PDA

View Full Version : Bet Sizing


BCOURTNEY
06-01-2008, 04:13 PM
Ok, you've made your selections. Now it's time to click the mouse button or go to the window. Anyone care to share successes they have had in bet sizing and producing a positive bankroll over a significant amount of time? What types of wagers do you make and are you flat betting, kelly betting, fractional kelly, what percentage of your bankroll do you wager? Do you produce a odds line? Do you account for the pool sizes and amounts in each one?

Personally, I beleive that proper bet sizing is about 3 or 4 times more important than selection or prediction methods in order to produce a successful profitable system, if you disagree with this then please share why this is. I have always been curious why the focus of handicapping software is almost exclusively concerned with selections as opposed to the wager and bet size. Share experiences.

Thanks,
B

ezpace
06-01-2008, 06:50 PM
Ok, you've made your selections. Now it's time to click the mouse button or go to the window. Anyone care to share successes they have had in bet sizing and producing a positive bankroll over a significant amount of time? What types of wagers do you make and are you flat betting, kelly betting, fractional kelly, what percentage of your bankroll do you wager? Do you produce a odds line? Do you account for the pool sizes and amounts in each one?

Personally, I beleive that proper bet sizing is about 3 or 4 times more important than selection or prediction methods in order to produce a successful profitable system, if you disagree with this then please share why this is. I have always been curious why the focus of handicapping software is almost exclusively concerned with selections as opposed to the wager and bet size. Share experiences.

Thanks,
B

Your right on the size and CONSISTENCY OF WAGER. THE BEST i have seen is at Daves Horse Street site. ..............bbmL...

Bruddah
06-01-2008, 07:34 PM
Ok, you've made your selections. Now it's time to click the mouse button or go to the window. Anyone care to share successes they have had in bet sizing and producing a positive bankroll over a significant amount of time? What types of wagers do you make and are you flat betting, kelly betting, fractional kelly, what percentage of your bankroll do you wager? Do you produce a odds line? Do you account for the pool sizes and amounts in each one?

Personally, I beleive that proper bet sizing is about 3 or 4 times more important than selection or prediction methods in order to produce a successful profitable system, if you disagree with this then please share why this is. I have always been curious why the focus of handicapping software is almost exclusively concerned with selections as opposed to the wager and bet size. Share experiences.

Thanks,
B

Let's say your system has the proper bet size to produce your desired results, over a period of time. The only thing you lack is proper selection. YOUR GOOSE is COOKED!!

Selection and proper bet sizing are equally important. It's like baking a cake. Too much flour or too much eggs, or not enough of either, will ruin the finished product and it won't be fit to eat. :ThmbUp:

HUSKER55
06-01-2008, 08:00 PM
I think it boils down to how well do you know the system you are using or how much do you trust the system you made for yourself.

For example, my system almost always produces 3 or 4 contenders but 3 or 4 times A WEEK it will only produce 1 horse and if the odds are over 3:1 I bet way more than normal and believe it or not I hit more often than not.

For me, there is the art of handicapping and the art of wagering and I am at a stage right now where I am unable to seperate the two. I pick my horse and then I try to make the most money I can with my pick.

I think that going with your second or third pick is a suckers bet because you are basically betting against yourself.

Maybe I am wrong. What do the rest of you think?

Thanks

husker55

Overlay
06-01-2008, 08:17 PM
I think that going with your second or third pick is a suckers bet because you are basically betting against yourself.

Even if you consider playing only your top-rated horse (which can certainly be a valid approach), you'll do better if you have a means of judging (by either experience, intuition, or quantitative data) whether the horse has a greater chance of winning than its odds indicate. The disparity between the odds and your estimate of the horse's winning probability can then also serve as a basis for wager sizing (as with Kelly or fractional Kelly, for example). And that same logic can be extended to horses below your top pick, to make them possible favorable betting propositions. It's not just about the absolute winning chances of any particular horse, but about those chances in relation to the horse's odds (whether it's the "best" horse in the field or not).

TrifectaMike
06-01-2008, 08:19 PM
I think it boils down to how well do you know the system you are using or how much do you trust the system you made for yourself.

For example, my system almost always produces 3 or 4 contenders but 3 or 4 times A WEEK it will only produce 1 horse and if the odds are over 3:1 I bet way more than normal and believe it or not I hit more often than not.

For me, there is the art of handicapping and the art of wagering and I am at a stage right now where I am unable to seperate the two. I pick my horse and then I try to make the most money I can with my pick.

I think that going with your second or third pick is a suckers bet because you are basically betting against yourself.

Maybe I am wrong. What do the rest of you think?

Thanks

husker55

Betting against yourself is dependent on how your selections are made. If your selections are made by following a probability curve, then you are not betting against yourself. However, if your selection process is based on measurement factors such as pace or speed ratings, then yes you might be betting against yourself.

Mike

BCOURTNEY
06-01-2008, 09:02 PM
Betting against yourself is dependent on how your selections are made. If your selections are made by following a probability curve, then you are not betting against yourself. However, if your selection process is based on measurement factors such as pace or speed ratings, then yes you might be betting against yourself.

Mike

Comment on the probability curve. If you think of insurance payouts upon death as the track take and monies to other players, do you think that insurance companies would do anything but probability estimates and averages when they issue their policies? Funny part is you cant argue with them about when you will die, they already know when that is, and don't worry if your 30-1 they will still accept you at a slightly rate for anticipated losses. Even 100-1's win 1 of 700 races (-66% returns ).

BCOURTNEY
06-01-2008, 09:13 PM
I think it boils down to how well do you know the system you are using or how much do you trust the system you made for yourself.

For example, my system almost always produces 3 or 4 contenders but 3 or 4 times A WEEK it will only produce 1 horse and if the odds are over 3:1 I bet way more than normal and believe it or not I hit more often than not.

For me, there is the art of handicapping and the art of wagering and I am at a stage right now where I am unable to seperate the two. I pick my horse and then I try to make the most money I can with my pick.

I think that going with your second or third pick is a suckers bet because you are basically betting against yourself.

Maybe I am wrong. What do the rest of you think?

Thanks

husker55

Depends on if horse racing is a regression or classification problem .. to you.
If you think it is a regression problem then you should be using a probability curve and potnetially playing multiple or no horses in a race. You should use a computer. If you think it's a classification problem, then you play according to the classes you develop. This is generally done pen and paper, but can be done on a computer also.

Cheers.

Murph
06-01-2008, 11:19 PM
Comment on the probability curve. If you think of insurance payouts upon death as the track take and monies to other players, do you think that insurance companies would do anything but probability estimates and averages when they issue their policies? Funny part is you cant argue with them about when you will die, they already know when that is, and don't worry if your 30-1 they will still accept you at a slightly rate for anticipated losses. Even 100-1's win 1 of 700 races (-66% returns ).How many 100-1 shots go off in a set of 700 races?

I would like to comapre your results to mine. Maybe some of the other d-base guys would like to check that out too.

I'll answer my own question now. In the past week Thorostats has recorded around 1380 races. It is a busy time for racing as many tracks large and small focus their spring racing in conjunction with the KY Derby and Triple Crown races.

In that set of nearly 1400 races, there are zero 100-1 shots winning any of them. For the month of May there are approx 5400 races. Again my data returns zero 100-1 winners. I do not know how many 100-1 shots went off in that group,(a limitation of the data that is being addressed.)

The same data set of all races for May returns 18 winners of 50-1 or greater with the largest payout at 163.20 Thorostats top selection had zero winners chosen from that group (if you were wondering about that.)

Hopefully, won the 2nd race at Penn National May 29th (http://www.drf.com/results/29/rPEN29.html?rn=953282)by 3 lengths over a soft turf course at 1 1/16M, carded for 3YO MdnClm Fillies for 25K. Trained by James Farley and ridden by J A Farley (maybe a 5# bug boy relative?), her tote board odds were 80-1.

Now, Thorostats only covers 91 North American thoroughbred tracks so I may not have the same data as you used to make your statement on 100-1 shots winning once in every 700 races. Plenty of posters here have a D-base much more powerful than mine. Maybe they can share their findings. We enjoy this type of thing so much you know. ;)

Murph

BCOURTNEY
06-01-2008, 11:50 PM
How many 100-1 shots go off in a set of 700 races?

I would like to comapre your results to mine. Maybe some of the other d-base guys would like to check that out too.

I'll answer my own question now. In the past week Thorostats has recorded around 1380 races. It is a busy time for racing as many tracks large and small focus their spring racing in conjunction with the KY Derby and Triple Crown races.

In that set of nearly 1400 races, there are zero 100-1 shots winning any of them. For the month of May there are approx 5400 races. Again my data returns zero 100-1 winners. I do not know how many 100-1 shots went off in that group,(a limitation of the data that is being addressed.)

