MMA Sabermetrics: Isolated Striking Power (ISOS)

Obviously, this is a given. And we have had discussions about such problems. It is always going to be a problem when it comes to trying to come up with statistics in this sport, and it will affect the meaningfulness. Many statistics likely will have this sort of asteriks next to it.

The dataset is subjective because FightMetric is determing what is a power strike and what isn’t. You are right.

In baseball there are two primary stat providers: Baseball Info Solutions (BIS) and STATS. A great example of this is a fielding statistic called UZR developed by Mitchel Lichtman. Comparisons of the metric run on the two datasets show a strong correlation but it certainly isn’t 1:1 and the discrepancies are tied to the classifications of batted balls (primarily).

This is going to be somewhat mitigated by the fact that you’re setting up a sample size floor so bias from individual fights shouldn’t be as prevalent.

Wanderlei is very low, but this is only for his stints in the UFC. He’s at a 0.1599, which is nearly at the bottom. But this has me wanting to look at what it would be with PRIDE fights included, which I’m going to compile just to see where he is.

I can’t give out specifics to everyone as the data is protected, but Wanderlei moves up into the middle of the pack if his PRIDE fights are included. Over a .22

Oh yes both his UFC fights ended in the first round, I apologise.

but I’m not sure what the point of this statistic is yet. I can conglomerate a set of numbers that conforms to what I think a “leaders” list should look like but that doesn’t prove anything about the validity of the statistic.

If this is supposed to be a purely descriptive statisitc (i.e. who throws the most power punches) then that’s fine. Personally, I’d be more interested in how the data correlates from fight to fight. Does Anderson Silva have a higher ISOS in fights that end quickly? In fights that end by TKO? Does ISOS offer any predictions as to what round the match would end?

I don’t know if the data needed can be readily extracted from the set you’re working with, but I’d be interested in seeing an attempt to quantify the trade-off between going for a leg lock versus giving up position and absorbing strikes.

I’m going to assume it has to do with his style in his early UFC stint than his more recent fights.

It’s not going to be easy, and I’m trying to work through some basic statistics before I begin hedging a path through some of the more complicated formulas to determine things like defensive ratings and offensive ratings.

This stat is really one of the more basic statistics that’s easy to calculate and determine, but it is also serving as one of the first posts here to get the ball rolling on discussion and ideas.

Baseball has an advantage. Hopefully, we can come up with some meaningful statistics.

It would be silly if they weren’t.

No, these are all power strikes in the standing (distance striking) department, so head kicks would count.

The only reason a fighter would have “generally a higher chance of ending someone’s night” is because they do happen to LAND more power strikes in correlation to their own attempts. As I stated, there are exceptions in that a strike may look very powerful and be tabulated as a power strike, but the fighter may not have a history of actual fight-stopping power. Caol Uno and Frank Mir are perfect examples of that.

My own opinion is that the stat has some hinting toward a “higher chance of ending someone’s night”, but it isn’t an accurate theory obviously, as you pointed out.

As I stated, on paper, this stat should simply give fans an idea of how effective a fighter is in landing power strikes in the striking portion of his fights. The higher the number, the more effective the fighter has been in throwing power and landing in correlation to the amount of attempts they’ve thrown. Anderson Silva is landing a high amount of power strikes in comparison to his total attempts, which is fairly impressive when you see him at over .40

This is a good idea, however, trying to quantify strikes that “hurt” fighters is shaky ground. How would one judge the kick that Cung Le gave to Frank Shamrock that essentially ended the fight. The kick was clearly blocked, but broke his arm and subsequently ended the fight.

Baseball’s ability to delineate the value of an outcome (run expectancy matrix) is so unique that it’s hard to apply something similar to MMA.

I actually have some models for this… but I need to determine multipliers. Any system would work at first generally… but as WestbergIDFC states below… it’s tough to quantify strikes that “hurt” fighters. We do have a determination of that in our database however, so I’ll see what I can come up with.

We actually discussed this, and it really came down to a discussion of B.J. Penn’s jab. But these conversations are exactly what we’re trying to bring out with the project. We want meaningful discussions on these things.

It’ll take us some time to get there. We need some base stats before we can start determining predictors like PECOTA does or even some of the Bayesian models.

With your access to FightMetric reports, I’d love to see more posts like this.

Yes, this does have an affect on this stat as well. Some numbers are probably skewed a bit because a striker wasn’t able to keep the fight where he wanted it.

Obviously, ground fighters aren’t on this list, or are very low on the list. This is why we will likely have a separate clinch and ground stat for Isolated Power.

We already have all of those statistics, although I suppose they are readily available to the public yet.

If we include non-UFC fights for Uno, he drops considerably. I believe his number stems from his earlier UFC fights.