ML in sports betting

>>8584330
When I need more money I just go to vegas, put $10000 on red. If I win, I leave (I made a profit). If I lose, I put my net losses on red and repeat. Gauranteed win.

>>8584337
:*(

>>8584330
Might work for horse racing since a lot of the horses are of the same bloodlines, fed and trained the same, etc. Its not a good idea for other sports without consideration of a fuckload of variables including day to day stats on the individual athletes because you'll probably train your algorithm to just always bet on whoever is winning instead of what makes them winners, such as training regimen, staying clean in season, proper sleep, not facing off the field/court drama, raw athleticism, age, their sports IQ for their position, their ability to work in a team, status on their relationship with coaches and teammates, if the coach has anything going with them,...

Way too many variables that can screw things up, and if you use something as simple as box scores and season stat sheets for things like passing yards or sacks, you won't be able to pick winners. Closest you'll get is being able to know for sure when some team is about to get BTFO by another team that completely out classes them all around.

Would be interesting though if you train it to browse news outlets and twitter profiles of athletes to try and get a read on if they're mentally there week to week for a game.

>>8584423
interesting thought on the mental thing.

also. one of my friends has a pretty solid baseball algorithm. makes good cash for him, pretty robust. Although baseball is a bit easier.

yeah, i see what youre saying about anonymous'ing the horses. thats what i'm doing for part of my ensemble. it's a neural net with similar horses

>>8584480
>>8584480
bump

>>8584330
I bet on UFC matches and so far have never lost, the majority of my 'skill' is just looking at somebody's coaches during camp and comparing their coach's stats of previous and current fighters they've trained.

For example Micheal Bisping's boxing coach had an excellent record of producing top fighters, so I bet on him to beat Rockhold and won $6,000 which was considered a foolhardy bet, but stats didn't lie.

I thought about making this into an algorithm but lately I don't have any time to watch fights or do research of camps for a while, plus some camps try and keep secret who they hire which is annoying.

>>8584670
interesting to note. How did you get the original data for the Bisping vs. Rockhold fight?

>>8584690
Found out through a Hoe Rogaine podcast that he was using the same boxing coach Cyborg was using. Researched that coach and discovered all his previous fighters, compared their stats to Rockhold's camp, threw down money.

For a while I tried to make a tautology/satisfiable logic argument out of these stats and it sort of worked but I got sidetracked and had no time to debug them. There is also much more wild variables in fighting than there is say, team sports where this would work much better by looking at all the asst coaches of a team and comparing their successes.

I've found it's coaches that make the difference, having super great players/fighters helps but as long as they are good enough to not give up you can bet on them if their coach's have superior records.

>>8584705
interesting. good to note man.

if there was a way to get the data easily, I could help you write an algorithm for it pretty quickly.