The same data set of all races for May returns 18 winners of 50-1 or greater with the largest payout at 163.20 Thorostats top selection had zero winners chosen from that group (if you were wondering about that.)

Hopefully, won the 2nd race at Penn National May 29th (http://www.drf.com/results/29/rPEN29.html?rn=953282)by 3 lengths over a soft turf course at 1 1/16M, carded for 3YO MdnClm Fillies for 25K. Trained by James Farley and ridden by J A Farley (maybe a 5# bug boy relative?), her tote board odds were 80-1.

Now, Thorostats only covers 91 North American thoroughbred tracks so I may not have the same data as you used to make your statement on 100-1 shots winning once in every 700 races. Plenty of posters here have a D-base much more powerful than mine. Maybe they can share their findings. We enjoy this type of thing so much you know. ;)

Murph


You are correct, the return was negative -61% percent not the -66%. I spoke quickly off the top of my head when providing an example of a long odds horse as a joke in relation to the insurance business ... and I think the context I presented it in required less precision than a database report, but since you need to have one ... (because I also enjoy this type of thing)

http://bpp.wharton.upenn.edu/jwolfers/Papers/Favorite_Longshot_Bias.pdf

"Our data include all 6.4 million horse race starts in the United States from 1992 to 2001." - Page 2. The data was provided by Equibase. :cool:

Murph
06-02-2008, 12:18 AM
You are correct, the return was negative -61% percent not the -66%. I spoke quickly off the top of my head when providing an example of a long odds horse as a joke in relation to the insurance business ... and I think the context I presented it in required less precision than a database report, but since you need to have one ... (because I also enjoy this type of thing)

http://bpp.wharton.upenn.edu/jwolfers/Papers/Favorite_Longshot_Bias.pdf

"Our data include all 6.4 million horse race starts in the United States from 1992 to 2001." - Page 2. The data was provided by Equibase. :cool:
AWSOME!! You're working with a few more races than I am!

You are too kind. I need to study a document like this and believe me I am grateful to you for sharing it. I will be lucky to build my database to one million races, I won't live that long. Unbelieveable report and they give you credits for working with info in this manner? That's fantastic, thank you again, BC.

On average, how many runners go off at 100-1 or better in 700 races? It may be a few daysuntil I can find that in the report.

BCOURTNEY
06-02-2008, 12:32 AM
AWSOME!! You're working with a few more races than I am!

You are too kind. I need to study a document like this and believe me I am grateful to you for sharing it. I will be lucky to build my database to one million races, I won't live that long. Unbelieveable report and they give you credits for working with info in this manner? That's fantastic, thank you again, BC.

The 61% stat is in the introduction to the paper section 1.

I really enjoyed the report, it isn't my report however, I found it from a
google search.

On a side note regarding any past horse racing economic models, and perhaps of interest, if you like papers like these, I recommend considering Harville not
Henery in terms of statistical analysis, Harville has more robust and accurate methods.

Assigning Probabilities to the Outcomes of Multi-Entry Competitions
David A. Harville
Journal of the American Statistical Association, Vol. 68, No. 342 (Jun., 1973),
pp. 312-316

098poi
06-02-2008, 12:33 AM
Thanks for the cool link! (Unfortunately my last few brain cells exploded trying to understand it all):eek:

Murph
06-02-2008, 12:37 AM
I really enjoyed the report, it isn't my report however, I found it from a
google search.

On a side note regarding any past horse racing economic models, and perhaps of interest, if you like papers like these, I recommend considering Harville not
Henery in terms of statistical analysis, Harville has more robust and accurate methods.

Assigning Probabilities to the Outcomes of Multi-Entry Competitions
David A. Harville
Journal of the American Statistical Association, Vol. 68, No. 342 (Jun., 1973),
pp. 312-316I am just getting my feet wet on this type of study. I am not at all convinced that advanced statistical modeling and engineering level math and physics can help me pick more winners. I am not the smartest kid on the block either but I know that folks would not waste so much time without some benefit involved. Maybe I need to have a closer look. The report will help me draw some valid conclusions.

Won't it?

jcrabboy
06-02-2008, 01:41 PM
Ok, you've made your selections. Now it's time to click the mouse button or go to the window. Anyone care to share successes they have had in bet sizing and producing a positive bankroll over a significant amount of time? What types of wagers do you make and are you flat betting, kelly betting, fractional kelly, what percentage of your bankroll do you wager? Do you produce a odds line? Do you account for the pool sizes and amounts in each one?

Personally, I beleive that proper bet sizing is about 3 or 4 times more important than selection or prediction methods in order to produce a successful profitable system, if you disagree with this then please share why this is. I have always been curious why the focus of handicapping software is almost exclusively concerned with selections as opposed to the wager and bet size. Share experiences.

Thanks,
B

Hi BCOURTNEY:

Essentially I bet Daily Doubles, pick 3's and pick 4's (I like the single takeout aspect of this wager) along with flat win bets. In my handicapping I focused on learning to pick winners at a decent rate (not so hot at choosing horses to finish in other money slots).

I bet one horse in each leg of DD, P3 Or P4 (a $2 wager). I usually hit at least one or 2 of these per day covering bet cost and then some.

I usually bet earlybird, so I use the Morning Line for my win bet calculations (this can make for pleasant surprises-it can also disappoint, but overall it has worked well for me). Win bets are calculated for the day based on having 200 times the amount of wager in my bankroll. I cap my win wagers at $50.00 (less at smaller tracks) regardless of bankroll size. I will pass a race for win betting if ML too low (I sometimes will make a place bet on these horses, but lately I am not getting much bang for the buck with these wagers, so I probably will be dropping them), but use horse in my serial betting.

I like to keep things simple and manageable.

I have enjoyed the threads you have started lately.

Jimmie

cj's dad
06-02-2008, 02:26 PM
Even 100-1's win 1 of 700 races (-66% returns ).

Isn't it actually -86%

Capper Al
06-02-2008, 02:37 PM
Ok, you've made your selections. Now it's time to click the mouse button or go to the window. Anyone care to share successes they have had in bet sizing and producing a positive bankroll over a significant amount of time? What types of wagers do you make and are you flat betting, kelly betting, fractional kelly, what percentage of your bankroll do you wager? Do you produce a odds line? Do you account for the pool sizes and amounts in each one?

Personally, I beleive that proper bet sizing is about 3 or 4 times more important than selection or prediction methods in order to produce a successful profitable system, if you disagree with this then please share why this is. I have always been curious why the focus of handicapping software is almost exclusively concerned with selections as opposed to the wager and bet size. Share experiences.

Thanks,
B

If I use my head, I use the half kelly.

http://www.suptips.com/system_kelly_criteria.htm

BCOURTNEY
06-04-2008, 10:30 PM
I think it boils down to how well do you know the system you are using or how much do you trust the system you made for yourself.

For example, my system almost always produces 3 or 4 contenders but 3 or 4 times A WEEK it will only produce 1 horse and if the odds are over 3:1 I bet way more than normal and believe it or not I hit more often than not.

For me, there is the art of handicapping and the art of wagering and I am at a stage right now where I am unable to seperate the two. I pick my horse and then I try to make the most money I can with my pick.

I think that going with your second or third pick is a suckers bet because you are basically betting against yourself.

Maybe I am wrong. What do the rest of you think?

Thanks

husker55

I think that handicapping is more of the art and that wagering is more of a science. I also agree with wagering on your postitive expectations regardless of your system used. Diversification or hedging speaks to uncertainty or ignorance and in any event if those strategies are being employed the best option is not to bet in my experience.

BCOURTNEY
06-04-2008, 10:33 PM
If I use my head, I use the half kelly.

http://www.suptips.com/system_kelly_criteria.htm

I use a fractional kelly system. Works very well.

cj
06-04-2008, 10:33 PM
My opinion...it is hard enough to determine if you are betting on genuine overlays or not. Trying to determine "how much" of an overlay is bordering on the impossible. I stick to flat betting when I think I have an edge.

BCOURTNEY
06-04-2008, 10:34 PM
Isn't it actually -86%

Returns versus percent winners. The returns were the percentage provided.

BCOURTNEY
06-04-2008, 10:36 PM
Even if you consider playing only your top-rated horse (which can certainly be a valid approach), you'll do better if you have a means of judging (by either experience, intuition, or quantitative data) whether the horse has a greater chance of winning than its odds indicate. The disparity between the odds and your estimate of the horse's winning probability can then also serve as a basis for wager sizing (as with Kelly or fractional Kelly, for example). And that same logic can be extended to horses below your top pick, to make them possible favorable betting propositions. It's not just about the absolute winning chances of any particular horse, but about those chances in relation to the horse's odds (whether it's the "best" horse in the field or not).

