A Critique of "Natural Odds"

[maddog42]

I have seen some very good handicappers on this site say something like:
If your horse doesn't have, say a 12.5% chance of winning in an 8 horse field
then it is a mistake to bet him. This seems to be the essence of "natural odds". This doesn't seem to fit in with my "overlay" mentality that you can make money on these fringe horses. I know that it takes the patience of Job to bet these kinds of horses, but that is my niche. I am calling this style of handicapping "Extreme Overlay Handicapping" and i feel there are others doing this and I would like to hear from you. If you don't want to post on this thread, then PM me.

Thanks
PS I would also like to hear any and all criticism of this style. I am certainly a fan of most styles of Overlay handicapping and welcome disagreements. I do not want to hear such BS as "forget Overlays" as I have had that argument before.

[pandy]

I think the fringe horses, as you call them, I just call them longshots, are easier to show a profit on than horses in the low odds range. Of course, you have to be good at spotting longshots that are being under bet.

[cj]

My own personal records led me to eliminate exactly the horses you are talking about. I bet plenty of bombs, but only if I think they better than natural odds to win the race. It really is up to what works for each person though.

[Capper Al]

I have seen some very good handicappers on this site say something like:
If your horse doesn't have, say a 12.5% chance of winning in an 8 horse field
then it is a mistake to bet him. This seems to be the essence of "natural odds". This doesn't seem to fit in with my "overlay" mentality that you can make money on these fringe horses. I know that it takes the patience of Job to bet these kinds of horses, but that is my niche. I am calling this style of handicapping "Extreme Overlay Handicapping" and i feel there are others doing this and I would like to hear from you. If you don't want to post on this thread, then PM me.

Thanks
PS I would also like to hear any and all criticism of this style. I am certainly a fan of most styles of Overlay handicapping and welcome disagreements. I do not want to hear such BS as "forget Overlays" as I have had that argument before.

Amen. This is the only absolute truth in the game that handicapping beats Random( or natural) odds. You should be able to find a thread on the One and Only Absolute Truth if you search the forum. Dave Schwartz also sells The Basics of Winning ( video or book can't remember) which is built around the same philosophy. We came to this belief independently of each other. I will say if you dutch and are willing to drop down to support any of your top four picks, in fields eight or more in size, you should have more betting opportunities.

Good luck

[pandy]

My own personal records led me to eliminate exactly the horses you are talking about. I bet plenty of bombs, but only if I think they better than natural odds to win the race. It really is up to what works for each person though.

So you didn't really eliminate them.

People who spew out stats showing that horses in a high odds range show a big minus ROI are completely missing the point. Naturally if you bet every 40-1 shot or every favorite your ROI will be higher on the favorites, but I'd bet that there are a lot more handicappers out there that can show a profit betting only on longshots than there are handicappers who can show a profit betting on favorites. There is no leverage with short priced horses, too many of them are underlays.

[GameTheory]

I have seen some very good handicappers on this site say something like:
If your horse doesn't have, say a 12.5% chance of winning in an 8 horse field
then it is a mistake to bet him. This seems to be the essence of "natural odds". This doesn't seem to fit in with my "overlay" mentality that you can make money on these fringe horses. I know that it takes the patience of Job to bet these kinds of horses, but that is my niche. I am calling this style of handicapping "Extreme Overlay Handicapping" and i feel there are others doing this and I would like to hear from you. If you don't want to post on this thread, then PM me.

Thanks
PS I would also like to hear any and all criticism of this style. I am certainly a fan of most styles of Overlay handicapping and welcome disagreements. I do not want to hear such BS as "forget Overlays" as I have had that argument before.

Yes, as long as the overlay really is extreme (the price vs the probability) because once you get to probabilities lower than 14% or so the margin of error is so thin. So I'm not going to be betting on a 10% horse that pays 12-1 or even 15-1 or 20-1, but at 30-1 or 50-1, then ok. Most of those kinds of plays are purely statistical spot plays for me -- any 30-1 horse is going to look like crap on paper (or have no record at all) -- you are just betting that some percentage of these horses (or their connections) with a certain characteristic will wake up and win. Purely a numbers game without really handicapping the race at hand. For normal sort of handicapping I think the "natural odds" advice is not bad...

[maddog42]

Let me clarify: I am betting horses that are really my 4th or sometimes fifth choice in most rankings if their odds are over 12-1. Sometimes I demand 15-1.

[cj]

So you didn't really eliminate them.

People who spew out stats showing that horses in a high odds range show a big minus ROI are completely missing the point. Naturally if you bet every 40-1 shot or every favorite your ROI will be higher on the favorites, but I'd bet that there are a lot more handicappers out there that can show a profit betting only on longshots than there are handicappers who can show a profit betting on favorites. There is no leverage with short priced horses, too many of them are underlays.

Horses that I don't think have better than natural odds? Yes, I eliminate them. Maybe there are people out there good enough to think a horse has a 4% chance to win is an overlay at 25 or 30 to 1, but I'm not one of them. My records prove it.

Now, give me a horse I think is better than natural odds going off at 30 to 1 and my ROI goes through the roof. I'm still wrong plenty, but right often enough. That isn't the case with the other scenario.

[cj]

Let me clarify: I am betting horses that are really my 4th or sometimes fifth choice in most rankings if their odds are over 12-1. Sometimes I demand 15-1.

Choice is a lot different than percentage IMO. I can have fifth choices that I think have no chance, and fifth choices that I think are close to the favorite.

[pandy]

Horses that I don't think have better than natural odds? Yes, I eliminate them. Maybe there are people out there good enough to think a horse has a 4% chance to win is an overlay at 25 or 30 to 1, but I'm not one of them. My records prove it.

Now, give me a horse I think is better than natural odds going off at 30 to 1 and my ROI goes through the roof. I'm still wrong plenty, but right often enough. That isn't the case with the other scenario.

I see what you mean. The horses I bet that are 15 to 30-1 or higher are horses that I feel have a much better chance of winning than those odds reflect. I don't look at the board and say, oh, this horse may be a good 40-1 shot because it should be 30-1. I use my computer system and other methods I've developed to pick out the live longshot horses in advance, and then bet them if the odds are high enough.

