Sabermetrics (Moneyball philosophy) in MMA? Would using stats help a fighter win fights?

[King59]

would using sabermetrics as a tool work for winning fights or is it purely for betting odds?

 

[Foghorn Leghorn]

The thing with sabermetrics is that the stats that it's based off of have hundreds or thousands of repetition of the same events.

Yes some fighters fight a similar style, but not nearly like pitching and hitting in baseball. They also have one fight every few months, whereas sabermetrics needs way more detail to predict properly (hence baseball being perfect for it).

Stats tell you something in mma. It's just that you'd need to be factoring in everything in the gym to even begin to have a better idea

 

[nebulousned]

i was trying to write a program that would help me make bets but i gave up because it's hard to find a place to scrape proper fight data.

i scrapped all i could from wikipedia but it doesnt tell you if it was a fluke win a close decision, a decisive thumping, etc

 

[PolarBearPaulVarelans]

The thing with sabermetrics is that the stats that it's based off of have hundreds or thousands of repetition of the same events.

Yes some fighters fight a similar style, but not nearly like pitching and hitting in baseball. They also have one fight every few months, whereas sabermetrics needs way more detail to predict properly (hence baseball being perfect for it).

Stats tell you something in mma. It's just that you'd need to be factoring in everything in the gym to even begin to have a better idea

Yes, exactly this. Statistics work when we have a sufficiently large sample size for events with very little deviation.

That's why stats conquered baseball first. It's really more of an individual sport (you hit and field by yourself, not relying on others), so we don't have to consider interactions between teammates, an intrinsically more high-variance, rarer, and more difficult action to model.

That doesn't work for MMA. It WOULD work if every top fighter fought every single other top fighter about 40 times each. And if their change between said fights was minimal.

Then stats like takedown percentage, takedown defense percentage, strikes absorbed, and everything else would actually matter.

But right now, what the fuck does takedown percentage in a single fight against fighter X 2 years ago tell you about that success against fighter Y now? Absolutely nothing.

Certainly no more than one could already discern through watching by itself.

 

[KBE6EKCTAH_CCP]

The idea is appealing, but statistics cannot apply to MMA, because you need a significant enough sample to derive statistics.

The number of fights of an MMA fighter in his whole career is not even sufficient statistically speaking to infer anything, and we haven't even started to speak about variables such as fighting styles, reach, KO % of fighters, etc, etc.