Horse Racing Forum - PaceAdvantage.Com - Horse Racing Message Board

Go Back   Horse Racing Forum - PaceAdvantage.Com - Horse Racing Message Board > Thoroughbred Horse Racing Discussion > General Handicapping Discussion


Reply
 
Thread Tools Rate Thread
Old 04-21-2019, 02:22 AM   #1
Raider
Registered User
 
Raider's Avatar
 
Join Date: May 2015
Posts: 141
Sample size to validate new angle?

I have an idea for a new angle and was wondering how many races I would need to validate it?
Raider is offline   Reply With Quote Reply
Old 04-21-2019, 10:14 AM   #2
Gakiss2
Data Warbucks
 
Gakiss2's Avatar
 
Join Date: Dec 2018
Location: Lexington, KY
Posts: 261
Race Lens

Quote:
Originally Posted by Raider View Post
I have an idea for a new angle and was wondering how many races I would need to validate it?
Not really an answer to your question but maybe a different way to approach it. If you subscribe to Race Lens they have a feature where you can evaluate the ROI of a custom angle. I don't know how many races they use but I think its relatively safe to assume it is a lot. I know they also provide a 'strength' factor which tells you how statistically sound the result is. I know a little about statistics, not a lot, but I'm pretty sure the results are more statistically sound when you have more data.

The downside is that your angle has to be expressed in terms of the characteristics they provide and you have to rely on their data. On both fronts though they don't skimp. You would have to take a look yourself for detail but I recall a LOT of options to choose from such as layoff time, class moves, trainer changes, horse coming off a win, and lots more. I haven't heard of any reason not to trust their data. Their site has a lot of cool features and its worth the price of admission to at least look around at what they have put together.

And tell them Gakiss2 sent ya! Just kidding, they wouldn't know who I am.
__________________
A profit at the race track isn't a profit until you spend it somewhere else.
Gakiss2 is offline   Reply With Quote Reply
Old 04-21-2019, 01:31 PM   #3
Raider
Registered User
 
Raider's Avatar
 
Join Date: May 2015
Posts: 141
THANKS Gives me another site to look at, can't have too much information. ROI would be a good way to look at the method, (combo of angles)

Last edited by Raider; 04-21-2019 at 01:37 PM.
Raider is offline   Reply With Quote Reply
Old 04-21-2019, 07:53 PM   #4
JerryBoyle
Veteran
 
Join Date: Feb 2018
Posts: 845
Quote:
Originally Posted by Raider View Post
I have an idea for a new angle and was wondering how many races I would need to validate it?
Unfortunately, I don't think you'll ever get a "definitive" number of races to determine if the angle has a true edge or if the results are just due to luck alone. At best, you can run tests that will give you a sense of how likely it is that your angle was due to luck, and then it's up to you to determine if you're willing to accept that chance couldn't have caused the results.

Most tests will depend on the number of events (races) played, as well as the average odds you receive on your bets. For longer priced horses, you will need a much larger sample size to feel confident that your results weren't due to luck.

Take a look at this post by Joseph Buchdahl on evaluating betting strategies: https://www.pinnacle.com/en/betting-...L2722PUAZM4Q8Q. In general, his articles are pretty great, though they mostly cover fixed odds betting on shorter priced events.

He also has another article on using the difference between the price available at the time you placed your bet vs the closing price to more quickly determine if you have a strategy with true edge. Though it was written in reference to fixed odds wagering, I suspect it could still be useful for parimutuel wagering. Article here: https://www.pinnacle.com/en/betting-...6JWJM5YKEJUWKQ
JerryBoyle is offline   Reply With Quote Reply
Old 04-22-2019, 01:50 AM   #5
Raider
Registered User
 
Raider's Avatar
 
Join Date: May 2015
Posts: 141
Quote:
Originally Posted by JerryBoyle View Post
Unfortunately, I don't think you'll ever get a "definitive" number of races to determine if the angle has a true edge or if the results are just due to luck alone. At best, you can run tests that will give you a sense of how likely it is that your angle was due to luck, and then it's up to you to determine if you're willing to accept that chance couldn't have caused the results.

Most tests will depend on the number of events (races) played, as well as the average odds you receive on your bets. For longer priced horses, you will need a much larger sample size to feel confident that your results weren't due to luck.

