Testing Overlays

[pandy]

As some of you know, I do a lot of testing of handicapping angles and methods, both for my own betting and for my books and software that I sell. Lately, I've been using a new website, trackyourplays.com, which makes it easier for me run multiple tests because it automatically tracks results and ROI. It's not my website.

One of the things I'm working on is a new set of speed and pace ratings generated from averages. In my book, Speed And Class Handicapping, I used "best of last two on the same surface" to create the average speed figure. But now I've added an average of the best three figures in last five starts on the same surface, and also created two different type of compounded pace figures based on the same pacelines. I don't think that using one paceline is the best approach and my testing verifies this.

Another thing I'm testing, something that has always been a puzzle for handicappers is, do you play your top pick, or do you handicap contenders and then look for value? So say you handicap a race and come up with four horses ranked in order of preference. Regardless of what method you're using to handicap, I think you could make a case that you should never bet the favorite. So if your top pick is the favorite, you should either pass the race, or look over your other three ranked horses to see if there's any value there. But I'm not sure that you should ever bet the second choice, either. The first two choices win around 54% of the races, but the value is usually in the horses that win less often... it's sort of a paradox because you have to find the best value among the horses that are less likely to win the race.

[GBL]

"I think you could make a case that you should never bet the favorite."

I am surprised that someone would think this way.

Yes, it's true that bettors should consider the aspects of risk vs. reward, ROI and etc. before investing but implying that one should 'never' bet the favorite is plain wrong and ridiculous.

There're true favorites and there're false favorites. Your statement implies all favorites are bad.

One need to distinguish between the true favorite and the false favorite. If you think that horse is a true favorite and in your assessment he has, let's suppose 50% chances of winning the race, and consequently you assigned him the odds of even-money. Now, the horse goes off as 8/5 favorite. Should you still go against him? That 8/5 makes him the overlay for you. Why not bet him then? For all intents and purposes you should jump on him and, actually, bet him heavy if you have slightest faith in your handicapping system.

[DeltaLover]

As some of you know, I do a lot of testing of handicapping angles and methods, both for my own betting and for my books and software that I sell. Lately, I've been using a new website, trackyourplays.com, which makes it easier for me run multiple tests because it automatically tracks results and ROI. It's not my website.

One of the things I'm working on is a new set of speed and pace ratings generated from averages. In my book, Speed And Class Handicapping, I used "best of last two on the same surface" to create the average speed figure. But now I've added an average of the best three figures in last five starts on the same surface, and also created two different type of compounded pace figures based on the same pacelines. I don't think that using one paceline is the best approach and my testing verifies this.

Another thing I'm testing, something that has always been a puzzle for handicappers is, do you play your top pick, or do you handicap contenders and then look for value? So say you handicap a race and come up with four horses ranked in order of preference. Regardless of what method you're using to handicap, I think you could make a case that you should never bet the favorite. So if your top pick is the favorite, you should either pass the race, or look over your other three ranked horses to see if there's any value there. But I'm not sure that you should ever bet the second choice, either. The first two choices win around 54% of the races, but the value is usually in the horses that win less often... it's sort of a paradox because you have to find the best value among the horses that are less likely to win the race.

Getting the average of previous figures is not useful for handicapping and betting reasons. In contrary, in most of the cases such an approach will suppress your chances of catching career best performances, which is where most of the betting interest lies in.

Also using constraints like “same surface” or “same distance” make your process more conservative focusing to an easier to attack problem that is also solved pretty well by the betting crowd. Focusing on more challenging problems, like surface, distance or circuit switches is way more attractive as this is where most of the crowd miscalculations occur.

As far as the second part of your posting, I think you are asking the wrong question. Your best betting selection should not be decided by its ranking (meaning whether is your top selection of the fourth one) but from the possibility of the crowd to misestimate its angles regardless of how its final price will be shaped in the market. The crowd is really good in ABC handicapping so trying to beat it by forming a superior ranking of horses (meaning where your top picks will win more often than the pubic favorite for example) is simply a dream pipe that is impossible to accomplish. The best you can do is to detect the way the public is forming the pools (something that changes constantly through time and also differs from track to track) and try to “correct” any potential mistake before the crowd realize it and encapsulate it to its betting.

[jay68802]

I set a minimum odds I am willing to take on the win bet, 5-2. If I do not get that, I look at the exacta with the second and third choice. Minimum odds on the exacta is 8-1.