Do you have an example of your fractional kelly method?

BCOURTNEY
06-04-2008, 10:38 PM
Hi BCOURTNEY:

Essentially I bet Daily Doubles, pick 3's and pick 4's (I like the single takeout aspect of this wager) along with flat win bets. In my handicapping I focused on learning to pick winners at a decent rate (not so hot at choosing horses to finish in other money slots).

I bet one horse in each leg of DD, P3 Or P4 (a $2 wager). I usually hit at least one or 2 of these per day covering bet cost and then some.

I usually bet earlybird, so I use the Morning Line for my win bet calculations (this can make for pleasant surprises-it can also disappoint, but overall it has worked well for me). Win bets are calculated for the day based on having 200 times the amount of wager in my bankroll. I cap my win wagers at $50.00 (less at smaller tracks) regardless of bankroll size. I will pass a race for win betting if ML too low (I sometimes will make a place bet on these horses, but lately I am not getting much bang for the buck with these wagers, so I probably will be dropping them), but use horse in my serial betting.

I like to keep things simple and manageable.

I have enjoyed the threads you have started lately.

Jimmie

For serial style betting do or have you considered the place or show pool?

BCOURTNEY
06-04-2008, 10:40 PM
My opinion...it is hard enough to determine if you are betting on genuine overlays or not. Trying to determine "how much" of an overlay is bordering on the impossible. I stick to flat betting when I think I have an edge.

That is fairly conservative, do you assess each race individually yourself then and not use a generalized / computer model?

Dave Schwartz
06-04-2008, 11:54 PM
Permit me to play devil's advocate here but I am playing (and advocating) an anti-kelly approach.

In the last six months at HorseStreet we have made tremdnous strides in the one-horse-per-race paradigm. We've concentrated on using the Monty Hall Problem to concntrate 2/3 of our top3's wins into a single horse and it is working.


Monty Hall Simulator
http://math.ucsd.edu/~crypto/Monty/monty.html

Monty Hall Explanation
http://montyhallproblem.com/


This has resulted in phenomenal hit rates with a play in every race.

(Yes, I know it sounds like I am a huckster. Can't help it 'cause it is true. Yes, we play a lot of chalk, but we hit enough price horses to be profitable.)

In reality, most of us have dialed back the hit rate a little to add value to the equation and have settled in at 27-35% hit rates with solid profitability.


With these hit rates (and, of course, profit) we have gone to a light chase-the-money strategy, which makes one almost unbeatable in session play. Personally, I use Kelly to correctly size my sessions but actually play this "anti-kelly" within the session.

There is no way that I can advocate a straight Kelly approach - fractional Kelly is just Kelly under a different disguise.

Yes, Kelly grows your theoretical bankroll faster than anything else. It just rarely works for mere mortals in the real world.

Kelly reduces your advantage. And I see that is a bad thing. If I can make a little less growth (say 2% less) but wager only 1/2 the money to do it, I consider that a huge improvement.


Regards,
Dave Schwartz

proximity
06-05-2008, 04:55 AM
Do you have an example of your fractional kelly method?

i guess you could say i use a conservative version of 1/4 kelly.

for basic win betting if you have an overlay i recommend betting the FIRST dollar amount listed (for the odds you make the horse) on barry meadow's win bet chart found in his book money secrets at the racetrack..... per $4000 (not $2000) of bankroll. (the only exception being skipping the $5 bet on a 5-1 shot going off @ 7-1).

mr meadow lists his chart as being for $2000, but says it is for races that he rates a "b" on his a-b-c scale. personally i don't grade the races a-b-c, i just assume all the races are c.....which is why i use 1/2 of his 1/2 kelly recommendation.

you may find this to be a bit (or more than a bit) conservative.

i readjust my bet charts at the beginning of each month.

proximity
06-05-2008, 05:08 AM
My opinion...it is hard enough to determine if you are betting on genuine overlays or not. Trying to determine "how much" of an overlay is bordering on the impossible. I stick to flat betting when I think I have an edge.

sometimes what we feel to be a big overlay is actually just a big mistake on our part. and many other times the bet is still an overlay, just not as big of an overlay as our line suggests.

i do know some good players who are (pretty much) flat bettors, but i've always been more comfortable betting more on my line's 6/5 shots than my line's 6/1 shots.

TrifectaMike
06-05-2008, 11:01 AM
Permit me to play devil's advocate here but I am playing (and advocating) an anti-kelly approach.

In the last six months at HorseStreet we have made tremdnous strides in the one-horse-per-race paradigm. We've concentrated on using the Monty Hall Problem to concntrate 2/3 of our top3's wins into a single horse and it is working.


Monty Hall Simulator
http://math.ucsd.edu/~crypto/Monty/monty.html

Monty Hall Explanation
http://montyhallproblem.com/


This has resulted in phenomenal hit rates with a play in every race.

(Yes, I know it sounds like I am a huckster. Can't help it 'cause it is true. Yes, we play a lot of chalk, but we hit enough price horses to be profitable.)

In reality, most of us have dialed back the hit rate a little to add value to the equation and have settled in at 27-35% hit rates with solid profitability.


With these hit rates (and, of course, profit) we have gone to a light chase-the-money strategy, which makes one almost unbeatable in session play. Personally, I use Kelly to correctly size my sessions but actually play this "anti-kelly" within the session.

There is no way that I can advocate a straight Kelly approach - fractional Kelly is just Kelly under a different disguise.

Yes, Kelly grows your theoretical bankroll faster than anything else. It just rarely works for mere mortals in the real world.

Kelly reduces your advantage. And I see that is a bad thing. If I can make a little less growth (say 2% less) but wager only 1/2 the money to do it, I consider that a huge improvement.


Regards,
Dave Schwartz

Dave, if you prefer not to answer I'll understand. The question is at what rate do your top three contenders win and at what average price?

Mike

Dave Schwartz
06-05-2008, 01:05 PM
Top three are just average... around 62%, but "the play" hits between 35% and 42%, with a play in every race.

The key is the decision-making engine called Monty Hall. I have spent the last two years making this adaptable to racing.



Regards,
Dave Schwartz

cj
06-05-2008, 01:27 PM
That is fairly conservative, do you assess each race individually yourself then and not use a generalized / computer model?

I'm using a database to find the plays, and the only thing I require is minimum odds. Some plays may be 3 to 2, some 5 to 1, etc. However, it doesn't matter if the horse selected is one click above the minimum or 50. It works well for me.

LottaKash
06-05-2008, 01:48 PM
IF YOU CAN MAINTAIN ROUGHLY 30% WINNERS @ 2/1,,,,THEN THE METHOD I USE WITH GREAT SUCCESS IS SOMETHING LIKE THIS....

START WITH ANY BANKROLL THAT YOU CAN AFFORD....
START BY BETTING 5% OF YOUR BANKROLL.....
IF A WIN....RAISE 1/2%....
IF A LOSS...LOWER BY BY 1/4%....
WHEN AT 8% STAY THERE AND KEEP BETTING AT THAT % OR TAKE PROFITS AND/OR RESTART AT 5%...
IF ON A MAJOR SCHNEID...DO NOT GO ANY LOWER THAN 2%.......

THIS IS, (IF YOU ARE MAINTAINING YOUR 33% WINS @ 2/1) A GREAT WAY TO TAKE ADVANTAGE OF YOUR WINNING AND LOSING STREAKS.......

IF DONE FAITHFULLY, IT WILL KEEP YOUR BANKROLL IN CHECK AT ALL TIMES, AND YOU WILL ALWAYS KNOW, THE SIZE OF YOU NEXT WAGER....

THIS METHOD OF BETTING HAS KEPT ME IN GOOD SHAPE FOR A GOOD WHILE NOW.......IN FACT, RECENTLY I HAVE DISCOVERED A NEW WAY TO INTERPRET (USING NUMBERS) AND PERCEIVE PACE IN A WHOLE NEW WAY....IT HAS BEEN GIVING ME THE SUCCESS THAT I HAVE ALWAYS DREAMED ABOUT....BUT DURING SOME OF MY TESTING, I WAS USING IT COLD AND BLINDLY WITHOUT ANY ADJUSTMENTS, AND i RAN INTO A 16 RACE LOSING STREAK, AND DESPITE THIS FAILING, MY BANKROLL IS STILLL QUITE HEALTHY AND IT IS BALOONING DAILY USING THIS FAILSAFE METHOD OF WAGERING......