Take a look at this post by Joseph Buchdahl on evaluating betting strategies: https://www.pinnacle.com/en/betting-...L2722PUAZM4Q8Q. In general, his articles are pretty great, though they mostly cover fixed odds betting on shorter priced events.

He also has another article on using the difference between the price available at the time you placed your bet vs the closing price to more quickly determine if you have a strategy with true edge. Though it was written in reference to fixed odds wagering, I suspect it could still be useful for parimutuel wagering. Article here: https://www.pinnacle.com/en/betting-...6JWJM5YKEJUWKQ

Thanks Jerry, that is exactly what needed. I am working on a horse selection angle, so

total number of events
total number of horses in events
total number of horse that fit angle conditions
total number of horse that achieve the desired results

If the % of horses that achieve angle results exceeds the % of horses that fit the angle conditions, dependent on sample size, the angle is validated. Statistical confidence is dependent on number of events. Just did a quick check, % of horse that achieved desired results is 30% of horses of horses that fit the angle. Need to refine it, but it is encouraging. I will need to do some more work on it, narrow the focus, but I think I am on to something. Need to expand the sample size

Funny thing is, it could be valid and I could still bet it wrong, LOL
Raider is offline   Reply With Quote Reply
Old 04-22-2019, 02:46 AM   #6
PaceAdvantage
PA Steward
 
PaceAdvantage's Avatar
 
Join Date: Mar 2001
Location: Del Boca Vista
Posts: 88,163
It's never enough. Just when you think you've tested enough, you take on another sample and BOOM...down the toilet it goes...
__________________
@paceadvantage | Support the site and become a today!
PaceAdvantage is offline   Reply With Quote Reply
Old 04-22-2019, 06:51 AM   #7
Raider
Registered User
 
Raider's Avatar
 
Join Date: May 2015
Posts: 141
I hear you, anything will work some of the time, nothing works all the time, h3ll if it did, no one would be on the site, be laying on the beach.
Raider is offline   Reply With Quote Reply
Old 04-23-2019, 10:04 PM   #8
mike_123_ca
Registered User
 
Join Date: May 2009
Location: Toronto
Posts: 66
I use Stats Race Lens as well and generally my sample size for validating angles is minimum 100 races over the past 12 months, with positive ROI and win percentage of +30%.
mike_123_ca is offline   Reply With Quote Reply
Old 04-24-2019, 09:26 AM   #9
Tom
The Voice of Reason!
 
Tom's Avatar
 
Join Date: Mar 2001
Location: Canandaigua, New york
Posts: 112,446
Quote:
Originally Posted by PaceAdvantage View Post
It's never enough. Just when you think you've tested enough, you take on another sample and BOOM...down the toilet it goes...
So true.
I had a spot play that showed a flat bet profit for three months. Over 250 plays. Profit to win and place, .90 roi for show.

I started a thread to post the play here and immediately went 0-39.

Don't test, play it now before it catches on to you.
__________________
Who does the Racing Form Detective like in this one?
Tom is offline   Reply With Quote Reply
Old 04-24-2019, 07:48 PM   #10
Raider
Registered User
 
Raider's Avatar
 
Join Date: May 2015
Posts: 141
Quote:
Originally Posted by Tom View Post
So true.
I had a spot play that showed a flat bet profit for three months. Over 250 plays. Profit to win and place, .90 roi for show.

I started a thread to post the play here and immediately went 0-39.

Don't test, play it now before it catches on to you.
I am going to have to go with a very small sample, to small to prove it works, but hey what the heck. Its for the TC races only. I think that the points system has changed the shape of the Derby. Using only the Points races, and doing a look back, it seems to prove that the race has changed. Need to see how it works in the preakness and belmont, but two that have met the conditions have done well in both,

I will be using it this year, it can't be worse than anything I have tried before
Raider is offline   Reply With Quote Reply
Old 04-25-2019, 11:30 AM   #11
formula_2002
what an easy game.
 
formula_2002's Avatar
 
Join Date: Dec 2002
Posts: 43,096
Quote:
Originally Posted by Raider View Post
I have an idea for a new angle and was wondering how many races I would need to validate it?
In his book "Horse Sense" Dr. Burton P. Fabricand, who was a Ph.D. from Columbia who published widely in the fields of atomic and nuclear physics, oceanography and finance formulated a theory of "Maximum Confusion" to determine when and when not to bet the favorite.
Briefly, he included 6088 races, 1327 were playable and profitable.
His incremental odds analysis is a method to comprehend and worth the effort.