I, like you, use more than 1 race to evaluate a horse. My latest experiment is using the track variant in the process also. I simplified the variant to five categories. You would be surprised at how often a horse shows good performances on tracks with the same variant. Then today gets a surface that does not have the same variant and throws in a clunker. The reason behind doing this was because of another thread here, Late Odds Changes. I was tracking the late money and noticed that a lot of times the late money was not showing in the first few races of the day. My thinking is that people are trying to get a feel for how fast or slow the track is and adjusting their picks accordingly. GL.

[DeltaLover]

I set a minimum odds I am willing to take on the win bet, 5-2. If I do not get that, I look at the exacta with the second and third choice. Minimum odds on the exacta is 8-1.



As discussed in depth in the Odds changes thread, in todays game it is meaningless to rely on pre-race minimum odds; is simply does not work like this!

[jay68802]

As discussed in depth in the Odds changes thread, in todays game it is meaningless to rely on pre-race minimum odds; is simply does not work like this!

Then what exactly tells you when to bet and when not to bet? To me, this is simple. You bet when the odds on the horse are above what your win % is.

[headhawg]

:popcorn:

[HalvOnHorseracing]

A person I worked with in the 1980's told me he had never been to a racetrack. He lives in Riverside and since I was at Santa Anita last week I asked him if he wanted to go and he said he'd come on Friday.

One of the things I was told very early in my racing career was that it is good betting to turn 4/5 shots into 7/2 shots. It can be done by using the horse in a combination bet.

Anyway there was an odds on horse in the 5th race on Friday. I told my friend I was just going to bet an exacta with that horse and the 1 horse. He bet $20 on the same exacta. It paid $9.80 for $2. In other words, we turned a 4/5 shot into almost a 4-1 shot.

Not the payoff of the century, but it makes the point that you can get a decent return on your investment, even with short-priced favorites, if you focus a combination bet properly.

My friend collected $98 for his $20 investment. Not bad for your first bet.

[pandy]

"I think you could make a case that you should never bet the favorite."

I am surprised that someone would think this way.

Yes, it's true that bettors should consider the aspects of risk vs. reward, ROI and etc. before investing but implying that one should 'never' bet the favorite is plain wrong and ridiculous.

There're true favorites and there're false favorites. Your statement implies all favorites are bad.

One need to distinguish between the true favorite and the false favorite. If you think that horse is a true favorite and in your assessment he has, let's suppose 50% chances of winning the race, and consequently you assigned him the odds of even-money. Now, the horse goes off as 8/5 favorite. Should you still go against him? That 8/5 makes him the overlay for you. Why not bet him then? For all intents and purposes you should jump on him and, actually, bet him heavy if you have slightest faith in your handicapping system.

I get what you're saying, but that's a theory. Yes, in some instances the favorite may actually appear to be a good bet, but the problem is, no matter how good you are at handicapping, there isn't enough leverage there to show a long term profit on favorites. If the takeout was in single digits, maybe, but with the takeout 15% or higher, there just isn't enough room for error, and the rebates are too low on win bets to make up the difference.

Favorites are a false temptation. You hit a few overlaid favorites and you think you can bet them, but in the long run, they're a losing proposition. Of course, sometimes you have to use the favorite in exotics, but that's different. As win bets, they are fools gold. Theory is one thing, real life results is another.

[pandy]

Getting the average of previous figures is not useful for handicapping and betting reasons. In contrary, in most of the cases such an approach will suppress your chances of catching career best performances, which is where most of the betting interest lies in. .

I'm not saying that you can't use a single paceline, but I am saying that my testing has shown that averages are better for ROI over the long run. One thing you have to be very careful about is using the last line. For instance, just in the last two weeks alone, I've seen quite a few horses win off or horrible lines, but they were coming off a layoff of a few months and they ran back to their best lines from several races back. So whatever was bothering them, the trainer fixed it. They all paid double digit odds, too.

[pandy]

A person I worked with in the 1980's told me he had never been to a racetrack. He lives in Riverside and since I was at Santa Anita last week I asked him if he wanted to go and he said he'd come on Friday.

One of the things I was told very early in my racing career was that it is good betting to turn 4/5 shots into 7/2 shots. It can be done by using the horse in a combination bet.

Anyway there was an odds on horse in the 5th race on Friday. I told my friend I was just going to bet an exacta with that horse and the 1 horse. He bet $20 on the same exacta. It paid $9.80 for $2. In other words, we turned a 4/5 shot into almost a 4-1 shot.