THE NEXT BET SIZE, INCLUDES ALL YOUR BETS, WIN BETTING AS WELL AS EXACTAS AND SUCH GIMMICK BETS, SO YOU HAVE TO DECIDE HOW TO DIVVY UP THE CALLED FOR WAGER.....i SUCK AT EXOTICS, BUT EXCEL AT WIN BETS AS THAT IS MY REAL STRENGTH.....SO OTHER THAN COLD EXACTAS, I AM STRICTLY A WIN BETTER.....(MAYBE SOMEDAY THAT WILL CHANGE)

BUT, YOU COULD SET UP A SEPARATE BANKROLL FOR YOUR GIMMICK BETS, AND BET ACCORDINGLY......BUT THE REAL POWER OF THIS METHOD IS IN THE WIN BET DEPARTMENT......

WELL, THAT IS MY WAY, AND WORKS WONDERS FOR ME, AND TAKES THE GUESSWORK OUT OF EQUATION WHEN IT COMES TO HOW MUCH TO BET NEXT...

ALSO, FOR BETS THAT I HAVE A LOWER EXPECTAION OF WINNING, i WILL BET 70% OF THE CALLED FOR WAGER....THUS IF IF YOU WIN YOU WILL STILL MAKES SOME CASH, BUT IF YOU WERE WRONG IT DOESN'T COST AS MUCH....BUT, YOU STILL MAKE THE ADJUSTMENT IN THE USUAL WAY UP 1/2% OR DOWN 1.4%

I AM A HARNESS BETTOR, AND OTHER THAN MEADOWLANDS AND THE OJC, BETTING INTO THE SMALL POOLS AT THE OTHER TRACKS THAT I PLAY, HURTS THE ODDS, SO I HAVE A CAP ON MY BANKROLL, AND WHEN IT TOPS OUT AT A CERTAIN AMOUNT, I TAKE PROFITS, AND START OUT AGAIN AT 5%.....THIS KEEPS IT FUN AND LOW KEY, AND THE BANKROLL KEEPS GROWING AND THERE IS NEVER ANY PRESSURE TO BET A LOT MORE OR A LOT LESS......AND I CAN CONCENTRATE ON THE NEXT PLAY, KNOWING FULL WELL AS TO WHAT I WILL DO NEXT.......

HUMBLY....:cool:

Dave Schwartz
06-05-2008, 02:33 PM
Lotta,

So, you are saying that you play horses that are basically break-even or a little loser and make money from this?


Regards,
Dave Schwartz

PS: Could you please turn off the caps lock? In BBS language, that is the equivilent to shouting.

InControlX
06-05-2008, 04:50 PM
If you can access a data base to test your plays over a good sample count, keep track of the highest count of consecutive wins (BOOM) and the highest count of consecutive losses (BUST) in the trial. Base you flat wager on about twice the BUST value, or = (Investment Bankroll/BUST x 2). This way you won't get hurt through the anticipated dry spells.

REAL EXAMPLE:

Samples = 633 (race count of play results)
Win % = 31.9 (winning percentage of plays)
BOOM = 6 (longest winning streak)
BUST = 14 (longest losing streak)

$1000 bankroll to risk.
Flat Bet = $1000/(14 x 2) = appr. $35

Of course, it always pays to study the heck of the BUST sequence vs the BOOM sequence for a common influence and tweak your play.

The point is that no matter how good the winning percentage and ROI may be if the BUST number wipes you out you lose.

ICX

jcrabboy
06-05-2008, 06:53 PM
For serial style betting do or have you considered the place or show pool?

Hi BCOURTNEY:

Not really as I am usually wagering earlybird. If I am at the track I will occasionally play a round robin show parlay using horses at longer odds that I believe might have a real shot at getting in the money.

Jimmie

pantoner
06-05-2008, 07:10 PM
Top three are just average... around 62%, but "the play" hits between 35% and 42%, with a play in every race.

The key is the decision-making engine called Monty Hall. I have spent the last two years making this adaptable to racing.



Regards,
Dave Schwartz

62 * 2/3 Is about 43%. Are you saying that you are playing one horse per race to get that number or betting two as one unit?

Dave Schwartz
06-05-2008, 07:12 PM
One horse.

andicap
06-05-2008, 08:20 PM
after studying the links, I feel I have a good grasp on the puzzle. So you are saying that by successfully translating the problem picking one of your top 3 horses you win about 2/3 TIMES your top 3 hit rate (62%) in picking a single horse (give or take for confidence intervals.)? That translates to over 40% winners.
(I won't spoil the puzzle for everyone else)

That is fascinating work especially since there is no opportunity in racing to "switch" horses so I would guess it is dependent on accurately assessing the probabilities of each the three horses or assigning them some value based on how they are being bet.

(Of courser I don't expect you to tell us!)





Permit me to play devil's advocate here but I am playing (and advocating) an anti-kelly approach.

In the last six months at HorseStreet we have made tremdnous strides in the one-horse-per-race paradigm. We've concentrated on using the Monty Hall Problem to concntrate 2/3 of our top3's wins into a single horse and it is working.


Monty Hall Simulator
http://math.ucsd.edu/~crypto/Monty/monty.html

Monty Hall Explanation
http://montyhallproblem.com/


This has resulted in phenomenal hit rates with a play in every race.

(Yes, I know it sounds like I am a huckster. Can't help it 'cause it is true. Yes, we play a lot of chalk, but we hit enough price horses to be profitable.)

In reality, most of us have dialed back the hit rate a little to add value to the equation and have settled in at 27-35% hit rates with solid profitability.


With these hit rates (and, of course, profit) we have gone to a light chase-the-money strategy, which makes one almost unbeatable in session play. Personally, I use Kelly to correctly size my sessions but actually play this "anti-kelly" within the session.

There is no way that I can advocate a straight Kelly approach - fractional Kelly is just Kelly under a different disguise.

Yes, Kelly grows your theoretical bankroll faster than anything else. It just rarely works for mere mortals in the real world.

Kelly reduces your advantage. And I see that is a bad thing. If I can make a little less growth (say 2% less) but wager only 1/2 the money to do it, I consider that a huge improvement.


Regards,
Dave Schwartz

BCOURTNEY
06-05-2008, 08:42 PM
Permit me to play devil's advocate here but I am playing (and advocating) an anti-kelly approach.

In the last six months at HorseStreet we have made tremdnous strides in the one-horse-per-race paradigm. We've concentrated on using the Monty Hall Problem to concntrate 2/3 of our top3's wins into a single horse and it is working.


Monty Hall Simulator
http://math.ucsd.edu/~crypto/Monty/monty.html

Monty Hall Explanation
http://montyhallproblem.com/


This has resulted in phenomenal hit rates with a play in every race.

(Yes, I know it sounds like I am a huckster. Can't help it 'cause it is true. Yes, we play a lot of chalk, but we hit enough price horses to be profitable.)

In reality, most of us have dialed back the hit rate a little to add value to the equation and have settled in at 27-35% hit rates with solid profitability.


With these hit rates (and, of course, profit) we have gone to a light chase-the-money strategy, which makes one almost unbeatable in session play. Personally, I use Kelly to correctly size my sessions but actually play this "anti-kelly" within the session.

There is no way that I can advocate a straight Kelly approach - fractional Kelly is just Kelly under a different disguise.

Yes, Kelly grows your theoretical bankroll faster than anything else. It just rarely works for mere mortals in the real world.

Kelly reduces your advantage. And I see that is a bad thing. If I can make a little less growth (say 2% less) but wager only 1/2 the money to do it, I consider that a huge improvement.


Regards,
Dave Schwartz

I think by implimenting a Monty strategy that you are avoiding the lure of choice that is exhibited by individuals investing in a horse race market. This can be a very good thing. The idea that trading activity alone reduces or eliminates behavioral biases in the market as a whole is unfounded. This is also counter-intuitive. There is a strong link between the lure of choice at the individual and with the lure of choice at the market level. The proportion of individuals exhibiting a lure of choice is an important parameter affecting the strength of the market. When you have contributions into the market which are simply errors in judgement in addition to selections which involve lure of choice this can have a very large effect and take prices in the opposite direction; even more so than just the errors in judgement alone.

Use of Monty is admission that we ourselves are irrational and subject to the lure of choice (because humans are so poor at estimating probabilities individually). The bias provided by lure of choice does in fact "survive" and exist in a larger market, something that is not intuitively obvious, and that "trading activity" alone does not eliminate the bias.

I'm not sure about your Monty implimentation, but I have some suggestions as to how you should run your selections in terms of number of doors for testing
reasons. Message private if you like.

On the note of Kelly. If you don't like it, I have to assume that the implimentation that you used was a garden variety form found on countless websites, that fail to capture the entire equation as given in original form, which quite frankly leads to vastly different outputs in terms of what is optimal. It's a fairly sensitive equation when "rough-shotted" as I see it done frequently has terrible consequences. I would encourage you to keep an open mind and perhaps entertain the idea that your implimentation of it was far off the mark and revisit it sometime when you get Monty out of the lab and are bored.

