Intro, accept it
One of the biggest issues with ranking MMA fighters is a mistaken belief that certain fighters are unbeatable. People think that the #1 guy never loses, and when he does they think the guy that beat him is now Captain Indestructible. History, statistics, and common sense would all beg to differ though.
For those that understand MMA is a sport, and even the best can be upset I’d like to move on and talk about expected win percentage (it is what I titled the rant after all). When I say expected win percentage I’m talking about the overall percent of the time Fighter A is expected to beat Fighter B.
For example, if I say Fighter A has a 60% expected win percentage against fighter B that means if they were to fight 10 times, Fighter A would be expected to win 6 of them. It’s certainly possible that Fighter A wins more than 6 times, or even loses every fight. However, the stat is telling us Fighter A should win 6(ish) out of 10.
This concept of expected win percentage is very important on the site because it’s the basis for a lot of comparisons. Under the covers the site looks at Fighter A and their rating, then looks at Fighter B and their rating. From those numbers it makes a determination about how often each fighter should win. Based upon that calculation and the actual fight outcome it then adjusts both fighters ratings accordingly.
History, learn from it
One of the things I try to do with the site is provide a level of transparency that rivals the partial tax returns offered by presidential candidates. One way I do this is via the Site Stats page. This page has a table that shows the results for fights across various rating gaps. For instance, if we look at the page quick (it’s ok to click it real fast, there aren’t any nasty roll over ads that are going to bite you) we’ll see that when fighters are within 70-80 rating of each other the underdog wins 40% of the time. That is to say that 6 out of 10 times the fighter rated 70 to 80 points higher than their opponent wins the fight.
Dance or no dance
Much like how some people have problems believing their favorite fighter can’t be beaten, many people have a problem understanding how the site works. They see the 60% number from the 70-80 rating difference and think the results are poor. What people need to realize is that the comparison that should be made is between how often the system thinks a fighter of a certain rating should win and how often that fighter actually does win.
There are a few ways I could illustrate this: a song, interpretive dance, or a bland HTML table. For now I’ll post the table, but keep your eyes out on youtube for “Expected Win Percentage by Evil Pooh and the Pigtails”.
| Range | Difference |
|---|---|
| 45 – 55 | 3.35% |
| 95 – 105 | 3.95% |
| 145 – 155 | 4.6% |
| 195 – 205 | 3.42% |
| 245 – 255 | 0.89% |
| 295 – 305 | 5.97% |
| 345 – 355 | 0.82% |
What the above table shows is the difference between the win percentage the site system expects and the actual win percentage achieved by fighters. For example, looking at fights where the fighters had a rating gap of 195-205 points the site was off by 3.42%. As a simple example, assume the site expected the higher ranked fighter to win 66% of the time, in actuality the higher rated fighter won 69.42% of the time.
Looking at the above it does appear that the system is a bit conservative in that historically the higher rated fighter has won a bit more than expected. However, the gap is only a few percentage points in most cases.
What have you done for me lately
The above table showed the results for all fights ever, so lets take a look at a more recent sample. This table will only show the results of fights that occurred since 1/1/2006.
| Range | Difference |
|---|---|
| 45 – 55 | 4.87% |
| 95 – 105 | 2.45% |
| 145 – 155 | 1.51% |
| 195 – 205 | 1.47% |
| 245 – 255 | 1.65% |
| 295 – 305 | 9.96% |
| 345 – 355 | 1.06% |
Again we see that for most rating ranges the results are within a few percentage points even for the most recent fights. The one exception is at 295-305, but due to the much smaller sample size it’s not as out of place as it first appears. Additionaly, it again errs on the side of the favorite winning.
And I read this why?
For a couple reason actually. Here’s what you should take away from this little rant:
- Fighters lose, even the best of the best of the best SIR~!!! Don’t think a guy is over the hill because of one loss and don’t think a fighter scoring one huge win makes them Captain Bignasty.
- Numbers need context. If the site shows Fighter A has a rating of 1802 and Fighter B has a rating of 1804 that does *NOT* mean that Fighter A has no chance in hell of winning so much as a round from Fighter B. It does not mean the site is claiming Fighter B will win every fight the two ever have. What it means is that the fighters are very close, but historically the higher rated fighter has won slightly more often.
- The site is running very close to where it should be from a statistical point of view. Don’t look at the 60% number and assume the site got 40% wrong. The truth is the site is actually getting a few too many fights right atm. *shrugs* What a terrible problem to have. 8P
To infinite and beyond
So where do we go from here? I’ve got a few tests to run on the system itself to lower the expected win percentage gap a bit. I’ve also thought about potentially adding the historical and maybe even the expected win percentage numbers to the Fighter Compare section of the site. That might help some people better understand just how closely matched some of these fighters are.
Should I add the win percentages to the fighter_compare? Are there any other site metrics you’d like to see? I had some from 2007 (year of the upset) which overall was about as expected. Just remember that lowering the sample size really skews the results. Don’t expect to see metrics for April of 2003 (although, yeah, I could provide them lol).
I like the idea of expected win percentage on the fighter compare, but having it show up only if the fighters are from the same weightclass, and both have at least 6 fights or so.
I’m a pretty huge fan of your site and have recommended it quite thoroughly to the people on sherdog. I would really encourage you to add the odds to the fighter compare, one for career, and 1 for 3 year accomplishments, as it helps in figuring out who one should bet on (Currently I just look in your sight stats to see odds etc, but the compare feature would be convenient).
MMA-elo is the only rankings that I even consider looking at.
That’s a good suggestion farnum. I’ll propbably put weight class and a couple other checks to decide whether to display the information or not.
Derek – Glad you enjoy the site and get some use out of it. 8)
Great! Thank you!
I always wanted to write in my site something like that. Can I take part of your post to my site?
Of course, I will add backlink?
Regards, Timur I. Alhimenkov