Not the payoff of the century, but it makes the point that you can get a decent return on your investment, even with short-priced favorites, if you focus a combination bet properly.

My friend collected $98 for his $20 investment. Not bad for your first bet.


I agree, I use favorites in exotics, but they're bad win bets over the course of time.

[pandy]

I set a minimum odds I am willing to take on the win bet, 5-2. If I do not get that, I look at the exacta with the second and third choice. Minimum odds on the exacta is 8-1.

I, like you, use more than 1 race to evaluate a horse. My latest experiment is using the track variant in the process also. I simplified the variant to five categories. You would be surprised at how often a horse shows good performances on tracks with the same variant. Then today gets a surface that does not have the same variant and throws in a clunker. The reason behind doing this was because of another thread here, Late Odds Changes. I was tracking the late money and noticed that a lot of times the late money was not showing in the first few races of the day. My thinking is that people are trying to get a feel for how fast or slow the track is and adjusting their picks accordingly. GL.


That's a good point. There was a race at Aqueduct Saturday, Go Get The Money won his debut going 7f over the Aqueduct track in 1:26.2 and earned a big fig, 94 bris, and Timeform also had him as the clear top figure. Gleason won his debut at 7f at Tampa and went 1:23.3, almost three seconds faster!...but his figure was 85. Obviously, different track, much different variant. But I'm very suspicious of high figures earned in slow times, as I've mentioned before. The public made Gleason the 4/5 favorite, despite the fact that he won his debut for a much lower purse, and Go Get The Money was 8-5. Both horses lost but Gleason ran a fine second, Go Get The Munny was last and completely outrun. I would never bet a horse at 8-5 odds that ran once and ran slow. 8-5 is for fast horses.

[green80]

"I think you could make a case that you should never bet the favorite."

I am surprised that someone would think this way.

Yes, it's true that bettors should consider the aspects of risk vs. reward, ROI and etc. before investing but implying that one should 'never' bet the favorite is plain wrong and ridiculous.

There're true favorites and there're false favorites. Your statement implies all favorites are bad.

One need to distinguish between the true favorite and the false favorite. If you think that horse is a true favorite and in your assessment he has, let's suppose 50% chances of winning the race, and consequently you assigned him the odds of even-money. Now, the horse goes off as 8/5 favorite. Should you still go against him? That 8/5 makes him the overlay for you. Why not bet him then? For all intents and purposes you should jump on him and, actually, bet him heavy if you have slightest faith in your handicapping system.

A horse you figure at even money and going off at 8/5 is just as good of overlay as the horse you figure at 10/1 and is going off at 16/1. If your system is correct you need to be all over him.

[pandy]

A horse you figure at even money and going off at 8/5 is just as good of overlay as the horse you figure at 10/1 and is going off at 16/1. If your system is correct you need to be all over him.

Again, that's a theory, and I get the theory, but it doesn't work in real life. Many years ago I said the same thing in some of my columns, but I was wrong. That's why testing is better than theory. A great handicapper, someone who is great at picking winners, is still not going to be able to show a profit on win bets on 8/5 shots, or favorites in general. If you can get to something like a 5% loss betting favorites, you're a master handicapper. Everytime I've had a good year and actually went back and checked the results, the favorites I bet only provided a mental stability that cut down on a series of losing bets, but they were not profitable. All of my profit came from longshot winners that I keyed in winning exotics. Most winning players get the bulk of their profit from several nice hits each year. 8/5 shots don't provide that, and you simply can't overcome the takeout with favorites.

There are also a lot more overlays in the higher odds range because it requires a really sharp handicapper to see that a 15-1 shot should actually be 6-1. But with favorites, everyone can see why the horse is favored.

Now, I agree, that it's good to have the ability to spot bad favorites and bet against them. But that doesn't mean that it's smart to bet on the horses that appear to be legit favorites. Bet against the bad favorites, pass the races where the favorite appears legit.

The best and fastest horse in the race still loses more than it wins, so you simply can't take short prices.

[pandy]

By the way, here's a simple test that anyone can do. After you handicap, narrow the race down to three horses, or four if the field is more than 10 horses. Now keep results of how you would do if you bet the shortest odds of those three horses, and bet the longhot odds of those three horses. If you are a good handicapper, the bet on the horse that goes off the longest odds, of the top three, should produce a higher ROI than the bet on the lowest odds horse.

[CincyHorseplayer]

A person I worked with in the 1980's told me he had never been to a racetrack. He lives in Riverside and since I was at Santa Anita last week I asked him if he wanted to go and he said he'd come on Friday.