Thanks for a nice alternative look at things.

Cheers,
B

BCOURTNEY
06-05-2008, 08:51 PM
John Maynard Keynes’ “The Beauty Contest”:

“...professional investment may be likened to those newspaper
competitions in which the competitors have to pick out the six
prettiest faces from a hundred photographs, the prize being
awarded to the competitor whose choice most nearly corresponds
to the average preferences of the competitors as a whole; so
that each competitor has to pick, not those faces which he
himself finds prettiest, but those which he thinks likeliest
to catch the fancy of the other competitors, all of whom are
looking at the problem from the same point of view. It is not
a case of choosing those which, to the best of one’s judgment,
are really the prettiest, nor even those which average opinion
genuinely thinks the prettiest.

We have reached the third degree where we devote our intelligences
to anticipating what average opinion expects the average opinion
to be.”

J.M. Keynes; General Theory, p.156, 1936

Dave Schwartz
06-05-2008, 08:52 PM
after studying the links, I feel I have a good grasp on the puzzle. So you are saying that by successfully translating the problem picking one of your top 3 horses you win about 2/3 TIMES your top 3 hit rate (62%) in picking a single horse (give or take for confidence intervals.)? That translates to over 40% winners.
(I won't spoil the puzzle for everyone else)

That is fascinating work especially since there is no opportunity in racing to "switch" horses so I would guess it is dependent on accurately assessing the probabilities of each the three horses or assigning them some value based on how they are being bet.





Andy,

Yes, you have it precisely correct.



(Of course I don't expect you to tell us!)

Well, I am doing fee-based seminars on the topic. The next such seminar is June 28. We've got a bunch of users who have attended these workshops already. One of them made a recent post on our BBS:

http://www.horsestreet.com/ubb/Forum4/HTML/001822.html


The first such seminar I did was an in-person seminar in Reno. Read about it here:

http://www.horsestreet.com/ubb/Forum4/HTML/001722.html

If you read towards the bottom, you will see some of the other players' results at that seminar. (BTW, we eventually determined that the 7th palyer was doing it all wrong. - That was facorsig - the one I took to the dentist. Now he has the issues corrected and is doing much better.)


The thing that I find amazing about this is that the entire Monty Hall Problem seems to fly directly in the face of any statistical analysis. I mean, really... three doors - each has 1/3 chance. How do you make that work for you? Yet it does. (Yes, I understand how and why and the whole Bayesian thing, nevertheless, it flies in the face of logic.)

The application to horse racing was hugely philosophical to begin with. Now it is reality. Has to be seen to be believed.

In reality, most of us have actually encouraged the hit rate to drop a bit in favor of a little more value. I expect that at least three guys from the first two workshops will be playing at the professional level in the near future.

One of the amazing by-products of this method of play is the horses you don't bet. For example, if my original 3 contenders produce 60% of all winners - that would be 60-for-300 bets - and my play is going 40-for-100, then the remaining 200 contenders divide up the remaing 20 winners, going 20-for-200. That is precisely how it works.

Typically, I will be on the favorite about 40% of the time. Those bets are barely break even. The favorites I don't bet lose about 40% per wagred dollar.

One of our users actually developed a method of play where he exploits the races where he doesn't bet the favorite! Seems the 2nd choices win like favorites in those races!

Ultimately, it is all about value.


Dave

PS: For anyone interested, here is the link to the workshop:

http://www.practicalhandicapping.com/desktop/hsuniv/index.htm#ProWorkshop:_Monty_Hall

TrifectaMike
06-05-2008, 09:01 PM
Permit me to play devil's advocate here but I am playing (and advocating) an anti-kelly approach.

In the last six months at HorseStreet we have made tremdnous strides in the one-horse-per-race paradigm. We've concentrated on using the Monty Hall Problem to concntrate 2/3 of our top3's wins into a single horse and it is working.


Monty Hall Simulator
http://math.ucsd.edu/~crypto/Monty/monty.html

Monty Hall Explanation
http://montyhallproblem.com/


This has resulted in phenomenal hit rates with a play in every race.

(Yes, I know it sounds like I am a huckster. Can't help it 'cause it is true. Yes, we play a lot of chalk, but we hit enough price horses to be profitable.)

In reality, most of us have dialed back the hit rate a little to add value to the equation and have settled in at 27-35% hit rates with solid profitability.


With these hit rates (and, of course, profit) we have gone to a light chase-the-money strategy, which makes one almost unbeatable in session play. Personally, I use Kelly to correctly size my sessions but actually play this "anti-kelly" within the session.

There is no way that I can advocate a straight Kelly approach - fractional Kelly is just Kelly under a different disguise.

Yes, Kelly grows your theoretical bankroll faster than anything else. It just rarely works for mere mortals in the real world.

Kelly reduces your advantage. And I see that is a bad thing. If I can make a little less growth (say 2% less) but wager only 1/2 the money to do it, I consider that a huge improvement.


Regards,
Dave Schwartz

I don't get the connection to the Monte Hall Paradox. I'm assuming you are using Bayesian inference, and a scoring strategy based on prior and posterior probabilities of each proposition (potential bet). The only connection to the Monte Hall Paradox is that both can be solved by Bayesian analysis.

Good luck... The only issue is the confidence level of your 3 pontential candidates. Your prior probabilities must be very good.

Mike

BCOURTNEY
06-05-2008, 09:31 PM
I don't get the connection to the Monte Hall Paradox. I'm assuming you are using Bayesian inference, and a scoring strategy based on prior and posterior probabilities of each proposition (potential bet). The only connection to the Monte Hall Paradox is that both can be solved by Bayesian analysis.

Good luck... The only issue is the confidence level of your 3 pontential candidates. Your prior probabilities must be very good.

Mike

I would pre-suppose that data that was not normally distributed (having fat tails) would adversely effect the door selection methodology as well, and thus the system would work best and perhaps only when presented with stable odds ranges, e.g 5-1 or less. I do adovocate his position in a sense that it attempts to seperate the lure of choice from errors in judgement which is desireable, I think it could be shown that Dave through his system is not allowing ( or limiting ) his customer's "ability" to be influenced by lure of choice and letting them focus on their errors in judgement for their developed handicapping method(s). I can think of a way to use 4 doors and actually "measure" lure of choice. Maybe that would be helpful in identifying errors that my system or I have made in judgement after seperating out the nonsense, his system would be very interesting taken to the next level. :ThmbUp:

BCOURTNEY
06-05-2008, 09:35 PM
I'm using a database to find the plays, and the only thing I require is minimum odds. Some plays may be 3 to 2, some 5 to 1, etc. However, it doesn't matter if the horse selected is one click above the minimum or 50. It works well for me.

Thanks cj, looks like you have a high band and low band filter. Computer then you. :)

BCOURTNEY
06-05-2008, 09:39 PM
Hi BCOURTNEY:

Not really as I am usually wagering earlybird. If I am at the track I will occasionally play a round robin show parlay using horses at longer odds that I believe might have a real shot at getting in the money.

Jimmie

You might consider or backtest the place pool for your 2 or 3 or 4 to 1's, seems to offer me more value in my selection methods, you might find the same.

Cheers,
B

Dave Schwartz
06-05-2008, 10:47 PM
I don't get the connection to the Monte Hall Paradox. I'm assuming you are using Bayesian inference, and a scoring strategy based on prior and posterior probabilities of each proposition (potential bet). The only connection to the Monte Hall Paradox is that both can be solved by Bayesian analysis.

Doesn't sound much like what we are doing.


Dave

Dave Schwartz
06-05-2008, 10:49 PM
I would pre-suppose that data that was not normally distributed (having fat tails) would adversely effect the door selection methodology as well, and thus the system would work best and perhaps only when presented with stable odds ranges, e.g 5-1 or less. I do adovocate his position in a sense that it attempts to seperate the lure of choice from errors in judgement which is desireable, I think it could be shown that Dave through his system is not allowing ( or limiting ) his customer's "ability" to be influenced by lure of choice and letting them focus on their errors in judgement for their developed handicapping method(s). I can think of a way to use 4 doors and actually "measure" lure of choice. Maybe that would be helpful in identifying errors that my system or I have made in judgement after seperating out the nonsense, his system would be very interesting taken to the next level.

COuld you explain what "lure of choice" means?

As for the rest of it, I am not sure what you said.


Dave

BCOURTNEY
06-05-2008, 11:05 PM
Doesn't sound much like what we are doing.


Dave

It is.

http://en.wikipedia.org/wiki/Bayes%27_theorem
Example #4: The Monty Hall problem

BCOURTNEY
06-05-2008, 11:08 PM
COuld you explain what "lure of choice" means?

As for the rest of it, I am not sure what you said.


