streaks in testing and playing

[Capper Al]

Do streaks win or lose in test or play mean anything if they are less than 30? (I just picked a number. You can pick 5, 10, 20 or whatever number you like. I picked 30 because it was once said that all that was need for sampling was 30.) Share some of your streaks while you'll talk. Mine are as follows:

• 13 winning days in a row using Giles' pace methods. Now I can't get two in a row.
  20 something race losing streak last year.
  Numerous successful test that flopped with real money.

Currently I'm running a test with 30 some races succesful. What to do?

[DeltaLover]

Originally Posted by Capper Al

Currently I'm running a test with 30 some races succesful. What to do?

Start betting real $$ and ramp up as you on, until your reach betting sizes that are affecting the payout...

(Paper results do not mean much though)

[Capper Al]

Originally Posted by DeltaLover

Start betting real $$ and ramp up as you on, until your reach betting sizes that are affecting the payout...

(Paper results do not mean much though)

Oh, I agree about paper results not meaning much. I am surprised on how positive test just vanishes when real money gets involved. But up to what size would you say a streak is meaningless? Everyone at one time has flipped a coin and gotten maybe 6 or 7 heads or tails in a row. And, of course, we know that the chances for either heads or tails over the long run is 50/50. Yet, our observations seem to suggest to us differently.

[Elliott Sidewater]

Al, there isn't quite enough info here for me to formulate the answer I'd like to give you, but I hope you are recording the odds on all plays during the test. A long streak of losses when all the plays are long odds (random field odds or higher) doesn't mean much, but when (for example) 10 or 15 plays in a row lose at 3-1 or less that means a lot. Just for illustration (not an argument from the math guys) let's say a 2-1 shot should win 1 out of 3 times. In one race count that as AN EXPECTATION of one third of a win. A 19-1 shot counts a 1/20 of an expected win and so on. In a long series of plays, add up these fractions to find the total number of expected wins - now compare this to the number of actual wins, the differences will clarify the picture for you, I'm sure. If the running total of expected wins is always or usually higher than the running total of actual wins, the system or method is unlikely to hold up in actual play.

Hope this helps,
Elliott

[Capper Al]

Originally Posted by Elliott Sidewater

Al, there isn't quite enough info here for me to formulate the answer I'd like to give you, but I hope you are recording the odds on all plays during the test. A long streak of losses when all the plays are long odds (random field odds or higher) doesn't mean much, but when (for example) 10 or 15 plays in a row lose at 3-1 or less that means a lot. Just for illustration (not an argument from the math guys) let's say a 2-1 shot should win 1 out of 3 times. In one race count that as AN EXPECTATION of one third of a win. A 19-1 shot counts a 1/20 of an expected win and so on. In a long series of plays, add up these fractions to find the total number of expected wins - now compare this to the number of actual wins, the differences will clarify the picture for you, I'm sure. If the running total of expected wins is always or usually higher than the running total of actual wins, the system or method is unlikely to hold up in actual play.

Hope this helps,
Elliott

I'm a medium to long shot player. I can go 20 races without a win. But by playing multiple horses (a.k.a. wheeling) I used to be able to hit a couple on a card.

The bigger point here is that whatever each of us do, we see streaks. How we categorize these is what is important. If a simple coin toss with only two outcomes can go 6 or 7 times in a row either heads or tails then how do we handle a find in our own research? When do we say that we found something, especially for the hobbyist without a 20,000 race plus database? Are we overemphasizing the patterns before our eyes ignoring the more complex scenarios in racing? What is the meaning of a streak or pattern? Is there an overarching point of view, like understanding the racing game, that supersedes all these streaks right before our eyes?

[thaskalos]

Originally Posted by Capper Al

Do streaks win or lose in test or play mean anything if they are less than 30? (I just picked a number. You can pick 5, 10, 20 or whatever number you like. I picked 30 because it was once said that all that was need for sampling was 30.) Share some of your streaks while you'll talk. Mine are as follows:

• 13 winning days in a row using Giles' pace methods. Now I can't get two in a row.
  20 something race losing streak last year.
  Numerous successful test that flopped with real money.

Currently I'm running a test with 30 some races succesful. What to do?

The selection rules of this test...did you have them in place before the testing started...or did you create the rules after reviewing the results of the races?