One of the things I was told very early in my racing career was that it is good betting to turn 4/5 shots into 7/2 shots. It can be done by using the horse in a combination bet.

Anyway there was an odds on horse in the 5th race on Friday. I told my friend I was just going to bet an exacta with that horse and the 1 horse. He bet $20 on the same exacta. It paid $9.80 for $2. In other words, we turned a 4/5 shot into almost a 4-1 shot.

Not the payoff of the century, but it makes the point that you can get a decent return on your investment, even with short-priced favorites, if you focus a combination bet properly.

My friend collected $98 for his $20 investment. Not bad for your first bet.

Growing up at River Downs I learned this was often the only way to make money! Or pass practically an entire day of races. I still play plenty of exactas where a single combination might be $20 and another $20 spread out to get a 7-1 return. I've gone back to more of this than win betting simply because better returns take off some pressure of having to be the Einstein of handicappers.

[CincyHorseplayer]

I get what you're saying, but that's a theory. Yes, in some instances the favorite may actually appear to be a good bet, but the problem is, no matter how good you are at handicapping, there isn't enough leverage there to show a long term profit on favorites. If the takeout was in single digits, maybe, but with the takeout 15% or higher, there just isn't enough room for error, and the rebates are too low on win bets to make up the difference.

Favorites are a false temptation. You hit a few overlaid favorites and you think you can bet them, but in the long run, they're a losing proposition. Of course, sometimes you have to use the favorite in exotics, but that's different. As win bets, they are fools gold. Theory is one thing, real life results is another.

Yep Pandy my last 3 years AB contenders less than 2-1.

2016-(-14%)

2017-(-7%)

2018(-11%)

The $6-8 range shows a modest profit. Then things jump up from there. Choosing the right odds range is simply a good foundation for nearly all bets IMO.

[pandy]

Yep Pandy my last 3 years AB contenders less than 2-1.

2016-(-14%)

2017-(-7%)

2018(-11%)

The $6-8 range shows a modest profit. Then things jump up from there. Choosing the right odds range is simply a good foundation for nearly all bets IMO.


5-2 seems to be a good cut off for someone who wants to keep their win percentage respectable. Certainly less than 2-1 is brutal in terms of ROI. You know, when you first meet and get a chance to talk to a professional horseplayer, it's usually not the conversation that you expect. They tend to talk more about betting than handicapping. And one of the first things they talk about is the take out. The professional bettors aren't always obsessed with who's going to win, but they're experts when it comes to the math: the size of the pools, the takeout on each type of wager, carryovers, the wagers at each track that tend to generate overlays, the odds range that they can bet to show a profit. Sometimes, that's why they can win, even though they're not necessarily a bettor handicapper than other people who lose. They understand what they're up against and they have the mathematical know how to maximize their chances by making the right bets.

[GBL]

I get what you're saying, but that's a theory. Yes, in some instances the favorite may actually appear to be a good bet, but the problem is, no matter how good you are at handicapping, there isn't enough leverage there to show a long term profit on favorites. If the takeout was in single digits, maybe, but with the takeout 15% or higher, there just isn't enough room for error, and the rebates are too low on win bets to make up the difference.

Favorites are a false temptation. You hit a few overlaid favorites and you think you can bet them, but in the long run, they're a losing proposition. Of course, sometimes you have to use the favorite in exotics, but that's different. As win bets, they are fools gold. Theory is one thing, real life results is another.

The answer was given based on the subject line "Testing Overlays". And overlays could include anyone including the favorites.

Now, that you've now put it in different light, it's impractical to think you can break the bank simply by playing win/place bets in longer terms. That's just not going to happen. Odds are stacked against you anyways. For good longer term ROI, exotics are more suitable to achieve the objectives. Secondly, if you play 20 races a day everyday, every system WILL fail. There just can't be any fool-proof system if you play too many races.

One of the keys to have an acceptable ROI and success rate in longer term is to limit the number of plays to minimum and only when the reward is acceptable against the risk. It's simply Risk vs. Reward strategy and limiting to your plays to when you feel the right opportunity arises. When you restrict your plays to only 2-3 races a day instead of 20 or 30, you're in a better position to play smarter and avoid unnecessary risks and losses.

Of course, that's also subject to one's own financial objectives and longer term goals.

