When do I lay it in?

[misscashalot]

I am into an ongoing actual betting study involving a 100% automatic bet that has no judgment. All bets for the day are made about ½ hr prior to first post (after scratches). At this point I do not consider I have enough of a sampling to consider this a true and fully tested study. Study started May 15th. All data is from actual bets made by me prior to the day. The numbers, at this time are: Win bets 108, Wins 40 (.37) Total return at track odds $267 ($2 win bets) for a $51 profit (roi +.24) Average win $6.70. Most consecutive races with no wins:11. My question is at what point or at what sampling number would you consider this a workable betting tool in order to lay it in if the numbers remain at this level?

[RichieP]

My feedback would be to start with a $200 real bank and begin making $2.00 win bets going forward today.

When you get to $300 tap up the bet a buck to $3.00 etc etc

This keeps you at 1% of bank betting level

Best of luck and nice work so far! :)

[jfdinneen]

misscashalot,

There are two key factors to consider when estimating sample size for testing a betting rule: confidence interval and confidence level. Confidence interval refers to the range within which you expect the correct strike-rate to fall and confidence level refers to how certain you are that this range holds the true value. In the present context, a confidence interval of +/-5% and a confidence level of 99% would be appropriate. Using these values, and without going into the exact calculations, gives a sample size of approximately 650 (643/644).
What this means is that you need to test the betting rule on 650 races drawn randomly from 2008 (i.e. different population from that used to discover betting rule). You can be 99% certain that the strike rate your betting rule generates from this new random sample will be within +/- 5% of the true strike rate.

Best wishes,

John

[green80]

Larger samples are always better, but I have found on win bets, after 200 sample races, the results stay fairly close. However the results of some methods or algorithms vary from track to track.

[GameTheory]

You also have to consider you may be working with a moving target, depending on how "fundamental" your system is. Which means, even a statistically verified system can suddenly stop working. It is not unusual (AT ALL) to have awesome results for 100 races (or 2000, or 1000, or 5000) and then to have everything go into the tank for just as long. (You could easily have an awesome YEAR be followed by disaster the next year.) And it is important to remember that any run of bets will have ups & downs. The key is to maximize the profits on the hits and minimize the losses on the misses.

But if it seems to be working good now, I'd start betting, with small amounts anyway. There are always wrinkles added by actual betting vs paper, even if you bet in advance. Might as well get used to that process.

[misscashalot]

Larger samples are always better, but I have found on win bets, after 200 sample races, the results stay fairly close. However the results of some methods or algorithms vary from track to track.

Yes, my results as stated above are exclusively from Bel. I am running the same for Arl and Hol and the results are very different in the negative.

[Dave Schwartz]

JF Dineen's answer is, of course, mathematically sound. (They always are.)

However, his estimate is too low for MY comfort zone. I have seen many systems that were profitable at the 800 wager mark become losers by 1,200 or 1,500.

My suggestion is that would be "certain" at around 1,800 races.

Of course, I would be "certain enough" to play at 650 races as he said.


Regards,
Dave Schwartz

[jfdinneen]

...My suggestion is that would be "certain" at around 1,800 races...

Dave,

Thanks for the kind words.

Assuming the same margin of error (i.e. confidence interval) of +/- 5%, a random sample of 1800 (1778) would lead to a confidence level of approximately 99.999%, which as you suggest is as good as "certain" in sports handicapping!

Best wishes,

John

[misscashalot]

JF Dineen's answer is, of course, mathematically sound. (They always are.)

However, his estimate is too low for MY comfort zone. I have seen many systems that were profitable at the 800 wager mark become losers by 1,200 or 1,500.

My suggestion is that would be "certain" at around 1,800 races.

Of course, I would be "certain enough" to play at 650 races as he said.


Regards,
Dave Schwartz

I like JFD's answer and your addition. My thoughts. I probably will bet about 50 races a month. Your 1800 estimate covers about 3 years which is a luxury at my age. My thoughts prior to posting originally here was 2 complete years at NYRA which covers each venue twice or about 1200 races. 2 Years is psychologically doable for me. I guess I'm saying we're all basically on the same track though I'd still like input from all interested. Thanks a lot and I'll keep you posted monthly. We'll ride this together.

