Years ago, Frank Rosenthal used to post football and basketball picks on his website. He died last year and in his later years he didn't give out many winners. But I learned from keeping notes on his picks over the years that he always knew when he had a winner. In fact, when he posted picks without a write-up, they weren't worth betting on. But when he got into explaining why he liked a certain game, the results were phenomenal. For instance: One thing that I can say about the way he was doing it was about the pointspread itself that he was looking at. Whether the number was a 6 or a 7 made a difference to him and I noticed that it more than just finding value. He also mentioned when a certain positional player was out and from keeping track of everything, I tried to figure it out.
I have been hooked ever since on figuring out how he was doing that and I am happy to say that I have found that it could be about building algorithms from simple trends.
You have to keep it simple, like when a team is playing without it's most productive running back and the entire league has been losing when it was in the same situation, or a when a string of 7-point underdogs all won their game, it is a simple trend. When you first start hearing rumors that the coach is going to be fired, or something in particular that a coach said and it suggest the team will win. The entire league follows patterns like that, and the algorithm is about putting a number of simple trends together on one game to ascertain who is going to win that game. Not betting on every trend because they carry no weight unless they associate with others and conglomerate onto one play and that's when you have the algorithm.
I have been hooked ever since on figuring out how he was doing that and I am happy to say that I have found that it could be about building algorithms from simple trends.
You have to keep it simple, like when a team is playing without it's most productive running back and the entire league has been losing when it was in the same situation, or a when a string of 7-point underdogs all won their game, it is a simple trend. When you first start hearing rumors that the coach is going to be fired, or something in particular that a coach said and it suggest the team will win. The entire league follows patterns like that, and the algorithm is about putting a number of simple trends together on one game to ascertain who is going to win that game. Not betting on every trend because they carry no weight unless they associate with others and conglomerate onto one play and that's when you have the algorithm.