Dave

These papers provide a good description. Also, a 4 door model of
Monty to boot. In summary I was paying your system a compliment
by reducing the lure of choice a typical person making a wager has.
Proof you have positive ROI now because of it shows its real prevalence in the horse wagering market.

http://www.lse.ac.uk/collections/operationalResearch/pdf/lseor02-54.pdf
http://ideas.repec.org/p/pra/mprapa/2749.html

Cheers.

facorsig
06-05-2008, 11:26 PM
If you read towards the bottom, you will see some of the other players' results at that seminar. (BTW, we eventually determined that the 7th palyer was doing it all wrong. - That was facorsig - the one I took to the dentist. Now he has the issues corrected and is doing much better.)


Can't I be remembered as the guy who flew 8,000 miles to the workshop instead of the guy with the root canal or the guy who screwed up Monty! Just to clarify, both my teeth AND my bankroll are doing better after the workshop!

Just a quick confirmation to others following this thread that Dave's proposed approach is successful. I posted some statistics on another thread, but effectively I have doubled my bankroll for both of two ADW accounts.

Fred

Dave Schwartz
06-05-2008, 11:32 PM
I did not mean to be insulting in any way. I simply did not understand what you meant.

What I have (essentially) done is deconstructed why Monty works. Everyone keeps telling mee it is simple Bayesian analysis. That may be so - I have only a vague understanding of Bayes. To me it seems akin to a lot of Venn diagrams.

Anyway, as I said, I deconstructed why it works from a philosophical standpoint. Once I understood it I began looking for an analogy to horse racing.

Ultimately, it worked out but I am always open to learning more. I still don't get "lure of choice" in our application since everything we do is deterministic in nature. "Lure of choice" sounds like people making intuitive or emotional decisions.


If you could clear that up I'd be really appreciative.


Regards,
Dave Schwartz

TrifectaMike
06-06-2008, 12:06 AM
I don't get the connection to the Monte Hall Paradox. I'm assuming you are using Bayesian inference, and a scoring strategy based on prior and posterior probabilities of each proposition (potential bet). The only connection to the Monte Hall Paradox is that both can be solved by Bayesian analysis.

Doesn't sound much like what we are doing.


Dave

My bad. I'm made too many assumptions. I quess I don't understand what you're doing... maybe you're playing a game with the toteboard as the adversary ...probably another bad assumption.

However, I do know that the Monte Hall Paradox can be easily solved by Bayesian analysis with proper selection of prior and posterior probabilities. And from the process you described it seems to allow for a Bayesian approach.

Mike

GameTheory
06-06-2008, 12:16 AM
The Monty Hall problem is an example of the Bayes Theorem in action in a very simple form. Monty works because the Bayes Theorem is true. So if you truly deconstructed how Monty works, then you must have "discovered" the Bayes Theorem. In a very general sense, the Bayes Theorem is a model for learning itself.

How to get this to work for you in horse-racing (in the way you describe) I have no idea. You *almost* explained it to me one day, and I played around with it some, but did not achieve any breakthroughs. The thing is, we have no new information to add "at the end" when handicapping except the toteboard odds, which if applied do what they always do but nothing magically new that shoots my hit rate and ROI through the roof. And any info beyond the toteboard odds is not "new" -- we would have to have left it out of the handicapping process deliberately just in order to use it in the Monty Hall process. It seems we could have applied it earlier and come up with the same results.

You seem to give the credit to the process rather than the data it uses. So I'd sure be curious what the secret is...

BCOURTNEY
06-06-2008, 12:20 AM
I did not mean to be insulting in any way. I simply did not understand what you meant.

What I have (essentially) done is deconstructed why Monty works. Everyone keeps telling mee it is simple Bayesian analysis. That may be so - I have only a vague understanding of Bayes. To me it seems akin to a lot of Venn diagrams.

Anyway, as I said, I deconstructed why it works from a philosophical standpoint. Once I understood it I began looking for an analogy to horse racing.

Ultimately, it worked out but I am always open to learning more. I still don't get "lure of choice" in our application since everything we do is deterministic in nature. "Lure of choice" sounds like people making intuitive or emotional decisions.


If you could clear that up I'd be really appreciative.


Regards,
Dave Schwartz

Oh I've found the disscussion anything but insulting, it is refreshing and enjoyable, few take the time to step outside the box, any comments I'm making are as a fellow concerned player with similar interests hoping to provide some constructive critisim. Others, perhaps like myself have explored these techniques in some form (albeit mine were a little more formal mathmatically) looking for edges. The lure of choice is a deterministic characterization of people's behavior that you can count on, and thus measure. In doing so it's influence on the market for a given race can be predicted and accounted for in both a selection and a wagering strategy.
It's influence interestingly is different in both regards ( both to selection and wagering ). Like you I prefer discrete solutions and so does my cpu. I like to estimate and divide up the public in a few ways by estimating not just their bias in a general sense, but actually formally breaking into pieces, and seperate biases, from errors in judgement, which is subtle but important to recognize in a general computer model if you want to play all or most races. In the end though I get a number to work with and plug in to an equation. I was curious if you had computed or derived the bias statistics that exist when your Monty method is not applied. My supposition was a measurement of a collection of just your users would even reflect the market as a whole? Additionally, if this was a consistent number based of some race characteriztion, then Monty could be skipped and turned into just another figure to plug in with the rest without having to simulate trials on each race?

BCOURTNEY
06-06-2008, 12:27 AM
The Monty Hall problem is an example of the Bayes Theorem in action in a very simple form. Monty works because the Bayes Theorem is true. So if you truly deconstructed how Monty works, then you must have "discovered" the Bayes Theorem. In a very general sense, the Bayes Theorem is a model for learning itself.

How to get this to work for you in horse-racing (in the way you describe) I have no idea. You *almost* explained it to me one day, and I played around with it some, but did not achieve any breakthroughs. The thing is, we have no new information to add "at the end" when handicapping except the toteboard odds, which if applied do what they always do but nothing magically new that shoots my hit rate and ROI through the roof. And any info beyond the toteboard odds is not "new" -- we would have to have left it out of the handicapping process deliberately just in order to use it in the Monty Hall process. It seems we could have applied it earlier and come up with the same results.

You seem to give the credit to the process rather than the data it uses. So I'd sure be curious what the secret is...

I think the secret is that the player has his lure of choice weakened as an innocent byproduct of the application of Monty. Now this doesn't reduce the ability for a player to make errors in judgement, but reduces bias. You get passed through a filter or "game" that makes you more of a Bayes Estimator of sorts and sensitive to probabilities and less human, this apparently is a good thing.

BCOURTNEY
06-06-2008, 12:30 AM
The Monty Hall problem is an example of the Bayes Theorem in action in a very simple form. Monty works because the Bayes Theorem is true. So if you truly deconstructed how Monty works, then you must have "discovered" the Bayes Theorem. In a very general sense, the Bayes Theorem is a model for learning itself.

How to get this to work for you in horse-racing (in the way you describe) I have no idea. You *almost* explained it to me one day, and I played around with it some, but did not achieve any breakthroughs. The thing is, we have no new information to add "at the end" when handicapping except the toteboard odds, which if applied do what they always do but nothing magically new that shoots my hit rate and ROI through the roof. And any info beyond the toteboard odds is not "new" -- we would have to have left it out of the handicapping process deliberately just in order to use it in the Monty Hall process. It seems we could have applied it earlier and come up with the same results.

You seem to give the credit to the process rather than the data it uses. So I'd sure be curious what the secret is...

Couldn't "at the end" represent the final pool totals, and the beginning the opening totals?

LottaKash
06-06-2008, 12:51 AM
Lotta,

So, you are saying that you play horses that are basically break-even or a little loser and make money from this?


Regards,
Dave Schwartz

PS: Could you please turn off the caps lock? In BBS language, that is the equivilent to shouting.


Sorry about the caps lock thing.......I wasn't trying to shout, just cute.....

No, what I am advocating, is having a good sound betting method, that has built in safeguards so that, when things are not going quite your way, it has a built in failsafe device.....What I mean, especially of late, I pick more than my fair share of winners, and they pay more than 2/1, most of the time....But, there are others here who are not quite as advanced as some of us, as they are still learning and struggling to get the hang of finding winners and making money doing it......This method allows you to always be in a good comfort zone, and that, is healthy to your mental well being and helps keep you confidence level up, and it allows you to always have some cash for your next wager.......Over the years, I have experienced Gamblers Ruin more times than I care to admit, and I would always save up and keep coming back for more.....But, I have to tell you, if one were to implement this betting method into his game, perhaps he would still lose, but it would take longer, and it would cut down on a lot of wasted cash, and allow for more wagers and possibly more winners... So for me, I wish I had sonmething like this many years ago, as it would have been good for my bankroll as well as my mental well being...Hey,most of us are trying to earn while we learn, so If one want to survive for any length of time, one must have at least a rational way of handling their money, and this method allows that......

