Fight Matrix: How it all works.. and more! (Long Read)

We’ve got tons of info if you look for it, particularly in the FAQ page and also at the parent MMA Ranks page.  However, many viewers tend to not explore our site or are just unwilling.  As such, there is a lot of misunderstanding, misconception and overall wrong beliefs and perceptions about the site.  I thought it an opportune time to hit on the basics and also address some common errors made by site visitors.

Just to make it catchy, we’ll call it the Ten Rules of Rankings: Fight Matrix Edition (Oct 2014)

1. All FightMatrix.com rankings are generated by software.

FightMatrix.com by its most general definition, is a web site business that houses professional MMA rankings and statistics.  IT does not actually come up with the rankings or ratings and neither do the site owners.

Yes, a computer application generates the rankings you see on this site.  Yes, I single-handedly wrote the code for this program.  Yes, humans are subjective.  I get all of that.  The application actually came before Fight Matrix (the website) and is a completely separate entity which I created and dubbed CIRRS (Computer Intelli-Rating and Ranking System), just because I thought I’d give it a name — I like naming just about everything, ask my wife.  My habanero pepper plant has a name, for example.

Fight Matrix started posting generated MMA professional rankings in early 2008, but I started working on the system in 2007, peddling the results over at the BoxRec forum, where I cut my teeth being an apprentice of sorts, helping, or at least believing I was helping in refining the rating system used there.  Some of the ideas and concepts I came up with are still in use today.  For those of you that don’t know, BoxRec is an unbelievable resource for boxing results, rankings and it has one of the most intellectual combat-sports forums on the Internet.

Despite its short comings, I’m a big supporter of the UFC ranking system, as I started work on the now, CIRRS, because I wanted to become an MMA fan and I had no idea “who was who.”  The system helped me learn the sport and the sport helped me learn the system.  I came in with no favorite fighter, no favorite style… no subjectivity.

2. The system is both result-driven and predictive.

If I had to put a ratio on it, I’d say the CURRENT rankings/ratings are 65% result-driven and 35% predictive.

Certainly, when a fighter has had a long career at the time, results take over.  However, for fighters with young careers that have seen nothing but wins, sometimes predictability governs a sizable portion of that fighter’s rating points.

We’ve seen certain early career trends correlate with fighters being better than their rating and rather than ignore this, we consider it, but to a limited extent.

We also observe things like inactivity, quality performance decay (aka: career coasting), veteran profiling (aka: journeymen fighters who ticky-tack earn points only to get beaten everytime they step up), outcome weighting (KO worth more than SD) and even go so far as to consider individual scorecards.  I have written pages and pages of code dedicated only to handling particular types of scenarios.

3. The rankings are not division-specific, but division-adjusted.

When we first started the site and our database was weak, a rating point was a rating point and we simply segregated fighters by division at which we think they fought.  In 2009, the data got to a point where it was good enough that we determined this method had quite a few shortcomings in cases where fighters jump divisions, or worse, jump multiple divisions.  A formula was implemented then, which has since been tweaked, that supports the ideal that more size (generally) is a good thing.

This isn’t perfect either, as we sometimes see fighters get desperate as they get older, drop too far down in division and get wasted.  Lately, we’ve seen this with the Flyweight division, as it attracted fighters like Banuelos, Makovsky and Pickett – quality guys, that earned all their points in a division higher and ended up disappointing us a bit.  Admittedly, this has irked me, but this is the type of problem that eventually works its way out and is probably best left alone to do so.  Anyway, what I’m trying to say is that mainly, this works.  Basically, when a fighter moves division, we divide the top-limit (except for Heavyweight – a “reference” weight is used) of the divisions involved, apply a “power” to it and modify the rating.

I doubt we will ever consider division-specific ratings.  If we ever receive division information for every bout in existence, I may consider it.  Until then, we use the method above and apply a fairly conservative ceiling in cases when fighters move to a new division and don’t beat anyone of substance.

4. Every time we run the ratings, we consider all MMA professional bouts in history.

This is important, because fight results change, bad data gets corrected and also because old fights are still being uncovered.  When fighters move in the rankings for no apparent reason, this is usually why.  Every week, the ratings engine moves from bout 1 to present and re-considers everything, using the latest version of the software.  We also refresh the Generated Historical Rankings almost every week.

5. The rankings can only ever be as good as the data.

Sherdog is our main source for fight data.  For old and Brazilian results, they are unparalleled.  Their accuracy is also very good.  They seem to lack excellence in select areas of the USA — such as the southeast and midwest, but they do eventually add these results.

Tapology has recently become a great resource for timeliness and they excel in those US regions that Sherdog can sometimes fall short.  One important data point they have that Sherdog does not, is divisional info which sometimes includes weigh-in info.  Unfortunately, neither site posts scorecards, something we utilize — because a 50-45 UD does and should mean more than a 48-47 UD.

MixedMartialArts.com actually does have scorecard data, but their overall coverage is sparse, as rumor has it, they charge commissions money to insert their own results — part of the “fee”, for being the desired database used by the states’ commissions.

But wait, there’s more!  Bad data doesn’t necessarily equate to inaccurate results on online databases.  How about bad decisions by judges and refs?  How about ridiculous matchmaking where #400 gets lucky and beats #5…. psst.. #400 should’ve never fought #5 in the first place.

