A wild thought about a possible odds' sweet spot

[markgoldie]

A thought crossed my mind the other morning while I was waking up. It involved sort of a convergence of different thoughts.

Thought one: As we know, there is an uncertainty principle surrounding every horse who competes in a race. A poor performance may occur regardless of the animal's supposed fitness and form. For example, there may be a hidden sickness; there may be bumping, jostling, or even colliding during the race; and there may be lameness that is a result of the race itself (breaking down during a race is just the most extreme example of this and less drastic but performance-affecting pain is likely much more common).

Thought two: The percentage of horses winning at a specific odds follows a more or less smooth curve, such that the lower the odds, the better the win percentage of the horse who carries such odds. However, as the odds become very low, the win percentage bumps up against the uncertainty principle (thought one). This is why "bridge jumpers" wind up jumping if they persist in playing monster favorites to show, looking for the dreaded $2.10 payoff.

Thought three: It's clear that huge favorites carry a significant advantage over all rivals by any means of rational handicapping. But since each horse encounters the uncertainty principle as he enters the starting gate, this handicapping advantage will prohibit him from achieving anything near a 100% win percentage. One could argue, I suppose, that the upper limit of possibility is somewhere around 85%. (This is simply a guess and maybe someone has a better estimation). At any rate, the question arises, how much of a handicappable advantage is more than enough?

With thoughts one, two, and three as a background, a possible question arises: Could there be an odds' "sweet spot" in which the handicappable advantage of the horse is still palpable and clear and yet the payoff is large enough to compensate for the losses due to the uncertainty principle? I'm thinking that this spot could be in the 2-5 or 3-5 range.

Now. Before you all jump on me, I realize that every situation is different. That there can be "good" 2-5's and "bad" 2-5's. But I'm talking about a general situation, taken over large numbers of races. I also realize that playing a steady diet of 2-5 horses will no doubt produce a negative ROI. But if such an odds were a sweet spot, could this be an area where a targeted refinement of play might produce some results?

Any thoughts, statistics, etc.?

[SchagFactorToWin]

re: Thought Three

I use a figure that I have dubbed the 'WTF Factor'. This is the percentage of picks that finished out of the money, didn't break (I'm a harness guy), and weren't DQ'd. It's for when you say "What the f***, why did that horse lose?"

My 2009 WTF Factor is currently 14.5%.

[CBedo]

I don't know if I'd call it "the sweet spot," and I'm sure Anderon & Formula2002 will have different answers, but personally, with every model I have built, I have not been able to turn a profit with horses whose odds are below the 6/5-7/5 range. Of course, I haven't truly ever tried to build a low priced overlay model.

[GameTheory]

I don't know if I'd call it "the sweet spot," and I'm sure Anderon & Formula2002 will have different answers, but personally, with every model I have built, I have not been able to turn a profit with horses whose odds are below the 6/5-7/5 range. Of course, I haven't truly ever tried to build a low priced overlay model.At prices that low, betting to place probably has a greater ROI...

[CBedo]

At prices that low, betting to place probably has a greater ROI...Interestingly so far, the place return has been worse than the win return for the lowest priced horses in the current model I'm working with. The distribution of finishes for the lower priced horses seems to skew to "win or run out." The number of place and show finishes is much lower (so far) than I would have expected.

[markgoldie]

The basic idea would be that at 2-5, say, there is litle difference in handicappable domination over the field than the 1-5; that the only difference might be more of field size, higher fig-related advantages, or a somewhat more favorable pace profile, such that the extra advantage would be nothing more than "overkill." But I could be wrong about this. There might be justfiable differences in some area that distinguish the 1-5 and 2-5. I'm not at all sure and that's why I threw this out there.

[CBedo]

i think I agree with you that there is a "terminal velocity" type win percentage that sets the top end of win percentage. Some invisible asymptotic line that win percentage approaches but never quite gets to.

You are probably also correct that the line does move somewhat based on some factors (field size is the first thing that pops into my head).

Now what do we do with this?

[Norm]

Should it matter why a horse is a 2/5 favorite ?? Some examples :

1). A 2/5 favorite in a 10 horse field is a different thing than a 2/5 favorite in a 5 horse field. Also, the opportunity for things to go wrong is greater in a larger field.

2). A horse is a 2/5 favorite because he has superior Beyers numbers and the locals religiously bet according to Beyers numbers.

3). A horse is a 2/5 favorite because a locally popular trainer is executing a recognizable pattern, everyone sees it and jumps on board.

4). A horse is a 2/5 favorite because of a huge class drop, everyone sees it and jumps on board.

5). A horse is a 2/5 favorite because an off-track bettor doesn't understand that a big $ bet has a huge impact on the pool at a small track. The horse is not really the "public" favorite, just the pick of one bettor.