I looked on Amazon and found the book at a price tag of $199.
Of course the first copy of the book I bought( I now have three copies) , back in the early 1960's cost me $4.95.
__________________
Peace on earth, good will to all
GOD BLESS AMERICA

" I pass with relief from the tossing sea of cause and theory to the firm ground of result and fact"
Winston Churchill
formula_2002 is offline   Reply With Quote Reply
Old 04-25-2019, 11:51 AM   #12
castaway01
Registered User
 
Join Date: Jul 2009
Location: NJ
Posts: 3,816
Quote:
Originally Posted by Raider View Post
I am going to have to go with a very small sample, to small to prove it works, but hey what the heck. Its for the TC races only. I think that the points system has changed the shape of the Derby. Using only the Points races, and doing a look back, it seems to prove that the race has changed. Need to see how it works in the preakness and belmont, but two that have met the conditions have done well in both,

I will be using it this year, it can't be worse than anything I have tried before
Well, the races changed when they implemented the current system because it took a lot of the 2-year-old speed horses out of the race and changed how the races are run (leading to more favorites). But that's not what you're asking.

If it makes you feel any better/worse, if it's just for the Triple Crown races none of us will live long enough to get a big enough sample size to statistically prove anything. If you have something that looks like it has worked in recent years, give it a shot.
castaway01 is offline   Reply With Quote Reply
Old 04-25-2019, 04:57 PM   #13
Raider
Registered User
 
Raider's Avatar
 
Join Date: May 2015
Posts: 141
Thats my thinking, all it can do is lose, and even if it does, there is always next year. Lady luck is telling me to go big, but we all know how shes is, the flirt.
Raider is offline   Reply With Quote Reply
Old 04-26-2019, 08:58 PM   #14
flatstats
Registered User
 
flatstats's Avatar
 
Join Date: Oct 2002
Posts: 149
Hmmm.

100 horses that went off at 2/7
100 horses that went off at 100/1

Which set of results from them would be the most reliable?

Obviously the first one, which means that if you want to "count" the number of runners to check if a sample size is valid you have to take into consideration the price.

I would recommend reading Steve Tilley's notes on sample size and stats 'usefuleness'.

For sample size you want to choose a consistent figure such as 10.0 and for usefulness/validity you want an Archie score of say 5+

Then there is sandboxing. You want to test data with past results, and then wait until a method has had sufficient live runs before continuing.

Sample Size
Sum up all the odds of the runners so for a 2/1 shot, a 3/1 shot and a 5/1 shot you would have:

1/(2+1) + 1/(3+1) + 1/(5+1) = 0.75

That figure (called the expected number of wins or Exp) is way too low. You want 10 be safe. Now you can see that if your method has mostly 100/1 shots then this is going to take a long time. But if your method contains short favourites you do not need "100's" of runners.

Archie Score
This is definitely something you want to calculate. It tells you how likley the results you have observed are due to luck or chance.

It is based on chi-squared and the lower the figure the more likely the results are BS, but the higher the figure the more likely you the figures are trustable.

Historic / Live
This is the hardest bit because anyone who has run a method and the initial figures are good, and the sample size (Exp) is great, and the Archie score is valid will totally convince themselves they are a 49er who started digging in California in 1847.

Unfortunately there are negative forces such as back-fitting and of course there is iron pyrite.

To protect against that you have to a) not back-fit at all (the hardest to avoid for the masses), b) sandbox your method.

To sandbox you have to research your idea and come up with a valid Exp and Archie figures. Then you sit on the results and observe what happens in the future. If your Live (future) Exp and Archie figures are valid then you should have a winning method and you should now start backing the method.
flatstats is offline   Reply With Quote Reply
Old 04-27-2019, 08:46 AM   #15
Raider
Registered User
 
Raider's Avatar
 
Join Date: May 2015
Posts: 141
I knew this would take some work, to prove or disprove, right now, my thinking is try it, dig deeper, get more samples.

Last edited by Raider; 04-27-2019 at 08:47 AM.
Raider is offline   Reply With Quote Reply
Reply




Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump

» Advertisement
Powered by vBadvanced CMPS v3.2.3

All times are GMT -4. The time now is 08:02 AM.


Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2024, vBulletin Solutions, Inc.
Copyright 1999 - 2023 -- PaceAdvantage.Com -- All Rights Reserved
We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program
designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites.