spilparc
12-07-2006, 03:38 PM
This idea was proposed by David Sklansky on his website. I thought I would repost it here.
http://tinyurl.com/y3jfsm
David SklanskyAdministrator
Carpal \'Tunnel
Reged: 08/28/02
Posts: 3420
Proposing a Profitable and Useful Horserace Betting Gizmo
12/06/06 07:17 PM
I doubt this can be patented, but if it can, you (hopefully) saw it here first (which officially protects it). With or without a patent, somebody out there should be able to transfer this idea, that should be simple to implement on a computer, onto something the size of a calculator. And then put my name on it, manufacture it for $10 a pop, and sell a couple of million of them for $39.95. That's fifty million bucks gross profit after subtracting my ten million endorsement fee.
Someone once said, "The race does not always go to the swiftest. Buts that's the way to bet." And it's true. Especially in thorougbred races as opposed to harness races. Sometimes the way the race is run causes the winner to be a horse other than the one who would have the best time in individual time trials for that distance. More often though the winner IS that horse, especially if his time trial is significantly faster.
Of course the tough part is knowing how well each horse would do in a time trial today on this particular track based only on past information about the horse. That is not my field of expertise and there are very few real experts out there. I think that Andy Beyer qualifies. And he has published speed ratings to help the public with this.
But those speed ratings predict the winner less than a third of the time. One reason is the aforementioned fact that how the race is run can cause the best horse to lose. A second reason is that the rating is "wrong". In other words it doesn't accurately reflect that horse's time trial speed on this track.
The third reason is the inherent standard deviation in any measurement of this type. I'm going to use a distance of a mile and 70 yards because a decent time is exactly 100 seconds. The point is that even horses that have consistently shown that 100 seconds is their expected time do deviate from this time in almost every race. And for the most part, barring exceptional circumstances, like muddy tracks, the deviation follows the bell curve normal distribution. Which means that a good handicapper should be able to assign to each horse both an average time and a standard deviation to that time. Eg 100 seconds with a SD of .55 seconds.
But not all horses have the same standard deviation. A larger SD can come either from the fact that there is less good information to go by(its young, its coming from another track, its changing distances, its coming off a layoff) or from the inconsistency of the horse itself. This subject can be discussed in more detail at a later time.
The point though is that this standard deviation has as much to do with the chances of a horse winning (or placing or showing or an exacta or trifecta coming in) as his average time. For instance a if the second fastest horse is slightly faster than the third fastest horse it still might be less likely to WIN if the third fastest horse has a higher standard deviation.
The bottom line is that, at least in theory, if we knew the expected time for each horse and the standard devistion in the time for each of these horses, the appropriate computer program could simulate a million races with these parameters and spit out the chances for any bet we care to make. Of course the accuracy of those odds would be tempered both by the inexpertise of the user and the fact that the theory behind the whole idea is not flawless.
In spite of those caveats, I am fairly sure that such a device could be very helpful in the hands of any good handicapper. And it would be purchased by many weak handicappers as well. So any electronic engineers out there wo are interested should let me know. But first a test.
Five Horse Race:
A 100 seconds. SD .41
B 100.1 seconds SD .55
C 100.2 seconds SD .32
D 100.3 seconds SD .35
E 100.4 seconds SD .85
I assume there are computer programs out there that would have no trouble quickly simulating at least a million races and giving me the price of all win, place, show, exacta, and trifecta combos. (I understand it could simply give me the price for each of the 120 permutations. I want the simplified answer.)
I know this might take a few hours. But remember to divide that into fifty million.
http://tinyurl.com/y3jfsm
David SklanskyAdministrator
Carpal \'Tunnel
Reged: 08/28/02
Posts: 3420
Proposing a Profitable and Useful Horserace Betting Gizmo
12/06/06 07:17 PM
I doubt this can be patented, but if it can, you (hopefully) saw it here first (which officially protects it). With or without a patent, somebody out there should be able to transfer this idea, that should be simple to implement on a computer, onto something the size of a calculator. And then put my name on it, manufacture it for $10 a pop, and sell a couple of million of them for $39.95. That's fifty million bucks gross profit after subtracting my ten million endorsement fee.
Someone once said, "The race does not always go to the swiftest. Buts that's the way to bet." And it's true. Especially in thorougbred races as opposed to harness races. Sometimes the way the race is run causes the winner to be a horse other than the one who would have the best time in individual time trials for that distance. More often though the winner IS that horse, especially if his time trial is significantly faster.
Of course the tough part is knowing how well each horse would do in a time trial today on this particular track based only on past information about the horse. That is not my field of expertise and there are very few real experts out there. I think that Andy Beyer qualifies. And he has published speed ratings to help the public with this.
But those speed ratings predict the winner less than a third of the time. One reason is the aforementioned fact that how the race is run can cause the best horse to lose. A second reason is that the rating is "wrong". In other words it doesn't accurately reflect that horse's time trial speed on this track.
The third reason is the inherent standard deviation in any measurement of this type. I'm going to use a distance of a mile and 70 yards because a decent time is exactly 100 seconds. The point is that even horses that have consistently shown that 100 seconds is their expected time do deviate from this time in almost every race. And for the most part, barring exceptional circumstances, like muddy tracks, the deviation follows the bell curve normal distribution. Which means that a good handicapper should be able to assign to each horse both an average time and a standard deviation to that time. Eg 100 seconds with a SD of .55 seconds.
But not all horses have the same standard deviation. A larger SD can come either from the fact that there is less good information to go by(its young, its coming from another track, its changing distances, its coming off a layoff) or from the inconsistency of the horse itself. This subject can be discussed in more detail at a later time.
The point though is that this standard deviation has as much to do with the chances of a horse winning (or placing or showing or an exacta or trifecta coming in) as his average time. For instance a if the second fastest horse is slightly faster than the third fastest horse it still might be less likely to WIN if the third fastest horse has a higher standard deviation.
The bottom line is that, at least in theory, if we knew the expected time for each horse and the standard devistion in the time for each of these horses, the appropriate computer program could simulate a million races with these parameters and spit out the chances for any bet we care to make. Of course the accuracy of those odds would be tempered both by the inexpertise of the user and the fact that the theory behind the whole idea is not flawless.
In spite of those caveats, I am fairly sure that such a device could be very helpful in the hands of any good handicapper. And it would be purchased by many weak handicappers as well. So any electronic engineers out there wo are interested should let me know. But first a test.
Five Horse Race:
A 100 seconds. SD .41
B 100.1 seconds SD .55
C 100.2 seconds SD .32
D 100.3 seconds SD .35
E 100.4 seconds SD .85
I assume there are computer programs out there that would have no trouble quickly simulating at least a million races and giving me the price of all win, place, show, exacta, and trifecta combos. (I understand it could simply give me the price for each of the 120 permutations. I want the simplified answer.)
I know this might take a few hours. But remember to divide that into fifty million.