[cj]

I see what you mean. The horses I bet that are 15 to 30-1 or higher are horses that I feel have a much better chance of winning than those odds reflect. I don't look at the board and say, oh, this horse may be a good 40-1 shot because it should be 30-1. I use my computer system and other methods I've developed to pick out the live longshot horses in advance, and then bet them if the odds are high enough.

I think we are on the same page. I'm not looking for horses that I think should be 50 to 1 but are going off at 90-1.

[maddog42]

My own personal records led me to eliminate exactly the horses you are talking about. I bet plenty of bombs, but only if I think they better than natural odds to win the race. It really is up to what works for each person though.

My records tell me that these horses don't win "natural odds" but somewhere in the 6-8% range. I am betting more of these horses in the 2 hole(underneath favorites) if there odds are above 20-1. This of course is a much higher hit rate and makes the long droughts tolerable.

[classhandicapper]

I think the big problem with line making is that we are all dealing with imperfect data and imperfect understanding. So we learn to not be too literal in our own odds lines and develop little rules that seem to improve the results.

I don't like to bet horses at less than 5-2 even though there are theoretically plenty of overlays at lower prices. Long term, my results indicate I'm better off not going below that. I do on rare occasions (I actually did it twice this weekend and both horses finished off the board :bang: , but I don't make a habit of it.

I also don't like to bet horses that are too far down in my ranking even when they seem to reasonably close to the top horse. I prefer to use horses like that to fill out exactas, triples etc... with just a little on top to keep my sanity. Those horse don't seem to win nearly as often for me as my figures suggest.

I think almost everyone develops little rules like that to help guide them.

[maddog42]

I think we are on the same page. I'm not looking for horses that I think should be 50 to 1 but are going off at 90-1.

Good point. Nobody can slice the Pie that thin and be accurate. I am talking
horses that I think have an 8% chance and are paying many times 20-1 or more. There are some cheaper races that I feel have 6 or 7 legitimate contenders. These overlays are often my 4th choice.

[maddog42]

Benter said in one of his papers/speeches that he combined public odds with his own odds line to create a more realistic oddsline. I think there may be a better more contrarian way to do it. Any thoughts.

[Grits]

Let me clarify: I am betting horses that are really my 4th or sometimes fifth choice in most rankings if their odds are over 12-1. Sometimes I demand 15-1.

Your fifth choice ran second in the Wood at 21-1. Your fifth choice ran second in the Blue Grass at 15-1. Meeting your rankings, I hope you had both underneath.

[jfdinneen]

maddog42,

When handicapping horse races, it is critical to focus on those markets in which there is at least one overlay. However, contrary to received wisdom, the professional sports trader does not just trade the specific overlays but instead trades one or more additional horses (possibly including an underlay) with the goal of spreading his risk while maximizing his long-term median income. An excellent worked example of this approach is detailed in Appendix V of Taking Chances (http://www.amazon.com/Taking-Chances-Probability-John-Haigh/dp/0198526636/ref=sr_1_9?s=books&ie=UTF8&qid=1347486802&sr=1-9) .

John

[maddog42]

maddog42,

When handicapping horse races, it is critical to focus on those markets in which there is at least one overlay. However, contrary to received wisdom, the professional sports trader does not just trade the specific overlays but instead trades one or more additional horses (possibly including an underlay) with the goal of spreading his risk while maximizing his long-term median income. An excellent worked example of this approach is detailed in Appendix V of Taking Chances (http://www.amazon.com/Taking-Chances-Probability-John-Haigh/dp/0198526636/ref=sr_1_9?s=books&ie=UTF8&qid=1347486802&sr=1-9) .

John

Thanks for the Book tip. Since I spend half my life making oddslines, I enjoy reading all sorts of varied opinions.

[maddog42]

maddog42,

When handicapping horse races, it is critical to focus on those markets in which there is at least one overlay. However, contrary to received wisdom, the professional sports trader does not just trade the specific overlays but instead trades one or more additional horses (possibly including an underlay) with the goal of spreading his risk while maximizing his long-term median income. An excellent worked example of this approach is detailed in Appendix V of Taking Chances (http://www.amazon.com/Taking-Chances-Probability-John-Haigh/dp/0198526636/ref=sr_1_9?s=books&ie=UTF8&qid=1347486802&sr=1-9) .

John
My idea of multiple overlays took a massive change in attitude when I read the 4 quarters of Horse Investing by Fierro. A couple of the False Overly points he made applied directly to me, and it is one of the best Books on Betting (not handicapping) I have ever read. It brought me into the black (just barely)at least for the year 2013. All of his betting line templates had either 3 or 4 contenders. They all totalled somewhere between 78 and 81%. I KNEW there were races that had at least 6 contenders and sometimes more. When I started testing my own betting lines, I could never get to these lofty percentages, but I did discover that at least on paper my top 4 contenders (in some categories) could win at a 74% clip and have some overlays. This also prompted me to make some 5 and 6 horse contender templates.

[Overlay]

I look at the full field, and I don't necessarily use "natural odds" as a cutoff, but I am wary of horses (regardless of what their fair odds may turn out to be) that rank in an "open-ended" bottom category on any of my rating elements, since, even if I assign them the value corresponding to that ranking, I still may be giving them too much credit by using that value in the calculation of their fair odds.

I also have more confidence in an overlaid horse if I can identify a misjudgment on the part of the public (and the reason for that misjudgment) in either making that horse an overlay, or making another horse in the race an underlay, which my compartmentalized approach generally allows me to do with accuracy.

[Robert Fischer]

It is a bit easier to think in terms of 'odds-ranges' rather than trying to be an expert from 1/9 - 900-1, but 'natural odds' are not a magical percentage per se.

Nothing wrong with including a horse like GREEN GRATTO in a pick sequence. You singled CONDO COMMANDO, you tossed READY FOR RYE, you thought the Wood was logical, and you were cold on CLEARLY NOW. You don't love GREEN GRATTO but there are a couple of reasons to use him in a variety of wagers if you are so inclined.

Benter said in one of his papers/speeches that he combined public odds with his own odds line to create a more realistic oddsline. I think there may be a better more contrarian way to do it. Any thoughts.

If you made the oddsline (or even if you determined that a single horse was an overlay) you have already made estimates as to the public's expected behavior.

Why upgrade THE BIG BEAST for being the 9/5 favorite that you have already estimated, and why downgrade a 50-1 GREEN GRATTO,- that you had already estimated to be about 50-1?