[Dave Schwartz]

Now, that you've now put it in different light, it's impractical to think you can break the bank simply by playing win/place bets in longer terms. That's just not going to happen. Odds are stacked against you anyways. For good longer term ROI, exotics are more suitable to achieve the objectives. Secondly, if you play 20 races a day everyday, every system WILL fail. There just can't be any fool-proof system if you play too many races.

Is that an immutable law?

It clearly can be done by some people. It's being done by Whales every day.

[steveb]

The answer was given based on the subject line "Testing Overlays". And overlays could include anyone including the favorites.

Now, that you've now put it in different light, it's impractical to think you can break the bank simply by playing win/place bets in longer terms. That's just not going to happen. Odds are stacked against you anyways. For good longer term ROI, exotics are more suitable to achieve the objectives. Secondly, if you play 20 races a day everyday, every system WILL fail. There just can't be any fool-proof system if you play too many races.

One of the keys to have an acceptable ROI and success rate in longer term is to limit the number of plays to minimum and only when the reward is acceptable against the risk. It's simply Risk vs. Reward strategy and limiting to your plays to when you feel the right opportunity arises. When you restrict your plays to only 2-3 races a day instead of 20 or 30, you're in a better position to play smarter and avoid unnecessary risks and losses.

Of course, that's also subject to one's own financial objectives and longer term goals.

i thought your first post was very good.
this one though i dont agree with.
i reckon churning is the way to go, and any pools are fine, if you can find bets above expectation.

i reckon it's much easier to win money from many small bets(i guess small is relative)even if your roi is far less, than it is from being selective
you can't spend roi.

i would be confident that ALL the big guys churn.
every race....every pool......where expectation is exceeded

and if you are just a hobby punter, it would not be much fun waiting.... waiting....waiting for your next play.
that's one boring hobby.

[GBL]

i thought your first post was very good.
this one though i dont agree with.
i reckon churning is the way to go, and any pools are fine, if you can find bets above expectation.

i reckon it's much easier to win money from many small bets(i guess small is relative)even if your roi is far less, than it is from being selective
you can't spend roi.

i would be confident that ALL the big guys churn.
every race....every pool......where expectation is exceeded

and if you are just a hobby punter, it would not be much fun waiting.... waiting....waiting for your next play.
that's one boring hobby.

"....if you can find bets above expectation."

That's the bottom line, I suppose.

Whenever you come across a race which offers the proposition you're seeking, you make a go. How often these opportunities comes up I think depends on how you define it.

The reason that WP-only isn't the way to go for long term profitable ROI (subjective, again), especially when you're not willing to consider favorites' overlay as an overlay (as the originator of the thread implied), is not practical given the fact that some 60% of the races are won by top three favorites. practically killing the chances of any long term positive ROI. A 10/1 shot finishing second behind 6/5 favorite, generally speaking, will still probably pay around 2/1 or 5/2 for place. But then you have to come up with that 10/1 to begin with. Much tougher than people generally think.

The better way in such situations is to play longer price horses in "exacta-as-a-place" meaning rather than play place, play exacta with the favorite on top with longer priced as under. And, play straight win on the longer priced horse.

If betting favorite is a losing proposition (which actually is true) betting longshots in every race is also a losing proposition. There got to be a good mix and balance of both in one's betting strategy. I would think that hit ratio of around 40% with mixed odds could yield good ROI on longer term. Lots of subjectivity.

[GBL]

"....if you can find bets above expectation."

That's the bottom line, I suppose.

Whenever you come across a race which offers the proposition you're seeking, you make a go. How often these opportunities comes up I think depends on how you define it.

The reason that WP-only isn't the way to go for long term profitable ROI (subjective, again), especially when you're not willing to consider favorites' overlay as an overlay (as the originator of the thread implied), is not practical given the fact that some 60% of the races are won by top three favorites. practically killing the chances of any long term positive ROI. A 10/1 shot finishing second behind 6/5 favorite, generally speaking, will still probably pay around 2/1 or 5/2 for place. But then you have to come up with that 10/1 to begin with. Much tougher than people generally think.

The better way in such situations is to play longer price horses in "exacta-as-a-place" meaning rather than play place, play exacta with the favorite on top with longer priced as under. And, play straight win on the longer priced horse.

If betting favorite is a losing proposition (which actually is true) betting longshots in every race is also a losing proposition. There got to be a good mix and balance of both in one's betting strategy. I would think that hit ratio of around 40% with mixed odds could yield good ROI on longer term. Lots of subjectivity.

