[JTrain]

What is your long-term winning percentage betting games based on SDQL trends?

What do you look for in query results that determine a good system?

[evo34]

Start broad (huge sample). Narrow down only accoriding to logical filters. Pay attention to edge decay. If things don't look good over the past 12 mos., you need to consider the fact that the angle may be dead. Assume that a 10% backtested edge will be less than 2% going forward (and that's an optimistic scenario, assuming you have not over-fitted anything).

[gojetsgomoxies]

i'm thinking you are looking for 54% or so win rate in reality - consistent with prior poster. maybe a little higher on less volume, but then less volume usually equals overfitting.

the one i am always grappling with is volaitlity of win rate.... like some of the bowl systems you did were excellent but then one year they go 2-9........ so how do you use that system in reality?

[evo34]

When I say "huge sample," I certainly don't mean 100 games. I.e., bowl systems better be 80% or better to even merit a sniff.