Win Probabilities... UFC: Weidman vs. Gastelum

latoya johnson

Double Black Card
@Green
Joined
Mar 27, 2016
Messages
1,416
Reaction score
323
upload_2017-7-22_12-55-34.png

Performance Review of Model, historical:
1. Correctly picks winner in 77% (96/125) of bouts;
2. Generates an average profit margin of 19% for all bets;
3a. Generates profitable bet spreads for 79% of events;
3b. Generates unprofitable bet spreads for 21% of events.


Performance Review of Model, UFC: Nelson vs. Ponzinibbio
1. Correctly picked winner in 75% (6/8) of bouts;
2. Generated a profit of 1% :(
floyd1.gif


Notes:
1. The model gives Weidman a mere 18% chance of victory. My intuition insists this is wrong, but I've learned with time to favor algorithms over intuition.
As a Weidman fan, I almost hope it is wrong. It's possible that the model is underestimating the significance of his size advantage.

2. The model ardently disagrees with oddsmakers concerning the win probability of Anders over Natal.
The model gives Anders an 80% chance of victory; oddsmakers give him a 52% chance of victory.
One explanatory factor is that Anders is a late replacement, but the algorithm doesn't account for this disadvantage.


LOCK OF THE NIGHT: Burgos @ 88%.
 
1. The model gives Weidman a mere 18% chance of victory. My intuition insists this is wrong, but I've learned with time to favor algorithms over intuition.
As a Weidman fan, I almost hope it is wrong. It's possible that the model is underestimating the significance of his size advantage.

Your intuition is right. There is no way Weidman is that big of a dog.
 
Your intuition is right. There is no way Weidman is that big of a dog.
Another explanation is that Weidman is better than his recent record suggests.

...but that's just our intuition. It can't be quantified.
 
Another explanation is that Weidman is better than his recent record suggests.

...but that's just our intuition. It can't be quantified.

I dont know how you model works but if it treats every L the same, then yeah its pretty flawed, and there is no reason for you to not consider that in a match up like this where it is very obvious it should be considered. Weidman has been very competitive in all of those losses and Kelvin has not yet fought any opponent of the caliber Weidman has lost to. Also the size difference is very significant like you said. I can live with Kelvin being the slight favorite, though, because of Weidmans cardio.
 
Your "model" is really messed up. Your % are all wrong for many of the fights.
 
I don't really understand the point of this if you aren't going to share your math?

You don't even note the different variables you are factoring in, even if you don't want to specify the weight you give to each


View attachment 252441

Performance Review of Model, historical:
1. Correctly picks winner in 77% (96/125) of bouts;
2. Generates an average profit margin of 19% for all bets;
3a. Generates profitable bet spreads for 79% of events;
3b. Generates unprofitable bet spreads for 21% of events.


Performance Review of Model, UFC: Nelson vs. Ponzinibbio
1. Correctly picked winner in 75% (6/8) of bouts;
2. Generated a profit of 1% :(
floyd1.gif


Notes:
1. The model gives Weidman a mere 18% chance of victory. My intuition insists this is wrong, but I've learned with time to favor algorithms over intuition.
As a Weidman fan, I almost hope it is wrong. It's possible that the model is underestimating the significance of his size advantage.

2. The model ardently disagrees with oddsmakers concerning the win probability of Anders over Natal.
The model gives Anders an 80% chance of victory; oddsmakers give him a 52% chance of victory.
One explanatory factor is that Anders is a late replacement, but the algorithm doesn't account for this disadvantage.


LOCK OF THE NIGHT: Burgos @ 88%.
 
I feel like Kelvin will win this
 
I dont get how +145 weidman translates to 18%, but +175 for elkins translates to 33%.

fail thread/
 
Back
Top