[cmoore]

The key is to maximize the profits on the hits and minimize the losses on the misses.



Here's a method for that.

http://74.6.239.67/search/cache?ei=UTF-8&p=fixed+percentage+minimum+betthing+method&fr=yfp-t-126&u=www.horseracinggold.com/Horse-Racing-Betting-Article-10.pdf&w=fixed+percentage+minimum+betthing+betting+method&d=fnAXgBlMTFr1&icp=1&.intl=us

[misscashalot]

Here's a method for that.

http://74.6.239.67/search/cache?ei=UTF-8&p=fixed+percentage+minimum+betthing+method&fr=yfp-t-126&u=www.horseracinggold.com/Horse-Racing-Betting-Article-10.pdf&w=fixed+percentage+minimum+betthing+betting+method&d=fnAXgBlMTFr1&icp=1&.intl=us

Thanx CM, another interesting addition to this subject.

[ranchwest]

I've never done any real studies, but I have it in the back of my mind that horses may show different tendancies based on season of the year.

So, I'd suggest following this through at least a year before being convinced its a good thing.

[Light]

100 races is far too few in my experience. There are just too many variables that will not surface in a 100 race span. I would note the ROI at 100 race intervals.After 600 races, most systems will have had their ROI's flattened to the point of disinterest. If not,you are ready to test it for real. Be prepared for a little dissapointment to its ROI as these systems never perform as well under the pressure of reality as they do under the comfort of fantasy.

[fmolf]

In this day and age with so much info out there so many different methods and wagering theories i would send it in now with a small bankroll as one member suggested.Everyone knows that this testing a method requires more extensive testing than what you have done to be truly sure of a methods viability.In this age with its almost daily changes in the way info is presented and new info becoming available all the time and all the computer sharpies out there send it home now before someone else catches on to it and depresses your odds.....goodluck miss......6th today at belmont i like the
:1: second off the layoff /blinkers on/second with lasix...ad the :5: and :8: to roundout the exacta....

[Handiman]

This may be far too simple of an observation, but bottom line is it will work till it doesn't. No one is more of a numbers and system guy than I am. You can develop your system and then test the hell out of it. Small sample...large sample...Ginormous sample, but when all said and done, the day you put it into action, things could change and it might not produce hoped for results.

Horseracing is a living, breathing animal much like the horses that make it up. It changes all the time. The best you can do is put it into action and then Keep records. And More records. And then some more records! And hope for the best.

Good Luck,
Handi :)

[formula_2002]

Your average mutual is small enough to make the following test interesting.
From your picks, select an odds range (as small as possible), with the maximum number of plays as possible. Calculate the win%

Then using the same odds range, determine the win percentage for all the horses in your sample.
Then calculate the z score.

Plan A:
If the score is less than 2, it may be worth to stop betting until you get it to 2 or more. If its more than 2, go to plan B.

Plan B
If its more than 2, it may be worth to stop betting and see where the score is after another 100 picks.
In the mean time, take your previous winnings and buy something. :)

[Track Collector]

This may be far too simple of an observation, but bottom line is it will work till it doesn't. No one is more of a numbers and system guy than I am. You can develop your system and then test the hell out of it. Small sample...large sample...Ginormous sample, but when all said and done, the day you put it into action, things could change and it might not produce hoped for results.

Horseracing is a living, breathing animal much like the horses that make it up. It changes all the time. The best you can do is put it into action and then Keep records. And More records. And then some more records! And hope for the best.

Good Luck,
Handi :)

I agree with Handi.

I am also a numbers guy, and things do seem to change constantly. What was very successful 5-6 years ago now may be a 20-25% loss. It has also been my unfortunate experience that promising patterns found in the database almost never turn out as successful in future plays. :mad: With the recent trends (banning of drugs previously allowed and shoe restrictions), it is almost as if none of one's data prior to 2009 is valid. Starting your studies with 2009 significantly reduces you sample sizes, which adds much more risk to any patterns which hint at profitability.