P.S. to the above in light of overlays: Just like overlaid favorites, not every high priced-overlay is a good, bet-able overlay. The 15/1 in your assessment system going off as 25/1 doesn't necessarily makes it a good bet.

[GBL]

"....if you can find bets above expectation."

That's the bottom line, I suppose.

Whenever you come across a race which offers the proposition you're seeking, you make a go. How often these opportunities comes up I think depends on how you define it.

The reason that WP-only isn't the way to go for long term profitable ROI (subjective, again), especially when you're not willing to consider favorites' overlay as an overlay (as the originator of the thread implied), is not practical given the fact that some 60% of the races are won by top three favorites. practically killing the chances of any long term positive ROI. A 10/1 shot finishing second behind 6/5 favorite, generally speaking, will still probably pay around 2/1 or 5/2 for place. But then you have to come up with that 10/1 to begin with. Much tougher than people generally think.

The better way in such situations is to play longer price horses in "exacta-as-a-place" meaning rather than play place, play exacta with the favorite on top with longer priced as under. And, play straight win on the longer priced horse.

If betting favorite is a losing proposition (which actually is true) betting longshots in every race is also a losing proposition. There got to be a good mix and balance of both in one's betting strategy. I would think that hit ratio of around 40% with mixed odds could yield good ROI on longer term. Lots of subjectivity.


Since every claim should be backed up by factual data, I thought it became necessary to publish some relevant statistics to highlight the difficulties and challenges one might face if he's hoping to make WP bets the basis of long term profitability especially when you're not willing to entertain favorites.

Here's some data to ponder from Dan Illman blog from the poster who maintains immaculate turf workbook for Gulfstream Park every year.

This was posted a day earlier (Monday 3/19/2018).

"Lately the Gulfstream Park turf courses are looking rather threadbare. Last week (W-T-F-S-Su) they ran 33 turf races. That is a lot of turf racing. Most were utterly forgettable. 18 races won by favorites (54%). One streak saw 6 straight chalks win. Only one winning horse went off more than 5.90-1 during the WHOLE WEEK. Field size was 8.4 last week. The course generally was not quick. Average winning time v. track record was about plus 4 seconds."

[pandy]

I agree with GBL that exotics are better for producing a positive ROI. But it's difficult to generate a positive exotic bet ROI with a key horse that has a minus ROI expectation. In other words, you still need to find that overlay. And, again, your chance of long-term success goes up if the key horse is not the favorite.

As for being ultra-selective and not making many bets, I don't agree with that. I think some players can make a lot of bets and still win. I read Mike Maloney's book and he seems to make a lot of wagers, and I know other players who are pros who make a lot of bets. One professional player who is currently active bets several pick 4's and pick 5's every day and has a lot of action, and wins.

There were three main points I wanted to stress from this thread, A). Favorites are generally bad bets in the long run. B). The horses that have the best long-term potential ROI are often not your top pick, or are not the horse that appears to have the best chance of winning. Your chances of winning actually go up if you only key the horses that have a lower win expectation. This is especially true in exactas and other vertical wagers where you can key a horse first or second. C). Testing, or actually betting and record keeping, is more reliable than "theory".

I remember many years ago reading that eating eggs can raise your cholesterol. But then I read a bodybuilding book by Vince Gironda, a natural professional bodybuilder who owned a famous gym in Hollywood. Gironda trained bodybuilders and movie stars who had to buff up quickly. He wrote a book called Unleashing The Wild Physique which was published in 1984. In that book, he said that the medical reports about eggs were false. He explained that for many years he had been having his pupils eat a dozen or more eggs a day as part of their training diet and their cholesterol didn't go up. Eventually, the medical profession started to test people and found the same thing, although eggs contain cholesterol, they don't increase your cholesterol. My point being, testing and real life results trump "theory".

[pandy]

By the way, thanks everyone for contributing to this thread, I appreciate it. GBLL mentioned that he didn't think you could win betting only longshots and that you need a mix. A mix is good because it helps cut down on run outs, but my testing has always shown that you can win, even on win bets, just by betting horses in the 6-1 and higher range. I truly believe that there are more overlays in the higher odds range because only the better handicappers are able to sniff out that live longshots.

One of the main reasons why most handicappers are just average, or below average, is that they are masters-of-the-obvious.

Years ago there was a website, I've mentioned it before, but unfortunately, I can't remember the name, but this was a service that gave out 20 to 20 computer generated longshot picks (on the morning line) every day. The picks were not just one angle, their system generated longshot picks based on different criteria that it showed could produce a positive ROI.

