Number of Races to Test an Angle

[NormanTD]

I know this has been brought up before, but supposed you're working an angle/spot play or some other methodology (and you're not a database sort of guy).

How may races would you test before you actually put some money into one of them?

Seems like I remember someone saying 100 races where you could have bet as opposed to just any 100 race sample. 100 betable races could mean several hundred (or thousand) actual races, right?

As always, thanks.

Steve AKA Norman TD

[ceejay]

I'd say it depends on the hit rate of the angle. A 10% hit rate will take more samples than a 25% one.

[ranchwest]

I think you either have to hit hot patterns, which don't last long, or you have to be very, very patient with your pattern research. After about two years you should know whether your pattern will work at a particular track at a particular time of year.

[Shacopate]

I think that small samples are okay as long as you use multiple tracks. Require no less than a 100 per each though. I'd say five tracks minimum.

[ranchwest]

Originally posted by Shacopate
I think that small samples are okay as long as you use multiple tracks. Require no less than a 100 per each though. I'd say five tracks minimum.

My opinion is the opposite. Each track is likely to be different. There are a few factors that would be valid across all tracks, but not many.

[Tom]

When something happens twice, I am on it the third time.
I am in this game to play races, not study data (I do that at work).
When something unusual happens in a trainer pattern, I start paying attention after the first time I see it.

[Kentucky Bred]

Also, if you are charting horses paying high mutuals you need many more races to really know.

I have one angle that was unbelievable this winter and has crashed since it started warming up. Can't figure out one reason except to note that it requires an 8-1 mutual.

I suspect it will all of a sudden start up again for apparently no reason. That means that the run outs of long price horses are so huge that it will require a lot more time to really know if something works.

Kentucky Bred

[MarylandPaul@HSH]

Not to hijack the thread, but hopefully to add to it....

Would you prefer:

a. A small very recent sample? (ie. 50 races at same track/distance)

or

b. A small targeted sample? (ie. last 50 races at the same track/distance/class/age/sex/etc...)

MP

[keilan]

Intelligent question asked IMO

I vote for (b)

[Tom]

Originally posted by Kentucky Bred
Also, if you are charting horses paying high mutuals you need many more races to really know.

I have one angle that was unbelievable this winter and has crashed since it started warming up.........
Kentucky Bred

Maybe the angle is best suited to winter racing. That would be a case where you don't want a large sample, if the data you are gathering now is irrelevant. If it starts hitting again this winter, forget the summer data and bet that bad boy.

[ranchwest]

MP,

This is going to sound like a copout, but it depends on the angle.

If the angle would be likely to have something to do with something hot, the more recent the better. Sometimes days or even races is too late.

If the angle is not likely to be closely related to something hot, the more specific look is better.

[Shacopate]

Whatever you do, don't start changing the original rules of your angle to allow some "close calls" to be counted as wins.

Then you're only predicting the past and cheating yourself.

[pmd62ndst]

Whether you sample big or small, I always like to split my races into two. Use one to test and then tune your theory and then use it on the other to see if it works.