIn Social Icon
  • Research Gate
  • Google Scholar