It's the unexpected behavior that threatens to shatter your preconceived value estimates.

[maddog42]

Before you guys run off and spend $172 on amazon for Fierro's book. Quinn's book The Best of Thoroughbred Handicapping-DRF press has a chapter on Fierro that summarizes his book very well and at a fraction of the cost. This is the SECOND expanded edition of this book. The first edition did not have a chapter on Fierro. So endeth the Public Service Address.

[Grits]

Before you guys run off and spend $172 on amazon for Fierro's book. Quinn's book The Best of Thoroughbred Handicapping-DRF press has a chapter on Fierro that summarizes his book very well and at a fraction of the cost. This is the SECOND expanded edition of this book. The first edition did not have a chapter on Fierro. So endeth the Public Service Address.

Md, thank you for this, you made me pull my copy out of the bookcase. ;)

[maddog42]

Your fifth choice ran second in the Wood at 21-1. Your fifth choice ran second in the Blue Grass at 15-1. Meeting your rankings, I hope you had both underneath.

Thank you for the kind words about the book. I am not sure I understand about the Blue Grass however. I didn't have any fifth choice. Were you speaking hypothetically from personal experience?

[Grits]

Hypothetically speaking. Too, I've put longshots underneath in exactas. With those, for whatever reason, I rarely cash. If I bet the LS to place, again, for whatever reason, I do better.

[maddog42]

Hypothetically speaking. Too, I've put longshots underneath in exactas. With those, for whatever reason, I rarely cash. If I bet the LS to place, again, for whatever reason, I do better.

I am not a very good bettor. I am a decent handicapper. Sound familiar Huh? I demand that at least one side of my exacta combo be an overlay, top or bottom.

[Appy]

I can certainly relate to a lot of what you've said in this thread maddog.
I'm better at handicapping than I am as a better, and am convinced learning to specialize is key to overall success.

The style of play I'm gravitating toward is not one of grinding it out, but more toward larger strikes. I've concluded there are 2 ways to realize big strikes. One is to risk enormous sums, the other is to play horses that can pay big.
I never risk enormous sums, but I do cash tickets for high priced winners on a relatively regular basis. Again, I credit specialization as key to knowing how and when to make the play.

Edit: Woops. Post too long. I erased last half. Sorry.

[badcompany]

My question is how do you determine whether something is an overlay when you don't even know the final odds you'll get?

[Clocker]

For anyone interested in Steve Fierro's work, there are a number of excerpts from his book here:

http://www.netcapper.com/TrackTractsArchive/TrackTractsArchive.htm

[Poindexter]

A couple of thoughts here.


I have no problem with 40-1 or 50-1 if I think they represent good value. I do think that once you get above 6-1 on your odds line you should start looking for double fair value. I have no problem betting a 4-1(my line) at 5-1, but I do not want to bet a 8-1(my line) at only 12 or 13-1, I would look for 16-1 plus. Also a lot of horses I may make 25-1 + I will not bet at any price and I will just line them out. I may make them 25-1 for defensive purposes, but would not be interested in them at 100-1, so of course those get weeded out. Also a lot of ways to get paid with a 40-1 shot(he doesn't just have to win-you have exactas, tris, supers......). If you are dealing with 20-1 or higher on your line you might want to require 2 1/2 times or triple your line.

Regardng blending in the public line, I have done it before(back when I was able to use a betting exchange), and probably would do it again if I was able to do so again(since betting exchanges enable you to get prices you cannot get at the track), but I do not think it is really needed. You just have to keep an eye on the money. Who is getting it and how much it scares you that they are getting the money. You can make tweeks from there. If I really respect the money on a bet down livee I might require 7 or 8-1 instead of 5-1 on my value horse and probably use the live horse in exotics with my value horse. If I think the money is the public just overbetting the obvious, I will probably make no adjustments to my play at all. Your value horse is going to be your value horse no matter who is live, you just want to make sure you are not ignoring some really pertinent betting information and may require a little or a lot of extra value based on what the board is telling you.

[Capper Al]

I am not a very good bettor. I am a decent handicapper. Sound familiar Huh? I demand that at least one side of my exacta combo be an overlay, top or bottom.

that's a good way to go. Add a place bet on your long shot if you are not dutching two shots.

[maddog42]

I didn't mean for this Thread to be a critique of Fierro, but I have praised him highly so here goes.

Fierro urges handicappers not to alter the templates even to adjust for scratches, jockey changes, alterations in the probable pace, Races taken off the turf, or other inconvenient factors that usually send handicappers scrambling....

This seems downright crazy. particularly the races taken off the turf. This is one of the main angles I use and it has become a specialty of mine. So many
handicappers are lost when this happens and i feel it positively gives me an edge.

My other problem that I have very mixed feelings about are Prohibited Favorites. "We have a prohibitive favorite in a race when: one of our betting line contenders is sent to the post at 3-2 and no other contender in the race is less than 5-1."
He claimed that "Whenever I set a line on horse at 3-2 or less, and the public agreed and bet the horse at less than 3-2, the other listed contenders were false overlays".
This is a very sticky situation for me, and about 75% of the time I would agree with this. He claimed to have lost an incredible 44 cents on the dollar in this type of race. I say 75% of the time because apparently there are some types I am terrible at handicapping.

[Stillriledup]

I thought about something recently that i never really thought all that much about over the years and that is the difference between horses you think are going to actually win and horses you feel are probably not going to win but you might bet anyway because they're a "good price".

Ask yourself how many wagers are you making on horses who you wouldn't bet if all the odds in the race were the same. Unless you're betting favorites, there's not too many runners you're going to play because you actually think they're the most likely horse to win. I think one of the downfalls of handicappers in the betting department is that they're betting on horses just looking for value and higher prices. I've been very guilty of that myself, passing over a favorite that i thought would "probably win" trying to get a price horse home. Betting on horses who you wouldnt bet if all the odds were equal is a very tricky thing to do and you have to be very careful when betting on horses you wouldn't otherwise bet.

Probably the best bets you could make are win bets on 2nd and 3rd betting choices when you really dislike the actual favorite. You're getting 2nd choice and 3rd choice odds on what you feel are the "most likely winner" and to me, that's the right place to start.

