Cracking mma-betting using physiology

[Deleted member 26468]

I'm sure you've all seen moneyball. Maybe you've heard of Tony Bloom(owner of Brighton Football Club).
A few people(math whizzes) have hacked sports and gained an edge via extensive data analysis and algorithms.

I need to collaborate with some gigabrains. I'm a braincel, but I've collected samples of fighter faces and bodies from fight week(aka the week of the fight), photos are relatively standardized and from identical angles(several of them). I have a pretty robust sample size, but need people who are good in data analysis or things like python so we can build a model that will help us pick fights from just looking at weigh ins(weigh in footage is ususally readily available from most fight orgs). We simply need to identify large physical mismatches that aren't reflected in the betting lines. If someone is well versed in GMM that would be awesome.

We want to consider as many parameters as possible. Static and dynamic (bone/facial structure) vs muscle/fat mass distribution. Height (is it positive/neutral/negative?)

I have compiled fighter faces and physiques in two categories, all photos used are official portraits from fight week. The two categories are:
1: High ranking UFC fighters with very strong records photographed from several angles leading up to a fight they WON.
2: Low ranking UFC fighters with weak records. Photographed from several angles leading up to a fight they LOST vs another low ranking fighter.

I have blended these categories into an "average" composite face/body for the winning and losing sample of fighters.
I need more eyes on this, and someone savvy enough to make a model that can plug in either fight week photos or stills from weigh ins, and give us predicted winners.

 

[Deleted member 21765]

I always taught that MMA betting can be profitable. I also know python + data analysis (didn't build a ML model, but pretty sure I can do it). Can you send me 1-2 photos as a sample?

 

[JohnDoe]

Doesn't work like that. You play the line, not the fighter.

 

[Deleted member 26468]

Doesn't work like that. You play the line, not the fighter.

The money that sets the line doesn't do analysis of fighter appearance

 

[Deleted member 26468]

I always taught that MMA betting can be profitable. I also know python + data analysis (didn't build a ML model, but pretty sure I can do it). Can you send me 1-2 photos as a sample?

(winners are the 2nd pic for each sample)








afbeelding converteren naar jpg

 

[Deleted member 26468]

These pictures were roughly standardized by: pupil-pupil and pupil-mouth distance.
I didn't include any heavyweight fighters for the samples. I feel they should have their own category as they're less weight restricted.

I also have more sample collections, including things like before vs after USADA(drug testing) samples, where I compared the exact same sample of fighter against itself, but one sample was old fight week pictures of fighters who were great with no drug testing, vs fight week pictures of the same pool of fighters after they completely deteriorated post introduction of strict drug testing.

This is where autism is appreciated. What do people notice? For me, winning fighters have proportionately thicker necks, larger/wider craniums(above the ears), stronger brow development(especially outer brow ridge, making the brows look more masculine, aka more straight/negatively tilted), less fat in the lower jaw, darker/more saturated facial hair, slimmer more "athletic" torsos, significantly slimmer in the hip/adonis belt area. So pretty obvious things you might expect, but surely there is a way to identify when a fighter has all of these advantages relative to their opponent in a certain match-up? These may be rare, but these are the ones we need to identify.

Large height mismatches I think make the match ups less predictable, since grip strength and power correlates with BMI. So you could have a less in shape fighter still be physically stronger because of the more compact build.

 

[Deleted member 26468]

To the best of my ability I used the following criteria for choosing which fight week I gathered the pictures from: for winners, I picked fights in which they had a dominant performance which also resulted in win via finish (tko or submission). For losers, I picked fights in which they lost clearly, and which also resulted in a loss via finish. So, winners won via finish in a fight they were objectively winning prior to the finish, and the opposite for losers.

And let me reiterate; the winning sample is comprised of UFC fighters who are/were in the top 10 of their weight class, I made the sample consist of as many unbeaten fighters(at time of the picture) as possible. Losers were unranked(low ranked) UFC fighters. I picked portraits from fights where they lost primarily to other unranked/low level fighters. If anyone has any pointers to improve the whole methodology, I'm all ears.

 

[zharupodrugu]

MMA is too unpredictable, fighters are literal alcoholics and drug addicts. Box is way easier.

 

[Deleted member 20064]

Who told you about Tony Bloom?

You are clearly from UK then.

It’s easier just to have extensive MMA knowledge and not be a casual.

For example i knew Islam was going to win his last 5 fights and I am 100% right.

 

[helpmegod]

Who told you about Tony Bloom?

You are clearly from UK then.

It’s easier just to have extensive MMA knowledge and not be a casual.

For example i knew Islam was going to win his last 5 fights and I am 100% right.

yes but Volk winning was a real possibility, no one expected him to do what he did on the ground