SOFTWARE TESTING

[HIGH ROLLER]

i am trying to test various spot plays, my question is what sample size do you need to test a systems worth?

let's say after 100 races a spot play shows a profit

is that enough? or what should it be?

[ranchwest]

Originally posted by HIGH ROLLER
i am trying to test various spot plays, my question is what sample size do you need to test a systems worth?

let's say after 100 races a spot play shows a profit

is that enough? or what should it be?

IMHO, the important thing is to split your samples into approximately equal sets and then test the consistency of the spot play over the sets.

100 races is not a very large sample at all. Of course, the rules of some spot plays are such that you can't anticipate many plays.

I've got about 400,000 entries in my database, so it is rare that I put together a spot play with only 100 qualifiers.

[formula_2002]

high, check out the file "significance testing" and the references, one of which is B. Fabricand.

http://globalwinningpicks.homestead.com/GLOBALWINNINGPICKSX.html