Sports bettors, for the most part, are risking 11 units to win 10, so you're not even doubling your money....in horse racing, bettors are not as "quick" to accept 1-1 odds, many bettors are looking for more price and what they deem as more value, but if pro sports bettors can make a living betting on 1-1 shots and 9-10 shots, i would imagine that its easier as a horse bettor to wager on horses who are between 4-5 and 3-1 and turn horse betting into "sports betting".

Its tough making a serious go of this game at the top of the oddsboard because you are betting on horses who are not most likely winners all that often, you're knowingly not betting on the best horse and when that's your approach, its a real fine line to walk.

[pandy]

That's why the best way to bet longshots is to have a method or methods that are good at picking longshot winners, then bet them if you feel that the horse is worth a bet at those odds....as opposed to just looking at the board and blindly betting on horses that are going off higher than you expected. There's a big difference. There are good longshot angles that work.

[Capper Al]

From my point of view, the best way to pick longshots is to have a system that picks favorites with a lot of proven elimination rules.

[Maximillion]

i would imagine that its easier as a horse bettor to wager on horses who are between 4-5 and 3-1 and turn horse betting into "sports betting".


if you can separate the good bets from the bad ones at those price ranges then more power to you.I know i cant.
if i handicap 50 races and cant come up with something over and above those price ranges ill pass all 50 every single time.

[Cholly]

There’s a conundrum involving the play of longshots: on the occasions of their success is the seed a positive, or a group of negatives. Did the longshot jump up with a surprisingly strong performance? Or did he only deliver another journeymen-like outing and benefit from being in the right place at the right time? Did several more talented horses concurrently fail to bring their race, whether it be due to bad starts, traffic problems, jockey blunders, track biases, unexpected pace developments, undiagnosed muscle soreness or more serious maladies, or any one of a hundred other things that can go wrong?

[maddog42]

Did the longshot jump up with a surprisingly strong performance? Or did he only deliver another journeymen-like outing and benefit from being in the right place at the right time? Did several more talented horses concurrently fail to bring their race,


Some of my research says that at least 2 of the top 4 contenders throw clunker races, Bounce, or what have you when the longshot 4th rank horse comes in.

[classhandicapper]

Probably the best bets you could make are win bets on 2nd and 3rd betting choices when you really dislike the actual favorite. You're getting 2nd choice and 3rd choice odds on what you feel are the "most likely winner" and to me, that's the right place to start.


That's my sweet spot.

[ReplayRandall]

Some of my research says that at least 2 of the top 4 contenders throw clunker races, Bounce, or what have you when the longshot 4th rank horse comes in.
Percentage-wise, how often does that occur for you?

[Cratos]

I have seen some very good handicappers on this site say something like:
If your horse doesn't have, say a 12.5% chance of winning in an 8 horse field
then it is a mistake to bet him. This seems to be the essence of "natural odds". This doesn't seem to fit in with my "overlay" mentality that you can make money on these fringe horses. I know that it takes the patience of Job to bet these kinds of horses, but that is my niche. I am calling this style of handicapping "Extreme Overlay Handicapping" and i feel there are others doing this and I would like to hear from you. If you don't want to post on this thread, then PM me.

Thanks
PS I would also like to hear any and all criticism of this style. I am certainly a fan of most styles of Overlay handicapping and welcome disagreements. I do not want to hear such BS as "forget Overlays" as I have had that argument before.
In the horseracing wagering pool the “natural odds” or random odds are somewhat skewed by racing rule of the “deadheat.”

According to that rule if more than one horse in the race finished the race at the same time and an individual winner cannot be determined; the race is declared between the finishers as a “deadheat.”

Therefore the “natural odds” of each horse will not be 1/n; they will be something less and in effect “conditional odds.”

[Saratoga_Mike]

Does anyone know what % of races end in a dead heat for win? Thx

[classhandicapper]

Probably the best bets you could make are win bets on 2nd and 3rd betting choices when you really dislike the actual favorite. You're getting 2nd choice and 3rd choice odds on what you feel are the "most likely winner" and to me, that's the right place to start.


That's my sweet spot.

[Magister Ludi]

Where N = number of entries, probabilities > 1/N are generally underbet and probabilities < 1/N are generally overbet. Odds entropy modeling can be useful for finding overlays.

[Cratos]

You don’t need to know the number of “deadheats”; it is an inherent part of the calculation if you understand statistics.

It is a probabilistic calculation; not a calculation of certainty which cannot happen.

[classhandicapper]

Did the longshot jump up with a surprisingly strong performance? Or did he only deliver another journeymen-like outing and benefit from being in the right place at the right time? Did several more talented horses concurrently fail to bring their race, whether it be due to bad starts, traffic problems, jockey blunders, track biases, unexpected pace developments, undiagnosed muscle soreness or more serious maladies, or any one of a hundred other things that can go wrong?

I make that distinction in my play.

If I bet a winning longer priced horse because I thought he was better than the favorite, I don't consider myself correct unless he wins fair and square. If the favorite was clearly best, but a trip I did not foresee was the reason he lost, after a brief celebration I try to figure out what I missed.

These kinds of things are part of the game. Sometimes they work in your favor and sometimes they don't. But I don't go into a race "hoping" the better horses run an unforeseen poor race or get some unforeseen bad trip. I like to bet against flawed horses I expect to run worse than people think for a specific reason.

[Saratoga_Mike]

You don’t need to know the number of “deadheats”; it is an inherent part of the calculation if you understand statistics.

It is a probabilistic calculation; not a calculation of certainty which cannot happen.

Do you have the %?

If 50% of races ended in dead heats vs. 0.2%, it matters.

[cj]

You don’t need to know the number of “deadheats”; it is an inherent part of the calculation if you understand statistics.

It is a probabilistic calculation; not a calculation of certainty which cannot happen.

The effect of dead heats in the natural odds calculation is minuscule. It increases each horses chances of winning very slightly, but hardly enough to be worthy of consideration in the way natural odds are being discussed in this thread.

[Saratoga_Mike]

The effect of dead heats in the natural odds calculation is minuscule. It increases each horses chances of winning very slightly, but hardly enough to be worthy of consideration in the way natural odds are being discussed in this thread.

I agree it's minuscule. But for the poster who loves precision to say it (i.e,, the occurrence rate) doesn't matter is befuddling.

