Created a MMA fight Picker GPT

[TCE]

Here is an example how the new model differs when making it picks.

ChatGPT 4o (prompting)
Winner: Lerone Murphy
Round: 3
Finish: Decision
Confidence: 65%

Fight Analyst Advanced
Winner: Edson Barboza
Round: 3
Method: Decision
Confidence: 60%

Just run it through with the new GPT, remember the Murphy/Barboza fight is 5 rounds main event.

Put it through 3 times. It had for me Barboza by Decision twice and Barboza by TKO in round 4.

 

[TSaL]

Just run it through with the new GPT, remember the Murphy/Barboza fight is 5 rounds main event.

Put it through 3 times. It had for me Barboza by Decision twice and Barboza by TKO in round 4.

Yep dam it forgot to add "5 round" at the end of the fighters names.

 

[TCE]

Yep dam it forgot to add "5 round" at the end of the fighters names.

Haha, what result did you get when corrected?

 

[TSaL]

• Predicted Winner: Lerone Murphy
• Predicted Round End: Round 5
• Predicted Finish: Decision
• Prediction Confidence: 65%

 

[TSaL]

I don't think we will see much increase in accuracy until gpt5.The have introduced memory but I don't think the custom GPTs have it. I will work on my instruction set to be a as big and through as I can get it before the inevitable release of something better.

 

[TCE]

I don't think we will see much increase in accuracy until gpt5.The have introduced memory but I don't think the custom GPTs have it. I will work on my instruction set to be a as big and through as I can get it before the inevitable release of something better.

Memory would be a very handy tool for this, with fight records etc. I think the last one only remembered until 1 and a half year ago.

 

[MichiganMMA1978]

I don't think we will see much increase in accuracy until gpt5.The have introduced memory but I don't think the custom GPTs have it. I will work on my instruction set to be a as big and through as I can get it before the inevitable release of something better.

Just go straight to the source and ask @SKYNET

 

[SKYNET]

Just go straight to the source and ask @SKYNET

 

[TSaL]

lol 40 is fukin poo. geeez

ok well it got the next 4 right...

 

[sonhow]

Alright gpt-4o dropped a few hours ago so I tried it out using prompt only. Here are the results for this upcoming fight:

https://www.tapology.com/ue/110359-4114857.png

Heeey thats pretty good once again. Although its concerning how it picked Bram over Themba. Williams a good pick in 50/50 fight and Murphy as well.

 

[Hazuki Ryo]

Here what's it saying for tonight's fights:
Predicted Winner: Magomed Ankalaev

Expected Round: Round 2

Manner of Victory: Knockout

Detailed Analysis:​

Magomed Ankalaev:​

1. Weight Cut Amount: 85 - Ankalaev cuts a significant amount of weight but manages it professionally, indicating a high level of conditioning.
2. Fighter Reach: 90 - With a long reach, Ankalaev can maintain distance and strike effectively.
3. Training Center: 88 - Training at a top-level center, he receives excellent coaching and sparring opportunities.
4. Natural Weight Class: Light Heavyweight - Score 95. His natural weight class aligns perfectly with his fighting category, giving him a natural advantage.
   Total Score: 358

Johnny Walker:​

1. Weight Cut Amount: 78 - Walker cuts a substantial amount of weight but has shown signs of struggle in the past.
2. Fighter Reach: 85 - Walker has a good reach, but not as advantageous as Ankalaev in striking exchanges.
3. Training Center: 82 - Training at a reputable center, but slightly behind in terms of resources compared to Ankalaev’s training environment.
4. Natural Weight Class: Light Heavyweight - Score 92. He is in his natural weight class but has not optimized his performance as well as Ankalaev.
   Total Score: 337

Conclusion:​

The data suggests that Magomed Ankalaev holds a significant advantage in reach and conditioning, with both fighters competing in their natural weight class. Ankalaev's superior training and weight management are likely to contribute to a knockout victory in the second round.

It was right !!!!! damn

 

[humperdoo]

is there a version of this for boxing?
it must be great to also have such AI just to see whether out speculation goes along with the evaluation of the AI.