Quote:
Originally Posted by TomG
Great opportunity to work through an example problem together. I'm thinking we could improve daringly's rough calc by adding league sacks?
BAL Sacks Allowed 1.5
BAL Sacks 2.7
NYJ Sack Allowed 2.9
NYJ Sacks 2.4
League Avg Sacks Per Game 2.5
Next we can think about what variables predict sacks to create our own xSacks metric. If we can all improve our handicapping by just 1% every week by the power of compounding soon we'll all be crushing
If you focus just on Baltimore forced sacks... 2.7 vs league average of 2.5, and NYJ allowed 2.9 vs average of 2.5, there are a lot of different ways to play with that. Additive, multiplicative, log 5, and more complex (but not necessarily better) approaches. We could build a multi-year database and build a precise xSacks model. We might believe the true answer is .61, .625, or what have you. We could put in our $250 bets and pound fists on our chests. "I captured 10% EV! $25 grocery dollars!"
Is it worth doing? Yes, but not so that you can place $250 bets. The methods (for me anyway) aren't worth the time, unless I can attack a much bigger market. Anyone can win on props with back-of-the-napkin calculations. Everyone recognizes -125 Under 5.5 as a great bet, but no one is taking -150.
The biggest successes I've had are with giant historical databases where you can really look at relationships. How much is outlier, how much do you expect it to revert? What is the best relationship between many variables for future prediction? If we could bet $10k on sacks, an xSack model would be a great project.
If you all are going to spend the mental effort, why not attack something worth beating?
If I were advising a new syndicate what to attack, I'd say go after any sort of in-game betting in a major sport (the big US 5, Soccer or Tennis). The market limits are huge, and the lines are still very weak. For sportsbooks that publish holds, the typical book hold for in-game is about 1/2 of the pre-game hold.