[Cratos]

The effect of dead heats in the natural odds calculation is minuscule. It increases each horses chances of winning very slightly, but hardly enough to be worthy of consideration in the way natural odds are being discussed in this thread.
I understand what you are saying, but it is not that simplistic, a dead heat calculation is different as n gets large.

However this problem is better resolved with a Bayesian calculation which someone like TrifectaMike might better with than me.

[cj]

I agree it's minuscule. But for the poster who loves precision to say it (i.e,, the occurrence rate) doesn't matter is befuddling.


Who said that, I missed it? But as I said, in the context of this thread it won't change anything.

Here is some data from 2009-2014, six years of racing in North America:

Total races: 290319
Races 1 win: 289711
Races 2 win: 607
Races 3 win: 1

Dead heats for win occur in about .2% of races

[Saratoga_Mike]

Who said that, I missed it? But as I said, in the context of this thread it won't change anything.

Here is some data from 2009-2014, six years of racing in North America:

Total races: 290319
Races 1 win: 289711
Races 2 win: 607
Races 3 win: 1

Dead heats for win occur in about .2% of races

Just as I guessed - always nice to see someone with real numbers and not fantasyland. Thanks.

[LottaKash]

I understand what you are saying, but it is not that simplistic, a dead heat calculation is different as n gets large.

However this problem is better resolved with a Bayesian calculation which someone like TrifectaMike might better with than me.

I wonder how, please tell me, in the real world betting arena, how knowing and making that calculation, in application, could or would improve anyone's ROI in a meaningful and useful way...As in more or less $$$...

[cj]

I understand what you are saying, but it is not that simplistic, a dead heat calculation is different as n gets large.

However this problem is better resolved with a Bayesian calculation which someone like TrifectaMike might better with than me.

Of course the more horses the more likely a dead heat, but again, it is really not relevant to this discussion.

[Saratoga_Mike]

I wonder how, please tell me, in the real world betting arena, how knowing and making that calculation, in application, could or would improve anyone's ROI in a meaningful and useful way...As in more or less $$$...

It wouldn't, as CJ already stated, but it could make oneself feel smart on a horse-racing message board.

[Cratos]

Who said that, I missed it? But as I said, in the context of this thread it won't change anything.

Here is some data from 2009-2014, six years of racing in North America:

Total races: 290319
Races 1 win: 289711
Races 2 win: 607
Races 3 win: 1

Dead heats for win occur in about .2% of races

You are missing the point; the OP’s question was to address this statement of:” If your horse doesn't have, say a 12.5% chance of winning in an 8 horse field then it is a mistake to bet him”.

That really doesn’t anything to do with your statistics because of the Law of Large Numbers.

This is a simple random variable calculation and the probability of any one horse winning does not increase randomly because the random probability in the universal set is fixed a 1 and is equally distributed over n.

A deadheat increases n; therefore to answer the OP’s concern with a deadheat the answer is 1/8+1 or 1/9 and would reduce any one horse random chance of winning in an eight horse field by approximately 11%.

[Saratoga_Mike]

Of course the more horses the more likely a dead heat, but again, it is really not relevant to this discussion.

Pushing my luck, but if it's easy, could you please post those stats for deadheat for 2nd? Thanks.

[cj]

Pushing my luck, but if it's easy, could you please post those stats for deadheat for 2nd? Thanks.

I already closed it out, I'll get it tomorrow for you.

[Cratos]

I wonder how, please tell me, in the real world betting arena, how knowing and making that calculation, in application, could or would improve anyone's ROI in a meaningful and useful way...As in more or less $$$...

LottaKash,

When you model you never think in those terms because it is inherent tin the outcome.

[Saratoga_Mike]

You are missing the point; the OP’s question was to address this statement of:” If your horse doesn't have, say a 12.5% chance of winning in an 8 horse field then it is a mistake to bet him”.

That really doesn’t anything to do with your statistics because of the Law of Large Numbers.

This is a simple random variable calculation and the probability of any one horse winning does not increase randomly because the random probability in the universal set is fixed a 1 and is equally distributed over n.

A deadheat increases n; therefore to answer the OP’s concern with a deadheat the answer is 1/8+1 or 1/9 and would reduce any one horse random chance of winning in an eight horse field by approximately 11%.

But you're assuming equal distribution, which as CJ already demonstrated (0.2%), isn't reality. Another one of your good intellectual exercises. Kind of enjoyed stats in grad school.

[Saratoga_Mike]

I already closed it out, I'll get it tomorrow for you.

Appreciate it

[cj]

You are missing the point; the OP’s question was to address this statement of:” If your horse doesn't have, say a 12.5% chance of winning in an 8 horse field then it is a mistake to bet him”.

That really doesn’t anything to do with your statistics because of the Law of Large Numbers.

This is a simple random variable calculation and the probability of any one horse winning does not increase randomly because the random probability in the universal set is fixed a 1 and is equally distributed over n.

A deadheat increases n; therefore to answer the OP’s concern with a deadheat the answer is 1/8+1 or 1/9 and would reduce any one horse random chance of winning in an eight horse field by approximately 11%.

Dead heats add winners, they don't subtract them. Of course it changes the payouts, but factoring in dead heats can only increase the random chance of a horse winning, not decrease it.

[Saratoga_Mike]

He's treating the dead heat as unique outcome and not crediting the winning horses. It's a theoretical discussion, nothing more.

[cj]

He's treating the dead heat as unique outcome and not crediting the winning horses. It's a theoretical discussion, nothing more.

Pretty sure every time I've seen a dead heat with two or more official winners, they all got a picture taken and the bettors were paid. :)

[Saratoga_Mike]

Pretty sure every time I've seen a dead heat with two or more official winners, they all got a picture taken and the bettors were paid. :)

You've never bet at MIT Downs! :)

[Cratos]

Of course the more horses the more likely a dead heat, but again, it is really not relevant to this discussion.

That might be what you conclude, but statistically that is not true. Random statistics tells us as n gets large the probability of a deadheat gets small.

Also whether you believe it or not, it is very relevant to this discussion if you are speaking of “natural odds” because by definition that infers that you include all “natural outcomes”.

[cj]

That might be what you conclude, but statistically that is not true. Random statistics tells us as n gets large the probability of a deadheat gets small.

