This year has been a relatively poor year in using the FightMatrix rating system for predictability when it comes to the UFC. Out of the 11 non-draw matches in UFC 127, the higher-rated fighter won in 5. We’re batting a tiny bit better on the year, but as it stands we are hovering around 50%.
.
Granted, the UFC does a fairly good job at matchmaking and the rating system here only has some elements of predictability. Still, I would expect the system to be closer to 60%, and hopefully we get there by the end of this year. In general, we are closer to 70%.
.
The Good
With the pre-bout ratings in mind, the “correct” bouts show a pretty good reflection of not only which fighters won, but by how they won. I’m particularly impressed with the system in regards to Quan/Reinhardt, a bout that would likely have caused a pure ELO system to fail. Even though I don’t like talking about fights in which one fighter was inactive — in this case, both were, and Quan had nearly triple the rating points of Reinhardt.
Fitch/Penn was an impressive result as well. I’m sure most people were doubting Penn’s initial placement of #4 at Welterweight, but a draw with #3 Fitch has cemented both their respective placements in the division.
.
The Indifferent
Fukuda/Ring was a questionable decision and also involved an inactive fighter. Ebersole/Lytle was a surprising result, one that a ELO system MAY have pegged, but I doubt it — not to mention Lytle’s recent knee surgery. I don’t think many people saw the Siver/Sotiropoulos result coming either. Hunt/Tuchscherer also shocked many and I have to credit the ratings in that Hunt was actually still a ranked fighter.
.
The Bad
Pearson/Fisher is the one result that the so-called experts were somewhat split on. The FightMatrix system went with the popular choice, but was wrong in the end. Fisher’s recent inconsistencies have caused him to jump all over the rankings (mostly down) as of late. Nevertheless, Fisher only had 26 spots on Pearson prior to the bout, and in the #50-100 segment of the Lightweight division — that is not a huge number at all.
.
Moral of the Blog
Don’t try to make a living by betting on the UFC! Stylistic differences in MMA also make software-based rating systems a bit generalistic by nature — not to mention that top fighters need to fight more often for a better picture of the ranks! An increase of predictability would surely make the rankings a total mess and by tradition, they need to show accomplishment over potential. Still, it’s worth noting there will always be up-and-comers, upsets, and stylistic differences that a computer will never, ever be able to predict.