where are the overlays?

[Billnewman]

Someone told me early in my love affair with this game, the favorite in the last race of every card is underbet. He explained its account of people trying to "get out" so they naturally are betting long shots. Thats always lingered in my mind but lately i've been thinking where else are the overlays? Which leg of the pic 3 or pic 4 give you the best of it if you get a longshot home? Another overlay sticks in my head from early in my avocation is the 1989 KD. Easy Goer paid $2.60 to place $3.40 to show. That's when i bought Dr.Ziemba's book. I lost a lot of money trying to bet show overlays. At the smaller tracks ive noticed if the favorite does not have the correct proportion of money across the board, theres a good chance it doesn't hit the board. I came to the conclusion the connections of the horse ie:owners their friends the grooms the backside guard, whomever they bet to place and show. Last winter i messed around with it, but i bet through my bookie(who has a sports website) so i didn't effect the odds. I was up for a while but in the end i lost a grand and quit.

[cashmachine]

I also read about this psychological angle long time ago and one of the first things that I tested with my program was whether people start betting more aggressively longshots at the end of the day. I don't know about American tracks, but in Hong-Kong situation is just the opposite: average win odds of winner is decreasing as day progress.

Regarding second idea from Ziemba's book: this idea worked before he published his book. His book become very popular and today just about everybody knows this idea and huge crowds of handicappers trying to exploit it. In Hong-Kong there are many professional syndicates, and when we have a situation that horse has large place odds in comparison with win odds, they bet such a horse with insanely large money so place odds for such a horse practically always drop so much after the race start that it makes those horses bitter underlay. So in Hong-Kong it is just a rule: if you see a horse who is a good bet according to Ziemba, he will become very bad bet after the race start. I read research paper from two students who tried to exploit Ziemba's idea on American racetracks few years ago and they reported same thing: after the race start those horses always become underlays.

[traynor]

I also read about this psychological angle long time ago and one of the first things that I tested with my program was whether people start betting more aggressively longshots at the end of the day. I don't know about American tracks, but in Hong-Kong situation is just the opposite: average win odds of winner is decreasing as day progress.

Regarding second idea from Ziemba's book: this idea worked before he published his book. His book become very popular and today just about everybody knows this idea and huge crowds of handicappers trying to exploit it. In Hong-Kong there are many professional syndicates, and when we have a situation that horse has large place odds in comparison with win odds, they bet such a horse with insanely large money so place odds for such a horse practically always drop so much after the race start that it makes those horses bitter underlay. So in Hong-Kong it is just a rule: if you see a horse who is a good bet according to Ziemba, he will become very bad bet after the race start. I read research paper from two students who tried to exploit Ziemba's idea on American racetracks few years ago and they reported same thing: after the race start those horses always become underlays.

Given the number of professional-level race analysis apps that calculate mutuel pool inefficiencies automatically, that should be no surprise.

[Magister Ludi]

Someone told me early in my love affair with this game, the favorite in the last race of every card is underbet. He explained its account of people trying to "get out" so they naturally are betting long shots. Thats always lingered in my mind but lately i've been thinking where else are the overlays?

As far as the United States racetrack betting market is concerned, you are on the right track. It is not as simplistic as the last race of every card being underbet. However, if you do an analysis of the ordinal rank of each race vs the odds rank of each horse on a track-dependant basis, you may get some very interesting results.

[Some_One]

Given the number of professional-level race analysis apps that calculate mutuel pool inefficiencies automatically, that should be no surprise.

In HK, usually the field quality gets better through the card (the class 5's start and the Class 1's end) with TT races being the exception. The Class 5's can be quite random at times, especially on the AWT.

[Robert Goren]

The last race theory may have worked 25 years ago, but with advent simulcast and ADW betting, I think it doesn't work anymore. You have to do your research in order to find a niche bet that only occurs one a week or less. Any more often and the "whales" have found it and will bet the heck out of it. If you are bright and creative, you should be able to find several of these niche bets. A clue, the ones I have found are between 5/1 and 9/1 odds. I suspect most are. The ones at higher odds win too seldom to bet and the ones at lower odds don't have a positive ROI. A warning, the samples you base your bets will be smaller than you like, so adjust your expected win% down sharply and see if at the median payoff it is still profitable. good luck and happy hunting.

[Robert Goren]

Of all the books I have read on horse racing betting, Ziemba's book cost me the most money. He sounded so convincing.

[traynor]

The last race theory may have worked 25 years ago, but with advent simulcast and ADW betting, I think it doesn't work anymore. You have to do your research in order to find a niche bet that only occurs one a week or less. Any more often and the "whales" have found it and will bet the heck out of it. If you are bright and creative, you should be able to find several of these niche bets. A clue, the ones I have found are between 5/1 and 9/1 odds. I suspect most are. The ones at higher odds win too seldom to bet and the ones at lower odds don't have a positive ROI. A warning, the samples you base your bets will be smaller than you like, so adjust your expected win% down sharply and see if at the median payoff it is still profitable. good luck and happy hunting.

I am impressed. The only point I disagree with (modest disagreement, not vehement) is the 9/1 top end. I think that is track specific, as well as race (class/grade/distance) specific. Meaning there are situations (specific by track and race) in which defining the range as 5/1 to 9/1 is too restrictive. It is dismaying to toss out a horse that wins easily at 15/1.

I agree that the low end is important, but lumping all >X into one category can be misleading. Discovering that "this high but no higher" point for each situation is useful. Of course, that presupposes cleaning data and correcting for outliers--the smaller the sample, the more distortion can be caused by an anomaly or two.

[Robert Goren]

I am impressed. The only point I disagree with (modest disagreement, not vehement) is the 9/1 top end. I think that is track specific, as well as race (class/grade/distance) specific. Meaning there are situations (specific by track and race) in which defining the range as 5/1 to 9/1 is too restrictive. It is dismaying to toss out a horse that wins easily at 15/1.

I agree that the low end is important, but lumping all >X into one category can be misleading. Discovering that "this high but no higher" point for each situation is useful. Of course, that presupposes cleaning data and correcting for outliers--the smaller the sample, the more distortion can be caused by an anomaly or two. That is why I said the median instead of the mean. I would also look at for a mode if I had enough data. I look at win% * Median winning odds+1. If it close to or above 1.00 then I dig deeper. I have rejected a lot of ideas. far more than I have kept. There maybe niches that include 15/1 shots, but I have not found one yet. I have temporally stopped looking while I work on another project, but I firm believe this is the best method for "small bettors" to be profitable.

[traynor]

That is why I said the median instead of the mean. I would also look at for a mode if I had enough data. I look at win% * Median winning odds+1. If it close to or above 1.00 then I dig deeper. I have rejected a lot of ideas. far more than I have kept. There maybe niches that include 15/1 shots, but I have not found one yet. I have temporally stopped looking while I work on another project, but I firm believe this is the best method for "small bettors" to be profitable.

The median is almost as likely to be distorted by an occasional outlier as the simple mean. I think the mean of the interquartile range (the average of the center half of the values) is a much more reliable descriptor, and definitely a better predictor for betting models.