Also whether you believe it or not, it is very relevant to this discussion if you are speaking of “natural odds” because by definition that infers that you include all “natural outcomes”.

I won't argue that point because I haven't studied it, but it doesn't seem to make sense in real world. Seems to me the more horses that are in a race, the more likely two will run the exact same time. But it could be that it is less likely those horses do it and finish first. If the winning time is a given, hard to imagine that the chances of another horse (or more) matching that time isn't increased as the number of other horses increases.

But as far as dead heats decreasing a horse's natural odds, that makes no sense in any world. Dead heats add winners, they don't decrease them. Winning purses are decreased as are payoffs, but the win still counts a win. You get a 1 whether the horse wins alone or as part of a 10 horse dead heat.

[Cratos]

I won't argue that point because I haven't studied it, but it doesn't seem to make sense in real world. Seems to me the more horses that are in a race, the more likely two will run the exact same time. But it could be that it is less likely those horses do it and finish first. If the winning time is a given, hard to imagine that the chances of another horse (or more) matching that time isn't increased as the number of other horses increases.

But as far as dead heats decreasing a horse's natural odds, that makes no sense in any world. Dead heats add winners, they don't decrease them. Winning purses are decreased as are payoffs, but the win still counts a win. You get a 1 whether the horse wins alone or as part of a 10 horse dead heat.
Winning purses and possible payoffs comes from fixed amounts before the race and if there is more than one winner (e.g. a deadheat) then each winner’s connection and bettor will get less because of the fixed amounts.

The reason that deadheats decrease a horse's natural odds is because it (the horse) didn’t win the race, it tied for a win and that is why n in the denominator is increased by 1.

I suppose you could say that the bettors chance of winning is increased by deadheats.

[cj]

Winning purses and possible payoffs comes from fixed amounts before the race and if there is more than one winner (e.g. a deadheat) then each winner’s connection and bettor will get less because of the fixed amounts.

The reason that deadheats decrease a horse's natural odds is because it (the horse) didn’t win the race, it tied for a win and that is why n in the denominator is increased by 1.

I suppose you could say that the bettors chance of winning is increased by deadheats.

I know how it works, which is why I know your statement is in error. In horse racing, a tie for first counts as a win. Next time you see a maiden dead heat winner still eligible for a maiden race, let me know.

[Saratoga_Mike]

Cratos - you're changing the goal posts by introducing a purse distribution discussion.

[Cratos]

I know how it works, which is why I know your statement is in error. In horse racing, a tie for first counts as a win. Next time you see a maiden dead heat winner still eligible for a maiden race, let me know.
We differ in our statistical understanding and that is okay; hopefully someone got something from each point of view.

[cj]

We differ in our statistical understanding and that is okay; hopefully someone got something from each point of view.

The difference is in our understanding of horse racing. A dead heat win is a win in this sport. There is nothing to debate about that. If you choose not to include it that way statistically for your own purposes, that is up to you, but that doesn't make it correct according to the rules of horse racing.

It also counts as a win for wagering purposes, which is what this thread is about. That said, no more dead heat talk. It will be deleted.

[LottaKash]

LottaKash,

When you model you never think in those terms because it is inherent tin the outcome.

Cratos, most likely, I am not following you in all of this....

Kill me please, as I am just a lowly pen and paper handicapper, but I just don't believe that the calculation in debate has or at least in my case (my own way of handicapping) any bearing on the upcoming situation (race) at hand....

I would be more concerned with runups, rail changes, timing errors and poor charting of the races as more of a concern to "my" ROI than knowing how many DH's there were historically....

I just don't believe that horse racing is all that complicated to begin with, let alone "that" calculation....If I didn't have any success at this game at all, I wouldn't have asked....Still unclear with your response tho...

[Cratos]

Cratos, most likely, I am not following you in all of this....

Kill me please, as I am just a lowly pen and paper handicapper, but I just don't believe that the calculation in debate has or at least in my case (my own way of handicapping) any bearing on the upcoming situation (race) at hand....

I would be more concerned with runups, rail changes, timing errors and poor charting of the races as more of a concern to "my" ROI than knowing how many DH's there were historically....

I just don't believe that horse racing is all that complicated to begin with, let alone "that" calculation....If I didn't have any success at this game at all, I wouldn't have asked....Still unclear with your response tho...
LottaCash,

We all differ here on the forum with different handicapping and different betting.

I am not attempting to change you; I am just voicing my point of view which with this post will end on this subject.

As I have stated, I am a handicapper that "model" races and that is as far as I want to go with that admission.

You should take from this thread whatever you found to be useful and dismiss the rest including what I have written if you don't like it because it is your money that is at risk when you bet.

[LottaKash]

LottaCash,

We all differ here on the forum with different handicapping and different betting.

I am not attempting to change you;

You should take from this thread whatever you found to be useful and dismiss the rest including what I have written if you don't like it because it is your money that is at risk when you bet.

Cratos, with all due respect, but where did you get the "change you" from my posts, as well as the "if you don't like it"....

I "was" asking to learn something, is all....I learned nothing it seems, as all I believed I was asking is how one could use that as an improver in an odds or handicapping model is all, and ultimately how knowing that could affect my ROI, long term....

Sorry you took it thataway.....

So asking again anyway, how could knowing how many Dead Heats, possibly aid "anyone" in the determining of realistic odds-value for any entrant....It is just about this one thing (#DH's), with me...Not your way, but only this...

I am still clueless in that regard....Sorry for annoying you...

[Cratos]

Cratos, with all due respect, but where did you get the "change you" from my posts, as well as the "if you don't like it"....

I "was" asking to learn something, is all....I learned nothing it seems, as all I believed I was asking is how one could use that as an improver in an odds or handicapping model is all, and ultimately how knowing that could affect my ROI, long term....

Sorry you took it thataway.....

So asking again anyway, how could knowing how many Dead Heats, possibly aid "anyone" in the determining of realistic odds-value for any entrant....It is just about this one thing (#DH's), with me...Not your way, but only this...

I am still clueless in that regard....Sorry for annoying you...
LottaKash, if I offended you I apologized.

Too many posts on this forum is "shooting from the hip."

My point was and still is, believe what you feel comfortable with.

This is a tough game and losing is not fun.

[MJC922]

A couple of thoughts here.


