tradesports.com

[bigpooch]

Does anyone know someone who makes a living trading sports
bets? It seems if someone is good at handicapping, he
should do exceptionally well with a decent bankroll. I
haven't opened an account at tradesports.com yet, so I just
want to know if any 2+2ers out there have an account there
or knows someone who does.

[biggiemcg]

I have an account at tradesports. I have been trading for about a year now. I have a model to predict win prob during a baseball game and post bid and sell offers on either side of my prediction during games. It's a lot of fun, but I don't have anywhere near the bankroll to make it a living. If you did and wanted to dedicate the time to creating an accurate handicapping system that can be quickly updated, you could definitely make some good money.

[Wildbill]

It would seem like there are a lot more variables than a simple chart could handle? I mean if you have the Dodgers up a run going to the ninth it is a lot more secure than the Royals right?

I am sure you could beat the games for a little, but the vig is quite high in their system. Further I am sure much of it is very mechanical and probably a lot based on your idea of just creating a probability chart and figuring the vig protects you should you be off a bit.

[bigpooch]

Although I read the other post, I don't believe you simply
have a chart but more likely a mathematical/statistical
model. If so, did you spend much time back testing on the
available data? If not, maybe it's just a hobby for you.
In any case, thanks for the input!

[biggiemcg]

Yes, it is a statistical model and not a "chart." I have data on inning by inning scoring for every game for the past 10 seasons (this and more available at retrosheet.org!!!) I actually have many seperate models, one for the end of each half inning. I ran a logit regression on the dependant variable homewin, with my main explanatory varibales being each teams score, the gambling odds before the game, the total number of runs scored to that point, and the gambling total (O/U) before the game. I also interact these variables in several different combinations. During a game, I plug the current game variables back into my formula. I then adjust the result by a few points if one team has a particularly strong portion of it's line-up coming up that inning, or one team chose to pitch a relatively weak reliever. I have tested the robustness of the model, by running my regression on smaller portions of my dataset, and it seems to be fairly robust. And yes, I do count on the vig cusioning me against any small errors that I might have made.

[bigpooch]

[mosta]

I've been interested in applying some stat modeling to intra-game markets--specifically, time-series models--because I assume that intra-game processes are less well studied (intra-game betting is new to online isn't it?), and because I want to learn ARMA modelling for personal curiosity. Anyone have any knowledge of attempts to use pure time series models? I was thinking of using basketball. I'd want to have data minute by minute (I can probably find a place to buy this, or maybe trying to automate extraction would be an interesting project itself), and I'm hoping to find that leads are over or undervalued.