The reason why a horse is a 2/5 favorite may be important ??

[Robert Goren]

And there is always the "being stiffed" factor.;)

[Norm]

And there is always the "being stiffed" factor.;)
Yes ! ... and then there is the "crap-happens" factor like in the 8th at Presque Isle on Sep. 9th when the front runner ran into a flock of geese in the back stretch, lost all action and could not finish.:lol: (It's only funny because I didn't bet on him ! :) )

[markgoldie]

i think I agree with you that there is a "terminal velocity" type win percentage that sets the top end of win percentage. Some invisible asymptotic line that win percentage approaches but never quite gets to.

You are probably also correct that the line does move somewhat based on some factors (field size is the first thing that pops into my head).

Now what do we do with this?

I'm sure that the terminal velocity (as you put it) is real and exists in the aggregate, which is why traditional bridge jumping (that is, before rebates) doesn't work, even though we're only talking about running third.

This would mean that there exists an "impossibility" odds, such that anything bet at that price or under would always result in a loss no matter what. Clearly, a 1-20 ($2.10) payoff qualifies since it qualifies for even running third, let alone win. We must strongly suspect that 1-10 ($$2.20) qualifies as well since as Schag Factor posts, his WTF? rate is running at 14.5% in harness racing and I have very strong reason to believe (which I won't bother going into here) that the same factor is higher in t-bred racing.

Now we come to 1-5 ($2.40). Again, very probably under the line. Here, 83.3% winners is the break even point and I don't think this win percentage can be maintained over a large sample. At $2.60 (1.5-5), we're most likely standing right on the line and at 2-5 ($2.80), we're probably just over it. So as a general rule, I think 2-5 is the minimum odds you can ever accept with any reasonable expectation of long-term profit.

At 2-5, you would think or hope that the handicappable advantage of the horse is ironclad. Again, we're talking about the aggregate and not the specific. Were this basically true, then Schag's WTF? factor should be the only thing that could beat you. In this scenario, the public money that tips a $2.80 payoff to $2.60 or $2.40 is more of a random wagering pattern which has no basis in handicapping. Thus, the $2.80 number might be what I postulated as a betting "sweet spot."

Another way of putting this would be that somewhere along the line, as the odds get lower, the "smoothness" of the curve which relates odds to win percentage breaks down as it approaches a ceiling of possibility. We know that we must stay away from the ceiling.

Now. All this is probably wrong. But why? Well, the most logical answer would be that even though we know that the curve breaks down eventually as to win percentage, the curve never stops as to handicappable advantage. That is, there is a material difference in advantage between a $2.80 payoff horse and a $2.60 or $2.40 animal and that just because a $2.40 horse can never be made profitable, it doesn't mean that a $2.80 horse can.

On the other hand, since we suspect that there should be little difference in advantage between the $2.80 and $2.40-$2.60 horses, there could be a possibility of target-filtering of the $2.80's to produce a profit. But since this is far from my area of expertise, I wouldn't know. Maybe others would.

[fmolf]

I'm sure that the terminal velocity (as you put it) is real and exists in the aggregate, which is why traditional bridge jumping (that is, before rebates) doesn't work, even though we're only talking about running third.

This would mean that there exists an "impossibility" odds, such that anything bet at that price or under would always result in a loss no matter what. Clearly, a 1-20 ($2.10) payoff qualifies since it qualifies for even running third, let alone win. We must strongly suspect that 1-10 ($$2.20) qualifies as well since as Schag Factor posts, his WTF? rate is running at 14.5% in harness racing and I have very strong reason to believe (which I won't bother going into here) that the same factor is higher in t-bred racing.

Now we come to 1-5 ($2.40). Again, very probably under the line. Here, 83.3% winners is the break even point and I don't think this win percentage can be maintained over a large sample. At $2.60 (1.5-5), we're most likely standing right on the line and at 2-5 ($2.80), we're probably just over it. So as a general rule, I think 2-5 is the minimum odds you can ever accept with any reasonable expectation of long-term profit.

At 2-5, you would think or hope that the handicappable advantage of the horse is ironclad. Again, we're talking about the aggregate and not the specific. Were this basically true, then Schag's WTF? factor should be the only thing that could beat you. In this scenario, the public money that tips a $2.80 payoff to $2.60 or $2.40 is more of a random wagering pattern which has no basis in handicapping. Thus, the $2.80 number might be what I postulated as a betting "sweet spot."

Another way of putting this would be that somewhere along the line, as the odds get lower, the "smoothness" of the curve which relates odds to win percentage breaks down as it approaches a ceiling of possibility. We know that we must stay away from the ceiling.