I have no problem with 40-1 or 50-1 if I think they represent good value. I do think that once you get above 6-1 on your odds line you should start looking for double fair value. I have no problem betting a 4-1(my line) at 5-1, but I do not want to bet a 8-1(my line) at only 12 or 13-1, I would look for 16-1 plus. Also a lot of horses I may make 25-1 + I will not bet at any price and I will just line them out. I may make them 25-1 for defensive purposes, but would not be interested in them at 100-1, so of course those get weeded out. Also a lot of ways to get paid with a 40-1 shot(he doesn't just have to win-you have exactas, tris, supers......). If you are dealing with 20-1 or higher on your line you might want to require 2 1/2 times or triple your line.

Regardng blending in the public line, I have done it before(back when I was able to use a betting exchange), and probably would do it again if I was able to do so again(since betting exchanges enable you to get prices you cannot get at the track), but I do not think it is really needed. You just have to keep an eye on the money. Who is getting it and how much it scares you that they are getting the money. You can make tweeks from there. If I really respect the money on a bet down livee I might require 7 or 8-1 instead of 5-1 on my value horse and probably use the live horse in exotics with my value horse. If I think the money is the public just overbetting the obvious, I will probably make no adjustments to my play at all. Your value horse is going to be your value horse no matter who is live, you just want to make sure you are not ignoring some really pertinent betting information and may require a little or a lot of extra value based on what the board is telling you.

At least to me anyway what you say is very sensible. In any of the research I've done once you're able to get a positive ROI in a top rank the drop-off to losing more than the take occurs quickly. Beyond top two in any ranking I think is generally silly to get involved with. To my way of thinking the ranking however can be anything you feel is relevant in the race at hand, trainer patterns or whatever it doesn't have to be comprehensive. The point is, whatever the ranking is based upon you believe (better yet, you've verified) the top rank is a potent factor in that scenario.

[seanm]

ML

your generalised comments re entropy through the SP ranges didn't match a limited study i did on 90k lines of racing data from australia, WA 2009 to 2012 inclusive. A simple 2 way anlaysis - Sp bands rows, Entropy bands cols.

despite that, ML i find your posts excellent and thought provoking as usual.

'entropy' itself i still find a fascinating concept. a 3 way analysis including 'field size' may give some further insights.

sm

[Magister Ludi]

ML

your generalised comments re entropy through the SP ranges didn't match a limited study i did on 90k lines of racing data from australia, WA 2009 to 2012 inclusive. A simple 2 way anlaysis - Sp bands rows, Entropy bands cols.

despite that, ML i find your posts excellent and thought provoking as usual.

'entropy' itself i still find a fascinating concept. a 3 way analysis including 'field size' may give some further insights.

sm

sm,

Thank you for your kind words. Trifecta Mike presented a competitiveness index (ci) in one of his posts:

sum {1/[O(i) + 1]^2}/N

where O(i) = odds of the ith horse
N = number of entries

The larger the value, the more uncompetitive the race. Use odds ranges for the columns and ci ranges for the rows and see what you get.

My own odds entropy model is much different than Trifecta Mike's but his is much easier to calculate.

[DeltaLover]

sm,

Thank you for your kind words. Trifecta Mike presented a competitiveness index (ci) in one of his posts:

sum {1/[O(i) + 1]^2}/N

where O(i) = odds of the ith horse
N = number of entries

The larger the value, the more uncompetitive the race. Use odds ranges for the columns and ci ranges for the rows and see what you get.

My own odds entropy model is much different than Trifecta Mike's but his is much easier to calculate.


How you define competitiveness?

[Cratos]

sm,

Thank you for your kind words. Trifecta Mike presented a competitiveness index (ci) in one of his posts:

sum {1/[O(i) + 1]^2}/N

where O(i) = odds of the ith horse
N = number of entries

The larger the value, the more uncompetitive the race. Use odds ranges for the columns and ci ranges for the rows and see what you get.

My own odds entropy model is much different than Trifecta Mike's but his is much easier to calculate.

Hi ML,

I find it interesting that you have an entropy model.

Therefore from an overview and without going into detail what do you think drives this lack of predictability in a horse race?

[Magister Ludi]

How you define competitiveness?

The most competitive possible race with an N-horse field would be one in which each horse’s probability of winning is 1/N. The least competitive possible race would be a match race where one horse’s probability of winning is just under 1 and the other’s probability of winning is just over 0. Needless to say, the probabilities in most races will be more dispersed than the former race and more concentrated than the latter race.

[traynor]

The most competitive possible race with an N-horse field would be one in which each horse’s probability of winning is 1/N. The least competitive possible race would be a match race where one horse’s probability of winning is just under 1 and the other’s probability of winning is just over 0. Needless to say, the probabilities in most races will be more dispersed than the former race and more concentrated than the latter race.

I think that is the area in which a bit of skepticism (mostly healthy and not obnoxious) tends to creep in. The unstated assumption is that an accurate method is used that assigns each entry a specific probability of winning that specific race, in competition with the specific group of other entries in that same race. Simplistic number massaging does not produce such a value, as that value is (most probably) understood by most who bet on horse races.

[Magister Ludi]

Hi ML,

I find it interesting that you have an entropy model.

Therefore from an overview and without going into detail what do you think drives this lack of predictability in a horse race?

Longer fields cause a greater concentration and homogeneity of probabilities, thus making prediction more difficult.

[Cratos]

Longer fields cause a greater concentration and homogeneity of probabilities, thus making prediction more difficult.
A good answer and I agree,hopefully some on this forum whose ineptitude in the understanding of statistics should read your posts.

[Capper Al]

sm,

Thank you for your kind words. Trifecta Mike presented a competitiveness index (ci) in one of his posts:

sum {1/[O(i) + 1]^2}/N

where O(i) = odds of the ith horse
N = number of entries

The larger the value, the more uncompetitive the race. Use odds ranges for the columns and ci ranges for the rows and see what you get.

My own odds entropy model is much different than Trifecta Mike's but his is much easier to calculate.

The whole premise to the natural odds thing is that we know nothing expect that handicapping beats random as the only measurable truth. Using something as esoteric as a competive index undermines the absolute truth and most likely will fail as most other false concepts regarding making your own odds line.

