Hey guys, What do you think about this sample size moving forward?

**sneak-a-peak** · 12-18-10, 07:49 PM

Its currently 104-70-5 (59.77%) this season.

**sneak-a-peak** · 12-18-10, 07:54 PM

It does include some 1q and 1h spreads.

**sneak-a-peak** · 12-18-10, 08:53 PM

Oh ya and one other important thing because I know someone will eventually ask.....

But without going into detail before the start of the 2008 season is non applicable.

**jamesbettor** · 12-18-10, 09:52 PM

Before betting using the plays generated by a model I would make sure that I came up with the model using one set of data and then proceeded to test the model with out of sample data. If the model proved to be profitable when forward testing on out of sample data then I would say it is ready to use for betting.

I like to use at least 5 seasons of data to create the model and then 5 seasons of out of sample data to forward test it. Some people may prefer more or less. In NBA, NHL and MLB you can probably get away with less due to the large volume of games that get played each season.

With your 3 seasons of past results, were they obtained using out of sample data? Or did you test your model using the same data that you used to create the model?

**sneak-a-peak** · 12-18-10, 10:25 PM

Originally posted by jamesbettor

With your 3 seasons of past results, were they obtained using out of sample data? Or did you test your model using the same data that you used to create the model?

I had a sample of the 2 previous seasons heading into this season.

So the test run is this season with the 104-70-5 record.

It started off at one point 80-46-4 so since its gone 24-24-1.

**roasthawg** · 12-18-10, 10:26 PM

Tough to say... I can come up with hundreds of models with similar results that will get crushed going forward. The results look promising but it's all about the criteria used to make the picks.

**sneak-a-peak** · 12-18-10, 10:31 PM

Originally posted by roasthawg

Tough to say... I can come up with hundreds of models with similar results that will get crushed going forward. The results look promising but it's all about the criteria used to make the picks.

Ya- thats usually the case and why I have not played it yet but this season has really opened up my eyes and got me thinkin.

The criteria is kind of unique- probably not what you would normally expect.

**classhandicapper** · 12-18-10, 10:57 PM

How do you guys create databases of several years of basketball data to test?