When we get criticism that is not constructive, I like to mention weather forecasting as a snide reply.  How many times has the weather been incorrect for your forecast and have you considered how many man-hours and weather stations/buoys are in existence?  This leads me into my next point.

6. Constructive criticism helps.

Telling us our rankings suck, doesn’t help anything.  You’re wasting your time, because I’m likely to delete your comment and move on.

Helping us find scenarios and/or patterns  that may lead to certain fighters being over- or under-rated and/or potential bugs IS helpful.  At the least, perhaps proper consideration was not given and a discussion begins.  At most likely, we may be missing or have bad data.  At worst, there is a software bug.

Proof is in the pudding:

A recent comment by a visitor in the Bantamweight section:

I wonder how you rank people. Matt Brown from Wisconsin at 135 isn’t ranked with a 8-4 record. His last 2 wins against guys you have ranked.John Castinada and Dominic Sindone. What’s the deal?

In actuality, Brown was 8-4 and we had him at 7-5.  His most recent fight (at that time) against Dominick Sindone was recorded as a loss and not a win.  We fixed this, and at next ranking, Brown was ranked at #182.  A simple search would’ve yielded we had the records of Brown and Sindone both wrong, but sometimes it is just easier to complain.

7. Yes, we calculate recognized title bouts with increased importance, if the bouts are competitive.

This is probably the most controversial aspect of the system.

The reasons for it are many, but simply put; title bouts that consistently match the top fighters in that particular community help provide data agreement.  You can read more about this in the aforementioned FAQ page, but this factor has the potential, NOT the guarantee, of helping both the winner and the loser of recognized title bouts.

The factor has been reduced over time, due to increase crossover from fighter communities (most notably US and Japan) and also because 5-round fights are becoming slightly more commonplace.  However, it is there.

8. A winner is always placed over the loser, but only directly after their fight.  Should either fight after that, it may no longer apply.

The most recent result is always given the utmost consideration.  Although styles make fights, unlike pure-mathematical rating systems, I decided long ago to honor the ladder-based philosophy, but not entirely.

Now, this condition may last longer than one fight because of something we introduced 2011…..

9. “Blocking”

I came up with the concept of blocking in 2011.  Prior to this, Fighter A could beat Fighter B by KO1 and technically, Fighter B could then beat 5 journeymen and soar past Fighter A.

A sidebar, if you want to see all the “bad” of not even having a semblance of the ladder-based philosophy, relying on a system that excels in predictability and does not have blocking, please check out MMA ELO.

I commend the efforts, but having Douglas Lima ranked above Robbie Lawler is not acceptable to me in any system.  Nor is having Ronda Rousey ranked #5 at Men’s Bantamweight or Ronaldo Souza ahead of Chris Weidman.  You can pick apart our rankings, no doubt… but to have such egregious, obvious errors means it may be best to start over.

Anyway, “blocking” essentially adds protection to the scenario in the first paragraph and ensures Fighter B has to at least do something credible in order to pass Fighter A.  Yes, it has downfalls and can sometimes produce oddities, but it is better than nothing.

10. We plan on sticking around for a while

OK, so I was out of legitimate things to dedicate a point too, but I think it is safe to say that the two-man team (Oleg and myself) that runs this site make nothing close to a full-time income for our efforts.  I guess for #10, I could’ve mentioned how we post an update every week like clockwork and list up to 600 fighters per division (and growing), but you probably already know this.

We’ve been around for nearly 7 years and I’d like to at least get to 10.  BTW, our 400th official current ranking update will be coming up soon.  At this point, it’s like a baby for me that I’ve watched grow up and we still have debates over the most minute rating changes… so I think its safe to say that we’re both still invested in this thing.

Furthermore, we’re still discussing and working on enhancements to the site’s content and giving consideration to potential ratings changes.

Bonus.  A “potential” future change to the ratings

You may know about our community adjustment, which is mainly for the Japanese population.  For whatever reason, we always overrated fighters in Japan, sometimes criminally.  Perhaps this was due to fight frequency or wrinkles in their scoring, or the fact that they don’t cut weight — leaving their Bantamweights fighting foreigners that enter the ring/cage looking like their Lightweights.

This was proven by the times that they would leave for seemingly greener pastures, only to (almost always) go down in flames to lower-rated fighters.  We resolved this by handing out periodic deductions to qualifying fighters.  At current, things on this front are almost close to stable.

Now, something else has reared its ugly head, which has always been there, but stronger than ever.  The dreaded “home advantage”.

On average, we’ve found that “home advantage” in MMA is nearly negligible.  Whether due to careful matchmaking, fan presence or whatever else, Brazil, Russia and especially Poland are not friendly to outsiders…. not at all.  Brazilian and Russian fighters almost always hold their own when leaving their country, but Poland is another story.  The data isn’t there, but certainly outsiders come in and typically suffer a losing outcome.

It seems rather selective to pull Poland out of the hat and point the sole finger at them, but they are the only ones that really stick out and its likely due mainly in part to the KSW promotion.

I’m not promising or suggesting any changes are near, but it is the top point of debate internally.

The digital industry does not stand still and you can also search people by name if necessary.