Now. All this is probably wrong. But why? Well, the most logical answer would be that even though we know that the curve breaks down eventually as to win percentage, the curve never stops as to handicappable advantage. That is, there is a material difference in advantage between a $2.80 payoff horse and a $2.60 or $2.40 animal and that just because a $2.40 horse can never be made profitable, it doesn't mean that a $2.80 horse can.

On the other hand, since we suspect that there should be little difference in advantage between the $2.80 and $2.40-$2.60 horses, there could be a possibility of target-filtering of the $2.80's to produce a profit. But since this is far from my area of expertise, I wouldn't know. Maybe others would.i always look at the trainerstats for his/her numbers with favorites and horses below even money...these numbers tell the story in addition to traditional handicapping....also a solid favorite can be leveraged in the exacta pools or as a single in a serial wager

[zerosky]

When trying to crack the German Enigma code during WW2, code breakers in England led by Alan Turing
used a formulae to calculate ‘Weight of Evidence’ which Is 10Log10 of the Bayes Factor, expressed a Deciban.

According to the theory, one Deciban is the smallest unit of belief that humans are capable.
How does this relate to the odds of horse?

First, calculate the Bayes Factor, which is the ratio of the odds with evidence
(in other words the odds on the tote, or your own derived odds) divided by the odds without evidence
(or random odds) Where the odds are in the form of P/Q (P =Probability of winning and Q = (1-P)

In a 7 horse race the random odds are calculated as one chance of winning out of seven chances = 1/7

The tote odds have to be transformed, as they are expresses as Q/P so assuming the odds are 2/1 = 1/2

Therefore the Bayes Factor = ½ divided by 1/7 which equals 3.5, multiplying by 10Log10 gives approximately 5.4 Deciban.

According to the following table, you can calculate the ‘Weight of Evidence’

>100 Deciban – Decisive
>15 Deciban – Very Strong Evidence
>10 Deciban –Strong Evidence
>5 Deciban –Substantial Evidence
>0 Deciban –Barely worth mentioning
<0 Deciban –No Evidence

I realise that this is of no practical use but it is an interesting byway never the less

[Overlay]

In Winning at the Races, Quirin tested the proposition that horses win races at an overall rate corresponding to their odds. Although the data supported that proposition, there were isolated odds ranges where the number of actual winners exceeded the number of expected winners to a degree that was statistically significant (the kind of factors that Quirin designated with a single asterisk to denote positive independent variables). The positive odds ranges he found in that regard were 1-5, 1-1, 5-2 and 6-1.

[misscashalot]

In Winning at the Races, Quirin tested the proposition that horses win races at an overall rate corresponding to their odds. Although the data supported that proposition, there were isolated odds ranges where the number of actual winners exceeded the number of expected winners to a degree that was statistically significant (the kind of factors that Quirin designated with a single asterisk to denote positive independent variables). The positive odds ranges he found in that regard were 1-5, 1-1, 5-2 and 6-1.
Did he cover roi for these positive odd winners?

[Overlay]

Did he cover roi for these positive odd winners?

Unfortunately, he didn't cover wagering outcomes for his statistics. However, from a total sample of 19,075 horses, there were 15 that went off at 1-5, of which 13 won. Assuming a minimum payoff of $2.40, the 13 winners would have returned a minimum of $31.20 on $30.00 wagered, for at least a 4% profit. There were 188 even-money horses, of which 88 won; 765 horses at 5-2, of which 193 won; and 1,058 horses at 6-1, of which 138 won, none of which would have produced a flat-bet profit at minimum payoffs, although (as stated), they won at a significantly higher rate than their take-adjusted odds said they should have, and produced a result which (although still negative) was better than the percentage corresponding to the take plus breakage.

[overthehill]

My only thought here is that horses are not machines and that over the years I have seen numerous odds- on horses go down to the tubes. I believe more so in claimers than in allowance and stake races. With my current selection strategies Im pretty unhappy with less than 5-1. There are just too many things that can go wrong in a race and im more interesting in relative value than the best horse in the race on paper.

[DeanT]

This is an interesting question that has crossed my mind from time to time. Generally, when we say something is a lock, what's a lock? 85% chance?

I watch the betfair markets and it is cool to watch the sharp players there who will fade low priced sure things, even when it comes to in running. Something could happen, and I think the fade4rs take advantage of that at times.

As for the overall markets for horse racing they are very efficient at both high and low odds levels, without too much of an ROI difference. This is data which includes in running betting where you can bet a horse at 1.01 ten feet from the wire if you wish. So this betting represents a more perfect information market than in our tote.http://www.probabilitytheory.info/images/horses_impliedvsactual_odds.jpg

http://www.probabilitytheory.info/topics/efficiency_betting_market.htm