[Capper Al]

Originally Posted by thaskalos

The selection rules of this test...did you have them in place before the testing started...or did you create the rules after reviewing the results of the races?

These rules are a result of observation. I'll test them this weekend.

[JJMartin]

Originally Posted by Capper Al

Oh, I agree about paper results not meaning much. I am surprised on how positive test just vanishes when real money gets involved.

I fail to see the logic in such an assumption. If for example your test records indicate a 15% (flat win bet) roi over a sample of past races, you mean to tell me that if you would have have bet say $2 to $10 per race you would have had a negative roi? What about all the other people that had bets in the same pool, they all lost too? To elaborate: you test 1000 past races for a +roi, then you test "forward" on another 1000, still +roi but only if you don't actually bet money?? That doesn't make any sense.

[Capper Al]

Originally Posted by JJMartin

I fail to see the logic in such an assumption. If for example your test records indicate a 15% (flat win bet) roi over a sample of past races, you mean to tell me that if you would have have bet say $2 to $10 per race you would have had a negative roi? What about all the other people that had bets in the same pool, they all lost too? To elaborate: you test 1000 past races for a +roi, then you test "forward" on another 1000, still +roi but only if you don't actually bet money?? That doesn't make any sense.

That was the point that it doesn't make any sense. You have had to experience it yourself to understand it I guess. It happens.

" There are more things in heaven and earth, Horatio,
Than are dreamt of in your philosophy. "

I think Shakespeare was talking about confounded variables in horse racing.

[DeltaLover]

Originally Posted by JJMartin

I fail to see the logic in such an assumption. If for example your test records indicate a 15% (flat win bet) roi over a sample of past races, you mean to tell me that if you would have have bet say $2 to $10 per race you would have had a negative roi? What about all the other people that had bets in the same pool, they all lost too? To elaborate: you test 1000 past races for a +roi, then you test "forward" on another 1000, still +roi but only if you don't actually bet money?? That doesn't make any sense.

Paper results are very unlike to reproduce their performance in future races. The most probable reason for the success of a strategy, while paper betting, is pure luck and nothing else..

[JJMartin]

Originally Posted by Capper Al

That was the point that it doesn't make any sense. You have had to experience it yourself to understand it I guess. It happens.

" There are more things in heaven and earth, Horatio,
Than are dreamt of in your philosophy. "

I think Shakespeare was talking about confounded variables in horse racing.

I think what happens is you become too conscious of the losses when they occur especially with streaks and it causes doubt and a lowering of confidence. In my research I have noticed that certain months have better results than others when looking at consecutive years (as tested with my method). And even though the overall result was consistent, there is no way to know when the streaks will occur and be arranged. As you stated, even with a 50/50 expected ratio, you can have more than 7 or 8 one-sided results. I recommend flat betting (small wagers) over a set period of days or set number of races and wait until the end before making an assessment.

[JJMartin]

Originally Posted by DeltaLover

Paper results are very unlike to reproduce their performance in future races. The most probable reason for the success of a strategy, while paper betting, is pure luck and nothing else..

That may be true in your experience but not mine. To be sure we are on the same page, what is your definition of paper betting?

[DeltaLover]

Originally Posted by JJMartin

That may be true in your experience but not mine. To be sure we are on the same page, what is your definition of paper betting?

Applying a betting strategy on historical data and reporting its PNL and ROI is a commonly accepted definition or paper betting...

[JJMartin]

Originally Posted by DeltaLover

Applying a betting strategy on historical data and reporting its PNL and ROI is a commonly accepted definition or paper betting...

Right. The only thing that I can think of that would create an inconsistency of results is betting according to post time odds live which you cannot accurately duplicate with testing past data. And I suspect most people use PTO to make their betting decisions, I do not.

[DeltaLover]

Originally Posted by JJMartin

Right. The only thing that I can think of that would create an inconsistency of results is betting according to post time odds live which you cannot accurately duplicate with testing past data. And I suspect most people use PTO to make their betting decisions, I do not.

The problem is not about the accuracy of the results of the paper betting but the fact that positive return will not be duplicated in the future... Almost all the betting strategies, will eventually converge to a specific negative ROI (usually between 0.89 - 0.95)