Let's Improve MMA: A Better Ranking System

Lev PisarskyDec 28, 2021


Editor’s note: The views and opinions expressed below are those of the author and do not necessarily reflect the views of Sherdog.com, its affiliates and sponsors or its parent company, Evolve Media.

* * *
My articles in 2021 have mostly fallen into one of three categories: analysis of a current fighter or match, a look back at an old-school fighter or match, or a criticism of the powers that run MMA. The last of these are the most popular and often the easiest to write. However, for my final opinion piece of the year, I wanted to try an inversion of that. I'm going to criticize an aspect of modern MMA and actually propose a solution. Moreover, this solution could be easily implemented tomorrow. There are no athletic commissions, powerful individuals or office politics to navigate. And the subject to be fixed is not a person, but rather an approach; namely, the current rankings system used by the Ultimate Fighting Championship and Bellator MMA.

Before we begin, let's acknowledge that no matter how perfect a ranking system is, a promotion can simply choose to disregard it. For instance, we've been hearing stories recently that Charles Oliveira wants to defend his UFC lightweight title against Conor McGregor, a man who has lost his last three fights at 155, all by stoppage, and whose last—and only—win at lightweight was over five years ago against Eddie Alvarez. It makes a total mockery of the ranking system, but as we've established, the UFC is first and foremost fighting for entertainment. A pay-per-view with Oliveira fighting McGregor will generate far more money than one featuring Justin Gaethje. A better ranking system offers no advantage if it's simply ignored, which will inevitably occur from time to time for any promotion.

With that caveat out of the way, what are we looking to fix? In general terms, that current rankings tell us very little, beyond which they're merely the subjective perceptions of a small group of people. For instance, what do we know about the ability of fighter ranked eighth versus ninth? Is the No. 8 ranked fighter better than No. 9 ranked fighter, or possibly worse, but does better against the rest of the top 10 or 15? Even such a basic point is rarely mentioned when rankings are discussed. Furthermore, is the eighth-ranked fighter a lot better than the ninth-ranked, or just barely?

What does it mean if they fight, and the No. 8 ranked fighter dominates the No. 9 ranked fighter? Does he or she stay at the same ranking, since the perception was that there is a large chasm between them, or does he or she advance, since the perception was that there is a small one? How are any of these obstacles handled? An individual ranker's whim? Flipping a coin?

What if I told you that there is a simple method for addressing all of these problems, and it's existed for over 60 years? I'm talking about Elo ratings. Created by Dr. Arpad Elo around 1960 for use in chess, this system gives every player a numerical score, with higher ones denoting a stronger level of play. A player will then gain points for every win and lose them for every defeat, with draws adding points if they occur against higher rated foes, and deducting points against lower rated. It's actually a remarkably simple system, based around win expectations. With a win denoting 1 point, a draw 0.5 points, and a loss 0 points, someone rated 100 points higher than their opponent will be expected to score 0.64 points on average. This is subtracted from their actual result and multiplied by K, which is the coefficient. Thus, if a 2400 rated player defeats a 2300 player, with a K of 20 for the game, he or she will win (1-0.64)*20= 7.2 points. As one can see, defeating higher rated opponents gives one far more points than lower rated ones. If a 2400 player defeats a 2600, against whom he or she is only expected to score 0.24 points, then they will win (1-0.24)*20= 15.2 points. Here is an online Elo calculator to play around with.

Now, this system, which is implemented for chess, table tennis, and a wide range of other competitions, would need to be reworked for mixed martial arts for very obvious reason. While Elo ratings work great for the many dozens of chess games or table tennis matches a player completes every year, a mixed martial artist may only fight 30 times in an entire career. Additionally, this means that improvement or regression is less likely to be captured by Elo ratings.

Luckily, there is a simple solution for this: Make the K much, much larger. With a K of 200, for instance, if a 2400 defeats a 2600, he or she will gain a whopping 152 points while the 2600 will lose that same 152, leaving the previously lower rated ahead in a list of top rated. The list of top rated, in turn, can be the basis of a new rankings system. Check out the current list of FIDE chess rankings. Eighth-ranked Wesley So, at 2772 points, is tied with sixth-ranked Levon Aronian and eighth-ranked Anish Giri, barely ahead of ninth-ranked Shakhriyar Mamedyarov. In MMA, the separation between rankings would of course be much larger. In chess, perfect play by both sides necessarily results in draws, which is not the case for MMA.

This is not the end of matters, though. For one, before implementing the Elo system, we still have to give top fighters ratings. This is partly a judgment call but not terribly difficult. Start by giving a few fighters ratings, and based on past results, interpolate everyone else's ratings, the same way it was first done for international chess players in 1970. Even if the initial ratings are incorrect, this will quickly be fixed over the course of many rated fights. However, we also want a way to distinguish between a one-sided beating and a close decision that could easily have gone the other way.

Again, this is not difficult to address. We can have different K values based on the result. Thus, for a victory by submission or knockout, we can have a full K value of 200. For a decision with a total separation of more than 5 points on the score-cards for a 3 rounder (for instance, cards of 30-27, 30-27, and 29-28 would have a separation of 7), the K value is reduced to 120. For scorecards with a separation of 3 or less, the K may be only 80, and for a split, 60. Obviously, one can play around with the specific K values, but the main idea should be clear.

This system addresses all the problems we originally outlined. With ratings, we would know exactly how much better the eighth-highest ranked fighter is compared to the ninth. Perhaps it's a small difference, with one being rated 2620 and the other 2610, or a large one, with the first being 2700. We also know that the rating system is not a measure of head-to-head ability, but one's average results against all fighters. And we now know exactly the significance of any fighter beating another, and the method in which they do so for rankings. Oh, and there is no subjective ranking by a small, largely anonymous group of people the UFC or Bellator select.

As a bonus, Elo ratings also mean we no longer have to juggle complicated head-to-head chains. For instance, Julianna Pena was fairly recently submitted by Germaine de Randamie but she just defeated Amanda Nunes, and Nunes defeated de Randamie via one-sided decision. How do we rank these three fighters? Elo ratings provide a simple answer. Furthermore, it's possible that the current champion is not the highest rated in the division. There is absolutely nothing wrong with this. Very often, the world champion in chess or table tennis is not the numerically highest rated, either. In that case, the highest-rated available contender becomes a very obvious next challenger.

Is the system perfect? Not by a long shot! Just like with present rankings, it's difficult to accurately rate fighters when they improve from bout to bout, or alternatively, regress with age and injuries. And large K values will still create huge swings based on a single performance, same as now. Moreover, ratings for martial artists first entering a major organization like the UFC or Bellator from a smaller one will always be very tricky. But it's a significant improvement over what we presently have. As noted, unlike most of my other gripes, fixing this doesn't require any employee to be replaced or fired or an athletic commission to make a major change.

Hopefully, an enterprising promotion will implement Elo ratings for MMA.