I don’t think you can calculate in the traditional sense because an event only happens once and when they play again, it’s not under identical conditions. That’s why a lot of times trend stats are useless. Some of them are fine but the underpinning for them is usually a bad match-up and not some mystical unexplainable trend.
There’s two easy methods to compare
1) closing line value. Compare the line you bet vs. closing line, calculate edge vs it, plot your profit vs. EV and do a statistical test to compare the distributions (e.g., Kolmogorov–Smirnov). Cons: Assumes the market is efficient at closing, can also cause issues when books take “stands” and don’t really move even if they’re getting one sided action. Books really don’t exist to get equal action on both sides and collect vig. Most of them are sports bettors themselves. The Bookmaker lines maker did a bunch of podcasts recently and explains how they approach the market.
2) Edge and EV vs. your model. I prefer this approach. If you’re using a mathematical model, you should be able to calculate your edge and EV vs. bet. For example, a bet I made today was Tyrell Hatton +112 over Ancer and Hatton +109 over Ancer both for 5k. I have Hatton a small favorite in this (51.9%) so at +112 that’s a 10.664% edge and should expect to return $535.71 based on how much I bet.
The +109 has a 9.098% edge and should expect to return $455.64.
Obviously those aren’t options. Either I’m winning $5450 at +109 and $5600 at +112 or losing 10k (or pushing). but I can plot all my golf bets EV vs. actual and after a time they should be close. You can do the same tests to calculate difference between distributions (Expected vs. actual) to see if what you’re seeing is accurate.
Not to **** on datagolf because I think they’re smart modelers, just not good bettors, but look at their betting log.
https://datagolf.com/betting-results
All bets: Expected value: 112.5, Profit: 7.6 over 1086 bets.
Either they’re the unluckiest people ever or their way over estimating their edge. This isn’t a one-time thing. They’ve posted their bets for several years and while they’re small winners every year , they’re also running way below EV. This is worse because they’re using Kelly staking and overestimating your edge there can lead to ruin. You're going to see swings of course. There's a graph I posted somewhere in this massive thread which I can't find around July where after almost a thousand Pitcher K prop bets I actually was exactly at my EV. There was something crazy like a 12 cent difference between them. Also at the end of every season my EV on runs scored in first, HRE, full game money lines, etc. are all really close to the actual value.
Sample size is always an issue, but comparing your model EV vs. your profit is probably the best metric. If you're just betting based on some qualitative metric or a hybrid then it's impossible IMO.
Last edited by fkjlhfdkjhkj.; 06-21-2020 at 10:04 AM.