Need Help\Advice Managing Database

[Jingle]

I would like some advice on managing a database to construct "spot plays".

Suppose I have a yrs. data on a current meet. Let's say I take the last 30 days to construct a "spot Play". Not using the last 30 days (same races) to validate the spot play, how far back should I go or how many races should I use to validate the "spot play" ie. 30 days, 60 days, 90 days, etc? I don't want the data to be too old but I don't want to use too few days either. I'm not talking about a speed-pace model necessarily.

I guess I'm wondering what is reasonable to construct the play and then validate it.

Your advice/help is appreciated. Thanks.

[Dave Schwartz]

Bill,

Truth be told, you will need years of data to find anything meaningful. Less than a couple of thousand occurrences would qualify as little more than coincidence.


Regards,
Dave Schwartz

[Jingle]

Thanks Dave.

For your info I used a yrs worth of data as an example--I have thousands of races but it just doesn't seem reasonable to go back to Belmont in the fall of 2008 to find out what was winning. As you know I've been wrong once before though and you're the data guy. I think you'll agree track profiles do change from time to time.