The Next Big Leap for the FightMatrix Rating System: Community Adjustment

4 Jun , 2011,

As the system matures and the MMA scene changes, we must adapt the rating system to handle new challenges.

Recently, a point that has been brought to my attention is the system’s tendency to overrate Japanese fighters who stay in Japan, specifically the male fighters.  In comparison to when the system was first created nearly four years ago, there are many more instances of Japanese fighters fighting in other countries.   Now, the data is finally there to prove what we thought all along… the system IS in fact overrating the Japanese population.  Not only that, but the problem is getting progressively worse and also affects foreigners fighting there.

The past data is there to prove this objectively.  Furthermore, there are underlying causes for this that can be proved objectively.  The Japanese MMA population itself is getting smaller in proportion to the rest of the world, as are the amount of fights taking place on Japanese soil — again, this is speaking proportionally.  There are also reasons for this that cannot be proven objectively such as the rest of the world’s embrace of the “weight drain” concept versus the Japanese fighters, who seem to be way behind in doing so, and maybe that other countries are simply “stepping up their game”.  Some potential reasons are in a gray area in that they cannot always be objectively proven in a consistent manner, such as those related to fight frequency, varying rules (Pancrase), different weight divisions, and so on.

Either way, the proof is clear.  The current rankings you see are overrating Japanese fighters by about 15%, and when I say Japanese fighters, I mean those from Japan fighting mainly in Japan.  Once fighters leave the country, this issue works its way out.

Now, the system is not first and foremost concerned with predictability, but there are predictive elements used in some situations and this is a desirable situation to use it in.  The core problem likely manifests itself  when fighters start with 10 points — on equal ground.  As we all know, all pro debuting fighters are not equal.  The best solution would be to start some populations of fighters above 10 points when they debut, but this “starter rating” is such a core principle that changing it would require a complete overhaul.

Given that, how do we get a handle on this?  The only fair method is to analyze past data and adjust ratings of these populations, in batches, once every few years.  This will be done retroactively for years already past  and we won’t be due for a wholesale “real time” adjustment for a few years.  This method would be no different than a statistician going back in time and using past data to project batting averages for baseball players who leave Japan to play in the MLB — except we cannot totally ignore Japan MMA, so we must consider it in the ratings and not leave it as a footnote, but it’s clear we have to adapt to better handle it.

Other nationalities will also be analyzed in the coming weeks, but only those with sufficient data.  Preliminary statistics show that Brazilian fighters may be underrated.  However, this may be more tightly linked to the weak reported fight histories (there are a lot of missing Brazilian fights).

Note:  The initial version will affect men only.