How many bets?

[Drew]

How many win bets does it take to get a good idea of your average win price and win percentage?

50?
100?
1000?

I'm curious what is a good sample size.

Thanks,

Drew

[sjk]

Formula_2002 provides a significance testing formula that addresses this issue. Go to his web page to find.

[Tom]

I use 20 race cycles. enough to mean someting, and not enough to bury current performance in older figures. I hope that I keep improving as I go along, so each 20 race cycle is one point in time. String them together and see haow you do over time. If you wait until you geta massive sample size, it will be meaningless and you will be an old man. You are betting NOW, what matters is NOW.

[oddswizard]

You need a run of 80 races for each racing condition. Any run over 80 will produce a difference of less than 1%.

[BlueShoe]

A run of 80 sounds very short to be meaningful.As a rough analogy,consider the game of craps.80 dice decisions would take about an 1 1/2 hours at a typical table.If pass or dont pass came up more often than the other side,the results would be worthless as far as being statisically valid.Would think that a much larger sample size is required for a true test of what can be expected long term.

[Tom]

You need a run of 80 races for each racing condition. Any run over 80 will produce a difference of less than 1%.

What year do you intend to start betting?;)
80 in each condition?
You mean 80 MSw, 80 MdCl, 80 NW1.....?

What do you do unitl you got that data base - not bet?

What happens if you change your metod, or add a new rating or tool - start from scratch? Or you get 80 in one class done and find out you are doing it wrong?

I figure it out as I go - real time data, make adjustments as I go.

[ryesteve]

What year do you intend to start betting?
The original poster didn't seem to be asking when he could start betting. It sounds more like he's going to feed his win% and average payoff into a bankroll management system, and wants to know how reliable his estimates will be after a given number of races.

[shanta]

I use 20 race cycles. enough to mean someting, and not enough to bury current performance in older figures. I hope that I keep improving as I go along, so each 20 race cycle is one point in time. String them together and see haow you do over time. If you wait until you geta massive sample size, it will be meaningless and you will be an old man. You are betting NOW, what matters is NOW.

I agree with this all the way. Take $200.00. Make $ 2.00 win bets. You will know soon enough if you are on the right track man. The races will give you the needed feedback.

Plus if u are tracking a sample thing based on off odds and think it will hold up under " live" conditions potential disaster awaits. last minute fluctuations in odds drops ( and raises) due to simul money and antiquated tote systems make this extremely dangerous.

my opinion anyway

good hunting and all the best :)
Rich

[Tom]

Richie,
The thing missing in compouter simulatins is - do you make the bet?
If you aren't betting REAL money in real time, the data doens't reflect reality.

[traynor]

oddswizard wrote:<You need a run of 80 races for each racing condition. Any run over 80 will produce a difference of less than 1%.>

In static situations with known, predictable return, that may be true. In racing, one $30 mutuel in that sample can thoroughly corrupt the whole sample, and create differences substantially greater than 1% when the sample is extended. In an 80 race sample, the only way to make it predictive is to correct for outliers. If not, the 80 race sample is a sequence of anomalies, not a model.
Good Luck!