Thru the years, my handicapping life has been like a roller-coaster ride, and I believe, that any of us who have been doing this for awhile, have been in those shoes also......The more successful handicappers are people who have paid dearly for their hard won success'....I have made some very stupid and large wagers and I have had good paying winners with only token amounts going on it......so, this method is perfect for keeping your money under control, and saves the devastation that comes from plunging....I remember on many occasions plunging in the final race of the day, hoping to score before I went home.........sometimes it would work, but mostly, I would go home talking to myself all the way.......Does this sound familar to any of you ?
If it doesn't, than I question where you are in your handicapping life......It would be like saying, I am a skilled and experienced Motorcycle rider, and I never laid my bike down....haha...

For me whenver my bankroll goes to $1,500 I take the profits and regress to a $1000 bankroll and go back to 5% and before I know it starts to take off again, before long I am back to 8% bets which is $120 for prime plays and $84 for everyday wagers.....then the 1500, take 500 profits and back to $50/35 bets and never a scary day in between.........provided you maintain your successful hit rate.....but say you lose 10 in a row, each time you lose you would be lowering your bet by 1/4% and you are cutting your losses while in the slump and when you start winning again (and you will if you are good enough), you will be increasing you wagers by 1/2%, so that each time you win you will have more going on the next bet without getting cocky or stupid...Just good money control.....This works perfect for me as I am now retired and live on a fixed income, and the extra profits provide me with the perks that I wouldn't have had otherwise.....living the dream, and at long last all those years of handicapping toil are fading into a memory......"Viva Hoss Racin"':jump:

I think for the middle of the road handicappers, getting 33% winners @ 2/1 is a very realistic goal, and this method will help those who are still learning, to keep their wits about them at all times, so that this game remains fun for them...and they won't be going back to the bank to refuel, as often as they may have in the past...that's all......

G.L. in everything

best,

Dave Schwartz
06-06-2008, 12:57 AM
Lotta,

What an intelligent post. Thank you.

:ThmbUp:


Dave

LottaKash
06-06-2008, 01:10 AM
Lotta,

What an intelligent post. Thank you.

:ThmbUp:


Dave

Thx for the Thx, Dave, coming from you, very meaningful......:cool:

TrifectaMike
06-08-2008, 12:14 AM
I don't get the connection to the Monte Hall Paradox. I'm assuming you are using Bayesian inference, and a scoring strategy based on prior and posterior probabilities of each proposition (potential bet). The only connection to the Monte Hall Paradox is that both can be solved by Bayesian analysis.



Originally Posted by Dave Schwartz
Doesn't sound much like what we are doing.


Dave



My bad. I'm made too many assumptions. I quess I don't understand what you're doing... maybe you're playing a game with the toteboard as the adversary ...probably another bad assumption.

However, I do know that the Monte Hall Paradox can be easily solved by Bayesian analysis with proper selection of prior and posterior probabilities. And from the process you described it seems to allow for a Bayesian approach.

Mike

Dave, I believe I do know what you're doing. Putting it in terms of Bayesian analysis, prior and posterior probabilities is not intutive. So let me give a simple description of your process without any technical terms.

1. You have three contenders, H1, H2, and H3 (Your 3 doors...the Monthy Hall thing)

2. The contenders cumulative win rate is 60% (You didn't define how the win rate is distributed among the contenders, but that is only a minor factor)

3. You select an initial win candidate ( I assume it to be H1. You are probably using a scoring or ranking function to determine the order of preference)

4. At this stage you want to reduce your contenders from three to two. You require some "new" information. In the Monty Hall thing, Monty chooses a door, which shows a "no" prize. You require a simliar process.

5. You need to find your Monty. How can you do this? What or who will be your Monty. I can think of several ways of doing it.

A) You can use another selection process ( different than the one that gave you H1,H2,H3)
B) You can setup an opponent player, who also makes selection (As accurate of you)

....there are more, but I believe that's the idea.

6. You will now use this "Monty" to tell you a probable loser contained in your contender list,H1, H2, H3.

7. You now have 2 contenders remaining. One is your initial preference (choice).

8. You now use what many consider the illogical choice (Monty Hall Paradox) switching from your initial choice to the remaining H, which is not your initial selection.

Dave, I'm not in any way trying to disclose your methodology. I'm sure you've invested lots of time and energy in making the process work for you. With that said I will not go further into this subject, unless Dave wants to give a more detailed description.

Mike

Dave Schwartz
06-08-2008, 12:22 AM
Mike,

You have summed it in a nutshell. There are a couple of small twists and turns, but basically you have it (except point #3, which you have backwards).

The logic behind it is just amazing, as are the results.


There are a couple of refinements but what you have defined is a great starting point. Try it. You may be very pleased with the results.


Dave

badcompany
06-08-2008, 12:32 AM
I think it was Joe Dimaggio who said, "Swing hard in case you hit something."

That's my take on betting.

Dave Schwartz
06-08-2008, 12:47 AM
Personally, I am a contact hitter: I am for steady, consistency.

Wizard of Odds
06-08-2008, 12:55 PM
Mike,

You have summed it in a nutshell. There are a couple of small twists and turns, but basically you have it (except point #3, which you have backwards).

The logic behind it is just amazing, as are the results.


There are a couple of refinements but what you have defined is a great starting point. Try it. You may be very pleased with the results.


Dave

"Backwards" could mean you choose H3 initially.....or
closer to the Monty Hall problem, one could initially choose amongst H1, H2, and H3 randomly.

Another thought is that - don't we really want the intial three to be the 3 highest contenders of the three highest ROI, with some consideration also of high win probabilities? - that delicate balance

Dave Schwartz
06-08-2008, 02:46 PM
Wait for the book. ;)

Wizard of Odds
06-09-2008, 07:08 AM
Wait for the book. ;)

Feel free to send me a Private Message :)

Overlay
06-10-2008, 06:45 AM
I think that going with your second or third pick is a suckers bet because you are basically betting against yourself.

Even if you consider playing only your top-rated horse (which can certainly be a valid approach), you'll do better if you have a means of judging (by either experience, intuition, or quantitative data) whether the horse has a greater chance of winning than its odds indicate. The disparity between the odds and your estimate of the horse's winning probability can then also serve as a basis for wager sizing (as with Kelly or fractional Kelly, for example). And that same logic can be extended to horses below your top pick, to make them possible favorable betting propositions. It's not just about the absolute winning chances of any particular horse, but about those chances in relation to the horse's odds (whether it's the "best" horse in the field or not).

After the Belmont this past Saturday, whenever this question/issue comes up from now on, instead of giving that rather protracted response above, I'm just going to provide the following link and say "Da' Tara".

http://www.paceadvantage.com/forum/showthread.php?t=47558

pantoner
06-11-2008, 01:51 PM
Wait for the book. ;)

When is the book coming out?

Dave Schwartz
06-11-2008, 02:12 PM
When is the book coming out?

Perhaps it could be on your Christmas list?

BCOURTNEY
06-11-2008, 06:05 PM
Permit me to play devil's advocate here but I am playing (and advocating) an anti-kelly approach.

In the last six months at HorseStreet we have made tremdnous strides in the one-horse-per-race paradigm. We've concentrated on using the Monty Hall Problem to concntrate 2/3 of our top3's wins into a single horse and it is working.


Monty Hall Simulator
http://math.ucsd.edu/~crypto/Monty/monty.html

Monty Hall Explanation
http://montyhallproblem.com/


This has resulted in phenomenal hit rates with a play in every race.

(Yes, I know it sounds like I am a huckster. Can't help it 'cause it is true. Yes, we play a lot of chalk, but we hit enough price horses to be profitable.)

In reality, most of us have dialed back the hit rate a little to add value to the equation and have settled in at 27-35% hit rates with solid profitability.


With these hit rates (and, of course, profit) we have gone to a light chase-the-money strategy, which makes one almost unbeatable in session play. Personally, I use Kelly to correctly size my sessions but actually play this "anti-kelly" within the session.

There is no way that I can advocate a straight Kelly approach - fractional Kelly is just Kelly under a different disguise.

Yes, Kelly grows your theoretical bankroll faster than anything else. It just rarely works for mere mortals in the real world.

Kelly reduces your advantage. And I see that is a bad thing. If I can make a little less growth (say 2% less) but wager only 1/2 the money to do it, I consider that a huge improvement.