The cut off was either 6-1 or 8-1 on the ML, I can't remember exactly, but I think it may have been 8-1. Many of the horses went off much higher. I subscribed to the site out of curiosity. They offered a guarantee, which was, if they didn't show a profit that month, the next month was free.

Anyway, I followed this site for about a year and they showed a profit ten out of 12 months and showed an overall profit even though they gave out about 160 horses a week and almost every horse went off as a longshot. The win percentage was only around 12%! They also had stats showing their results over a three year period and they showed a profit all three years.

They took the site down and were thinking of selling their system as software but I never heard from them again.

[CincyHorseplayer]

Just wanted to add one last thing about decision making. When I first started to track contender selection ie 1st choice 2nd choice 3rd choice etc I noticed the drop off after the first 2. Top 2 would hit in the 25-30% range then 3rd choice would be 12-15% and 4th choice 8-10% etc. And I realized that many 3rd/4th choices and even 2nd choices weren't worth betting anyway so I started tracking them simply by main contenders and marginal contenders. Similar results over the last 1,000 races. Main contenders represent 21% of the entries and win 55% of the races. Marginal contenders represent 29% of the entries and win 31% of the races. It was what I had suspected all along. Marginal contenders do not win enough to develop a betting approach with them. And the mutuels are not any higher than the 9.07 on main contenders. And with the precipitous dropoff in hit rate evidenced by the 12-15% and 8-10% the average mutuel would have to be fairly astronomical to justify betting these horses. This reality has led to me to a conclusion I wholly believe=that value handicapping can lose itself quickly and become even a futile quest for value if not grounded.

That said with main contenders representing 21% of entries there are plenty of opportunities to bet value. I make practically no distinction between my top choices in a race. On handicapping principle I can split hairs if necessary but when win betting I simply bet the horse not taking action. If both aren't getting play I bet both. End of strategy. From 2001 to 2012 my average mutuel stayed with .50 of 9.40. Since then it has hovered within .50 of 12.23. All I did was restrict my win betting criteria to the above. Nearly 4 months into the new horse betting year my average mutuel is happily over $15, but waves of favorites the last few weeks are bringing that back into line.

None of that has anything to do with exotics. Lower mutuel horses always weave their way through payoffs and can still produce the desired return on bets. But I find the above to be a solid foundation for all plays to have a solid expectation. Or you don't bet. Period!

[CincyHorseplayer]

I agree with GBL that exotics are better for producing a positive ROI. But it's difficult to generate a positive exotic bet ROI with a key horse that has a minus ROI expectation. In other words, you still need to find that overlay. And, again, your chance of long-term success goes up if the key horse is not the favorite.

As for being ultra-selective and not making many bets, I don't agree with that. I think some players can make a lot of bets and still win. I read Mike Maloney's book and he seems to make a lot of wagers, and I know other players who are pros who make a lot of bets. One professional player who is currently active bets several pick 4's and pick 5's every day and has a lot of action, and wins.

There were three main points I wanted to stress from this thread, A). Favorites are generally bad bets in the long run. B). The horses that have the best long-term potential ROI are often not your top pick, or are not the horse that appears to have the best chance of winning. Your chances of winning actually go up if you only key the horses that have a lower win expectation. This is especially true in exactas and other vertical wagers where you can key a horse first or second. C). Testing, or actually betting and record keeping, is more reliable than "theory".

I remember many years ago reading that eating eggs can raise your cholesterol. But then I read a bodybuilding book by Vince Gironda, a natural professional bodybuilder who owned a famous gym in Hollywood. Gironda trained bodybuilders and movie stars who had to buff up quickly. He wrote a book called Unleashing The Wild Physique which was published in 1984. In that book, he said that the medical reports about eggs were false. He explained that for many years he had been having his pupils eat a dozen or more eggs a day as part of their training diet and their cholesterol didn't go up. Eventually, the medical profession started to test people and found the same thing, although eggs contain cholesterol, they don't increase your cholesterol. My point being, testing and real life results trump "theory".

