I’ll have Fight Scores up on this blog as soon as I get the Fight Metric data. I also have plenty to say about the main event, but first I need to talk about my picks and FPR…
If you want to throw tomatoes at me for the VERY high-confidence projections for Chael Sonnen and Rory MacDonald, I don’t blame you. Part of me wants to throw tomatoes at myself.
There are clearly some flaws in my model and how I determine who will win fights. There is no excuse for coming out and saying Chael Sonnen had an 89 percent chance of beating Rashad Evans. Even though I said myself that I didn’t buy that projection, my model should never say that in the first place.
So there are no excuses. My picks straight-up really weren’t all that bad (8-4) but the problem is in the percentages. The fighters FPR recommended a bet on didn’t do well at all… accounting for three of the four losses. The fighters I recommended were 3-3 tonight but again, the two highest confidence picks went down in flames.
With that said I’ll say a couple things in my defense. This blog is based on stats and FPR in particular is based on Fight Metric stats. I defy you to go to the Fight Metric stats for Chael Sonnen, Rashad Evans, Rory MacDonald, and Robbie Lawler, and tell me with a straight face that the stats favor Evans and Lawler to win their respective fights. You can’t do it. The stats clearly indicate that Sonnen and MacDonald are better fighters.
Now, I get it. Statistics need context. Renan Barao’s statistics might be better than Joey Beltran’s, but that doesn’t mean Barao would beat Beltran in a fight. I would love nothing more than to break down each fighter’s statistics and separate them by weight class, by opponent, and by where the fight takes place (top position, bottom position, standing distance, clinch, etc.). Perhaps my model should have looked at what happens if Sonnen doesn’t take Evans down. If that was the case the model might have recommended a bet on Evans. It’s tough to say.
The problem is: I don’t have that data. It’s not available on Fight Metric. If I want to get that kind of data, I need to go through the tape of the fights and collect it myself. I don’t know about you, but I don’t have the time to do that for every fighter. I wish I did but I don’t.
The other problem is: the statistics are already so limited for most of the fighters already. When there is a small sample of data, taking segments of that data only makes things worse. In MMA the biggest obstacle to proper statistical analysis is a lack of data. It’s not a good idea to exacerbate that problem by only looking at portions of the already limited data I have.
Does that mean I shouldn’t make any changes? Obviously not. There are changes that absolutely have to be made to my model. I’m not 100 percent sure what those should be yet, but I have some good ideas. One thing is for sure – you’re not going to see as many fighters considered a 90 percent favorite to win from this point forward.
All of that said, FPR still has a very good overall record, both with straight-up picks (unless you think being ranked #23 on Sherdog is bad) and against the spread (although today put a major dent in that). I can’t revel in the good times and then panic and change everything during a cold streak. That’s just bad form.
So if you’ve lost faith in my ability to predict fights, I don’t blame you but even the best have an off night now and again. I don’t claim to be the best but I still think I can contribute valuable insight in the MMA community. Time to look “under the hood,” tighten some bolts, and come back with a better model next time. See you then.