Regards,
Dave Schwartz

Dave,

I was re-reading this thread and I'm interested in something and want to understand better where you are coming from. You advocate that your selection methods are anti-kelly. Kelly methods apply to the wagering strategy after the selections are made and expectations ( or roughly speaking the information error rate is determined ) are calculated. Why or how are you co-mingling these items? Do you not view them as seperate processes or are they one in the same in your opinion? The selections might alter the wagers, but the wagers or betting methods don't really change the selections do they?

Cheers,
B

GameTheory
06-11-2008, 06:19 PM
Dave,

I was re-reading this thread and I'm interested in something and want to understand better where you are coming from. You advocate that your selection methods are anti-kelly. Kelly methods apply to the wagering strategy after the selections are made and expectations ( or roughly speaking the information error rate is determined ) are calculated. Why or how are you co-mingling these items? Do you not view them as seperate processes or are they one in the same in your opinion? The selections might alter the wagers, but the wagers or betting methods don't really change the selections do they?His betting methods are anti-kelly. I don't think he said his "selections" were anti-kelly. Pick up "The Opponent Method" from horsestreet.com and you will understand...

BCOURTNEY
06-11-2008, 06:35 PM
His betting methods are anti-kelly. I don't think he said his "selections" were anti-kelly. Pick up "The Opponent Method" from horsestreet.com and you will understand...

I visited the site, I found various papers but none titled the opponent method, though there is discussion of it's application on some of the public forums. Maybe you or Dave could provide a link?

GameTheory
06-11-2008, 06:38 PM
I visited the site, I found various papers but none titled the opponent method, though there is discussion of it's application on some of the public forums. Maybe you or Dave could provide a link?It's for sale:

http://www.horsestreet.com/products/index.html

Scroll down.

TrifectaMike
06-25-2008, 01:17 PM
Monty with Unequal Probabilities (Dave's Monty Hall Solution)

Sorry Dave, I couldn't resist it. But hey this should help your
marketing effort.

I'll keep this simple...no math involved. The results I'll present
will not be exact (Unless I use a Bayesian analysis...
good enough for this discussion).

Strictly speaking the Monty Hall Paradox has an initial assumption
that a winning prize is placed behind a door at random. Thus, each
door has an equal chance of being the winner.

Suppose we have information that the doors have different probabilities of
being the winner... Using Dave's three contender (H1, H2, H3) approach, each with a known win rate (w1, w2, w3)... For example, by using good recording
procedures, Dave knows that over time H1 has a win rate of w1 = 40%, H2 has a win rate of w2 = 35%, and H3 has a win rate of w3 = 25%. (I just gave Dave credit for having all winners within his three contenders...He's not that good, but it does make it easier to describe the results).

Let's pick H1 as our intial choice.( It seems to be the most likely choice).
Daves's Monthy Hall Engine reveals H2 as the "goat" (loser). He has selected
a horse with a 35% win rate as the loser. What do you do? Rejoice! Because
you have selected the 40% horse, H1, which has the highest initial probability of being the winner or do you switch to H3?

Here are the results:
1. Stay with your initial choice, H1, win 40% of your wagers.
2. Switch to H3, win 60% of your wagers.

The choice is rather clear.

But is this best we can do?

No! We can do better.

Let's pick H3, the horse least likey to win. Daves's Monty Hall Engine reveals either H1 or H2 as the "goat" (loser)

Here are the results:

1. Stay with your initial choice, H3, win 25% of your wagers.
2. Switch to either H1 or H2 (Dependent on the "goat" selected by Dave's Monty Hall Engine) and win 75% of your wagers.

Dave, nice job. Especially since you've done it all intutively.

Mike

Wizard of Odds
06-25-2008, 01:31 PM
Monty with Unequal Probabilities (Dave's Monty Hall Solution)

Sorry Dave, I couldn't resist it. But hey this should help your
marketing effort.

I'll keep this simple...no math involved. The results I'll present
will not be exact (Unless I use a Bayesian analysis...
good enough for this discussion).

Strictly speaking the Monty Hall Paradox has an initial assumption
that a winning prize is placed behind a door at random. Thus, each
door has an equal chance of being the winner.

Suppose we have information that the doors have different probabilities of
being the winner... Using Dave's three contender (H1, H2, H3) approach, each with a known win rate (w1, w2, w3)... For example, by using good recording
procedures, Dave knows that over time H1 has a win rate of w1 = 40%, H2 has a win rate of w2 = 35%, and H3 has a win rate of w3 = 25%. (I just gave Dave credit for having all winners within his three contenders...He's not that good, but it does make it easier to describe the results).

Let's pick H1 as our intial choice.( It seems to be the most likely choice).
Daves's Monthy Hall Engine reveals H2 as the "goat" (loser). He has selected
a horse with a 35% win rate as the loser. What do you do? Rejoice! Because
you have selected the 40% horse, H1, which has the highest initial probability of being the winner or do you switch to H3?

Here are the results:
1. Stay with your initial choice, H1, win 40% of your wagers.
2. Switch to H3, win 60% of your wagers.

The choice is rather clear.

But is this best we can do?

No! We can do better.

Let's pick H3, the horse least likey to win. Daves's Monty Hall Engine reveals either H1 or H2 as the "goat" (loser)

Here are the results:

1. Stay with your initial choice, H3, win 25% of your wagers.
2. Switch to either H1 or H2 (Dependent on the "goat" selected by Dave's Monty Hall Engine) and win 75% of your wagers.

Dave, nice job. Especially since you've done it all intutively.

Mike

If I understood this correctly, then one never ultimately bets on H3? One ends up betting on either H1 or H2?

... and of course, the goat is certainly not 0% probability.

TrifectaMike
06-25-2008, 09:18 PM
If I understood this correctly, then one never ultimately bets on H3? One ends up betting on either H1 or H2?

... and of course, the goat is certainly not 0% probability.

That's a question for Dave to answer, if he wishes to do so.

Mike

Wizard of Odds
06-25-2008, 09:30 PM
That's a question for Dave to answer, if he wishes to do so.

Mike

Fair Enough...but parsing what you wrote, that is the clear implication

Robert Fischer
06-25-2008, 09:31 PM
Bet Size should be small enough to avoid randomness in accord with your bankroll size.

That is the most important issue.

I am making positive expectation wagers. If I lose it is because I was wrong about my probabilities, not because of RANDOMNESS.

Getting to that point is the first step.

You may even improve upon that by wagering slightly more according to your better opportunities.


Right now I am using Bankroll*Value*hit%*0.033
rudimentary to simply multiply by a factor of 0.033, but it works for me at this point.

Note that I also crudely take hit% into accout twice. Once in the Value equation ($2price*Hit%/2) and once again on it's own. This pigeonholes my wagering to high% high value wagers and helps to avoid ruin from randomness.

Dave Schwartz
06-25-2008, 09:52 PM
:( Sorry, guys. You just have to wait for the book.

Cratos
06-25-2008, 10:08 PM
Ok, you've made your selections. Now it's time to click the mouse button or go to the window. Anyone care to share successes they have had in bet sizing and producing a positive bankroll over a significant amount of time? What types of wagers do you make and are you flat betting, kelly betting, fractional kelly, what percentage of your bankroll do you wager? Do you produce a odds line? Do you account for the pool sizes and amounts in each one?

Personally, I beleive that proper bet sizing is about 3 or 4 times more important than selection or prediction methods in order to produce a successful profitable system, if you disagree with this then please share why this is. I have always been curious why the focus of handicapping software is almost exclusively concerned with selections as opposed to the wager and bet size. Share experiences.

Thanks,
B
You raise an interesting question and my experience comes from the evolution of making the wrong wagers. I was a good handicapper long before I was a good bettor.

However I evolved to learn that wagering and bet size come from three primary factors: (1) bankroll, (2) patience and (3) discipline. You must have the cash to wager, therefore the bankroll. Patience is the ability to wait until the betting opportunity presents itself through your astute handicapping and you must have discipline to stick with your wager and exploit that opportunity.

One quick example for me was last year when I went to the BC at Monmouth to bet two horses which I shared on the PA forum. I only bet about 50-60 time a year and after losing a bag of “wooden nickels’ on Curlin in the Derby, I wagered only a handful of “wooden nickels” on him in the Preakness.

Because I don’t bet at odds less than 3-1 and I don’t bet exotics I didn’t have another opportunity to bet Curlin again until the BC Classic at Monmouth. However I also wanted to bet Discreet Cat who was running on Friday with the idea of rolling my winnings (if I had won) onto Curlin in the Classic. That didn’t happen because Discreet Cat went off at odds lower than my betting threshold and he lost.

On Saturday I decided to increase my bet size from 1 bag of “wooden nickels” to 2 bags because of not betting Discreet Cat on the previous day. Needless to say I was shock to see Curlin to go off at 4-1 and was ecstatic when he crosses the finish because I had just made and won with my largest bet size ever.