Created a MMA fight Picker GPT

[TCE]

This AI fight picker is actually fucking amazing haha. I'm top again for the event on my picks league, and everyone on there picked Pyfer, except me. Another nice £100.

 

[theincognito]

This AI fight picker is actually fucking amazing haha. I'm top again for the event on my picks league, and everyone on there picked Pyfer, except me. Another nice £100.

I'm pretty sure it has correctly picked main and co-main on every event I have tested it on thus far with the exception of Strickland (and many think he should have gotten the decision)

 

[TCE]

I'm pretty sure it has correctly picked main and co-main on every event I have tested it on thus far with the exception of Strickland (and many think he should have gotten the decision)

It seems to get most the fights on each card correct, as well. Including correctly picking the round and method on many.

Since I've been using it, I've only had like 2 or 3 of my picked fighters lose, each event.

 

[DalchaLungiambula]

This AI fight picker is actually fucking amazing haha. I'm top again for the event on my picks league, and everyone on there picked Pyfer, except me. Another nice £100.

How did you manage to work through the problem of it seeming to changes it's picks every day?

 

[TSaL]

first place in both my leagues and 85th overall.

 

[TSaL]

How did you manage to work through the problem of it seeming to changes it's picks every day?

I have been trying to make it more consistent if you enter in the same fight. What might help if you add the rounds for the fight (example: Fighter1 vs Fighter2 5 round title fight)

 

[DalchaLungiambula]

I have been trying to make it more consistent if you enter in the same fight. What might help if you add the rounds for the fight (example: Fighter1 vs Fighter2 5 round title fight)

Out of interest how long would it take a layperson to build a "specialised" ChatGPT bot like yours from scratch?

The ability to gather data in such a large scale and in a relatively unbiased manner could be game changing in so many different pursuits.

 

[theincognito]

How did you manage to work through the problem of it seeming to changes it's picks every day?

IMO that is useful because it's pretty much saying those fights are too close to call. But it will give the same answer 99% of the time when it's sure. It predicted Hermansson every single time I asked it about last night's card (and he was quite a big underdog).

 

[wwkirk]

Created a MMA fight Picker GPT: https://chat.openai.com/g/g-a4l5RVxni-fight-analyst-advanced

I gave it instructions to look at a fighters then compare then on a few different criteria and give them a weighted score from 1-100. I wouldn't use it to bet on fights or anything. Let me know what you think and if you have idea on how to expand it.

Are you a sophisticated programmer or something, or is this GPT the AI version of an Angefire build-your-own website?

 

[TSaL]

Are you a sophisticated programmer or something, or is this GPT the AI version of an Angefire build-your-own website?

Just a simple minded man here.

 

[TCE]

How did you manage to work through the problem of it seeming to changes it's picks every day?

I ran all the fights 3 - 5 times, and went with the fighter who the AI picked the most. Mind you, think there was only 1 or 2 fights it couldn't decide on.

The Hermansson/Pyfer fight, I put through like 7 times, and it picked Hermansson every time. I was very skeptical at first, but thought fuck it, I'm picking Jack... This AI tool is amazing, and it's only going to keep improving from here.

I've credited the TS loads here, but he's created an absolute gem.

 

[TSaL]

Here is our current standing in topology:

 

[WoozyFailGuy]

Here is our current standing in topology:

View attachment 1029485

Excellent picks thus far, made money on a little parlay with it.

 

[TSaL]

https://www.tapology.com/ue/107060-3912071.png

 

[Vulve]

hitting 78% is actually insane, im extremely impressed

 

[Levi_]

https://www.tapology.com/ue/107060-3912071.png

View attachment 1029820

Have you calculated how many of its correct picks were favorites vs underdogs? It would show far greater predictive value if it can choose underdog wins consistently (and give us motivation to bet based on these picks).

 

[TSaL]

Have you calculated how many of its correct picks were favorites vs underdogs? It would show far greater predictive value if it can choose underdog wins consistently (and give us motivation to bet based on these picks).

I think the last main event was an upset it guess correctly.

 

[sonhow]

https://www.tapology.com/ue/107060-3912071.png

View attachment 1029820

Hmmm first time I'm seeing some big differences with my own thoughts....
Let's see how well it holds up. This is the biggest test of the ai picker so far in my humble opinion lol.

 

[TSaL]

Hmmm first time I'm seeing some big differences with my own thoughts....
Let's see how well it holds up. This is the biggest test of the ai picker so far in my humble opinion lol.

I'm using a new AI gpt that I built, but I do select a few fights to run them twice and run on the original fight picker gpt. Most of the time they agree and when I find one that doesn't I will ask it (gpt) how I could modify its instructions to fix this issue. Then I add the instruction to the gpt. They are almost building themselves.

 

[Levi_]

I think the last main event was an upset it guess correctly.

Looks like it did. That might be a good thing to track.