[maddog42]

All this talk of "Entropy" reminded me of a question I've been meaning to ask. On another gambling website one of the members said that Entropy was the name that Benters old partner Alan Woods used on this forum. Looking up that name, he did claim to be Woods. Of course Woods is deceased now and Entropy no longer posts. Any Opinions?

[ReplayRandall]

Let's get to the bottom-line with the most clear and concise post from Trifecta Mike that sums it up:
Originally Posted 7-15-12 by Trifecta Mike

Time to get serious. Lots of words and very little substance so far.

1. A purely par based methodology can get the take down to approximately 7%
2. Introducing a few properly selected factors and using Maximum Likelihood Estimation can get you an approximate -3 to +3% (This is track dependent).
3. Incorporate the tote odds and you are there ...purely positive.

No further details coming.

Mike (Dr Beav)

[ReplayRandall]

Here's how to exploit data using Maximum Likelihood Estimation by Dr Paul Parsons, who is a quantitative analyst at Botsphere, a division of BetVictor, where he builds mathematical and statistical models for sports pricing and business operations:
http://www.sportstradingnetwork.com/exploiting-data/

[seanm]

Originally Posted by Magister Ludi
Longer fields cause a greater concentration and homogeneity of probabilities, thus making prediction more difficult.

Cratos : A good answer and I agree,hopefully some on this forum whose ineptitude in the understanding of statistics should read your posts.

Cratos I am puzzled, so you may help me here.

ML ends in saying "..... thus making prediction more difficult."

Does it apply to
1. the handicapper of the handicappers
2. the handicapper of the horses
3. both 1 & 2.

sm

ps i went back and cruched another 30k lines of data looking for the effects of entropy and i wish i could see what ml sees. on a more positive note, my thoughts are being provoked and pushed in another "entropy" direction and hopefully i can shake that "ineptness" tag sometime soon.

[Robert Fischer]

All this talk of "Entropy" reminded me of a question I've been meaning to ask. On another gambling website one of the members said that Entropy was the name that Benters old partner Alan Woods used on this forum. Looking up that name, he did claim to be Woods. Of course Woods is deceased now and Entropy no longer posts. Any Opinions?

I'm not sure if he was still actively posting by the time I've joined, but I've read quite a few older threads.
I was always under the impression that he was legit.
There are many 'hall of fame' players, authors and figures related to the game that have passed through this site.

[traynor]

Here's how to exploit data using Maximum Likelihood Estimation by Dr Paul Parsons, who is a quantitative analyst at Botsphere, a division of BetVictor, where he builds mathematical and statistical models for sports pricing and business operations:
http://www.sportstradingnetwork.com/exploiting-data/

"Maximum likelihood estimation is a versatile and powerful technique which can extract best-estimate parameter values from a dataset irrespective of the statistical distribution of your model. And best of all you can run it without too much fuss in a simple spreadsheet like Excel."

I think some statisticians may fail to grasp the simple fact that building useful models is more complex that crunching a lot of numbers. In horse racing, the more numbers crunched, the more unlikely it is that the results can be meaningfully applied as predictors of individual events.

Crunching large sets of numbers may provide solace to those who believe that doing so enables them to predict the outcomes of future groups of events. In sweeping generalities, perhaps. Applied to a finite number of events in subsets (as in, "races in which wagers are actually made by the modeler/statistician"), not so much.

[Magister Ludi]

I think some statisticians may fail to grasp the simple fact that building useful models is more complex that crunching a lot of numbers.

I wholeheartedly agree. Mining overlooked data, using coarsely granular and creative data analysis, and building models with unique tools can be much more robust and adds information to public odds.

[traynor]

I wholeheartedly agree. Mining overlooked data, using coarsely granular and creative data analysis, and building models with unique tools can be much more robust and adds information to public odds.

Definitely. "Appropriate layering" is possibly the most difficult hurdle in predictive modeling.

[ReplayRandall]

Definitely. "Appropriate layering" is possibly the most difficult hurdle in predictive modeling.
Yes it is.......But once that hurdle is cleared, the sky's the limit.. ;)

[traynor]

Yes it is.......But once that hurdle is cleared, the sky's the limit.. ;)
And that makes it all worthwhile.

[seanm]

Traynor : I think some statisticians may fail to grasp the simple fact that building useful models is more complex that crunching a lot of numbers.

Traynor, how true.

I am not a statistician by any stretch. I did Quantitative Methods 1 & 2 in the mid 80's as part of my degree, hardly a leg up for this sort of task.

I remember TM's "speed models", I followed his lead and used 100k lines of data and got 1 of 2 models to a 99.8% r2 value, the other circa 98%.

Other general purpose models I got r2's anywhere from 20's to 40's.

My excitement was quickly dulled by the poor predictiveness, strike rate & ROI.

Traynor, it took me the last 12 months, 12 hours a day 5 days a week to understand how complex this model building task really is. It's indeed a hard slog.

[maddog42]

I'm not sure if he was still actively posting by the time I've joined, but I've read quite a few older threads.
I was always under the impression that he was legit.
There are many 'hall of fame' players, authors and figures related to the game that have passed through this site.

I got the same impression. No real evidence but he seemed to fit the character.

[maddog42]

Traynor : I think some statisticians may fail to grasp the simple fact that building useful models is more complex that crunching a lot of numbers.

Traynor, how true.

I am not a statistician by any stretch. I did Quantitative Methods 1 & 2 in the mid 80's as part of my degree, hardly a leg up for this sort of task.

I remember TM's "speed models", I followed his lead and used 100k lines of data and got 1 of 2 models to a 99.8% r2 value, the other circa 98%.

Other general purpose models I got r2's anywhere from 20's to 40's.

My excitement was quickly dulled by the poor predictiveness, strike rate & ROI.

Traynor, it took me the last 12 months, 12 hours a day 5 days a week to understand how complex this model building task really is. It's indeed a hard slog.

I shop for books and sell them on Amazon. I came across a textbook called Multivariate Statistics. I buy books, sometimes very expensive books and then sell them. I was really working my way through that book and only understanding about half when the damn thing sold. Crap I had to ship it off.

[Stillriledup]

I'm not sure if he was still actively posting by the time I've joined, but I've read quite a few older threads.
I was always under the impression that he was legit.
There are many 'hall of fame' players, authors and figures related to the game that have passed through this site.
http://www.paceadvantage.com/forum/showthread.php?t=11543&page=1&pp=15