[MarketMaker]

I am becoming quite obsessed with trying to figure out if models I have created are to be profitable long term and have seen Justin7 mention that how the plays would have done if bet as action points is a possible indication of this.

I would like opinions on if the following results indicate any kind of edge going forward or if this is simply not enough information.

W 73
L 55
Avg W 8.438
Avg L -7.290
Total Win 215

W 76
L 52
Avg W 10.993
Avg L -9.759
Total Win 328

Now neither of these include the vig for the loses which would be -40.1 for the first set and -50.75 for the second. One set constitutes plays on the 1h and the other 2h.

[Pokerjoe]

You might want to reconsider whether you want to weight a 10 pt variance from the spread as double that of a 5 pt variance.

Still, it's better than weighting a 1/2 point spread victory as equal to a 20 pt cover.

[bztips]

Read Ganch's post on z-scores here

[That Foreign Guy]

Most sports teams don't play to the spread so for weighting I would suggest you look at the game strategy - what milestones do coaches use when making decisions? For example in the NFL they might be 0.5, 3, 7, 10, 14 because they're a certain number of scoring plays (so a 9 point win is equal to a 8 point one but better than a 7 which is better than a 6 etc) .

Another option might be to rate significance by the % of times a team comes back from that sized deficit in the final quarter or something.

[ThaddeusB]

I would like opinions on if the following results indicate any kind of edge going forward or if this is simply not enough information.

W 73
L 55
Avg W 8.438
Avg L -7.290
Total Win 215

Assuming these are 50/50 propositions, the odds of getting at least 73 successes SOLELY by chance in 128 trials is 6.63%. (But keep in mind the probability of no more than 55 successes is also 6.63%).

W 76
L 52
Avg W 10.993
Avg L -9.759
Total Win 328

Again assuming 50/50 propositions, the math says 2.08% (4.16% 2 sided) by chance alone.

If the two examples are based on the same principle, you can probably combine them, and say 149/256 total. The odds of those results being explained by chance alone are a mere .51% (1.02% 2 sided).

The action point bias adds a bit more evidence in favor of luck not explaining the results. Thus, I'd say its safe to say the PAST results are unlikely to be due to chance alone (which is not to say, they couldn't have been exaggerated by chance). And, of course, the past edge isn't guaranteed to exist into the future, but yah, I think there is definitely a real edge here.

[pats3peat]

that shiznit is crazy