I agree with all of this down to the eggs Pandy!:ThmbUp:

[GBL]

Just wanted to add one last thing about decision making. When I first started to track contender selection ie 1st choice 2nd choice 3rd choice etc I noticed the drop off after the first 2. Top 2 would hit in the 25-30% range then 3rd choice would be 12-15% and 4th choice 8-10% etc. And I realized that many 3rd/4th choices and even 2nd choices weren't worth betting anyway so I started tracking them simply by main contenders and marginal contenders. Similar results over the last 1,000 races. Main contenders represent 21% of the entries and win 55% of the races. Marginal contenders represent 29% of the entries and win 31% of the races. It was what I had suspected all along. Marginal contenders do not win enough to develop a betting approach with them. And the mutuels are not any higher than the 9.07 on main contenders. And with the precipitous dropoff in hit rate evidenced by the 12-15% and 8-10% the average mutuel would have to be fairly astronomical to justify betting these horses. This reality has led to me to a conclusion I wholly believe=that value handicapping can lose itself quickly and become even a futile quest for value if not grounded.

That said with main contenders representing 21% of entries there are plenty of opportunities to bet value. I make practically no distinction between my top choices in a race. On handicapping principle I can split hairs if necessary but when win betting I simply bet the horse not taking action. If both aren't getting play I bet both. End of strategy. From 2001 to 2012 my average mutuel stayed with .50 of 9.40. Since then it has hovered within .50 of 12.23. All I did was restrict my win betting criteria to the above. Nearly 4 months into the new horse betting year my average mutuel is happily over $15, but waves of favorites the last few weeks are bringing that back into line.

None of that has anything to do with exotics. Lower mutuel horses always weave their way through payoffs and can still produce the desired return on bets. But I find the above to be a solid foundation for all plays to have a solid expectation. Or you don't bet. Period!

pandy,
Thanks for positive and healthy exchange. Agree with many points you raised.

[AltonKelsey]

Shouldn't have to say it, but the number of bets you make has nothing to do with winning or losing IN ITSELF.

Some of you are confused. Many ?

A small sample is more likely to WIN even if the long term roi stinks.

A large sample will almost surely lose.


Thus, the confusion, thinking that many bets is somehow a negative.

For positive long term roi's its exactly what you want, as many plays as possible

[CincyHorseplayer]

Shouldn't have to say it, but the number of bets you make has nothing to do with winning or losing IN ITSELF.

Some of you are confused. Many ?

A small sample is more likely to WIN even if the long term roi stinks.

A large sample will almost surely lose.


Thus, the confusion, thinking that many bets is somehow a negative.

For positive long term roi's its exactly what you want, as many plays as possible

Absolutely. I think a 100 bet cycle covers all the swings of variance. The more of these you can go through a year the better.

[ultracapper]

Again, that's a theory, and I get the theory, but it doesn't work in real life. Many years ago I said the same thing in some of my columns, but I was wrong. That's why testing is better than theory. A great handicapper, someone who is great at picking winners, is still not going to be able to show a profit on win bets on 8/5 shots, or favorites in general. If you can get to something like a 5% loss betting favorites, you're a master handicapper. Everytime I've had a good year and actually went back and checked the results, the favorites I bet only provided a mental stability that cut down on a series of losing bets, but they were not profitable. All of my profit came from longshot winners that I keyed in winning exotics. Most winning players get the bulk of their profit from several nice hits each year. 8/5 shots don't provide that, and you simply can't overcome the takeout with favorites.

There are also a lot more overlays in the higher odds range because it requires a really sharp handicapper to see that a 15-1 shot should actually be 6-1. But with favorites, everyone can see why the horse is favored.

Now, I agree, that it's good to have the ability to spot bad favorites and bet against them. But that doesn't mean that it's smart to bet on the horses that appear to be legit favorites. Bet against the bad favorites, pass the races where the favorite appears legit.

The best and fastest horse in the race still loses more than it wins, so you simply can't take short prices.

Not a thought out of place.

Great post.

[PaceAdvantage]

Absolutely. I think a 100 bet cycle covers all the swings of variance.I can personally attest to the fact that this is VERY false...:lol:

[thaskalos]

There is a psychological aspect to horse-betting that I feel has been largely ignored in this discussion. Even more important than adopting the most "profitable" betting techniques is the all-important matter of selecting a style of play which matches with our temperament and personality as players. Before we decide to "ignore the legitimate favorites, even in our exotics wagering"...we should spend a little time determining if we are individually capable of doing such a thing consistently in our actual play. TALKING about doing something isn't the same as actually DOING it. "Most knowledge is self-knowledge" the ancient sages have declared...and I feel that this applies most strongly and directly in gambling.

In my own play, I have learned enough about myself as a player to know that limiting my bets to horses that go off at odds of 6-1 or higher would drive me absolutely crazy...especially in today's horse-betting landscape. How many legitimate 6-1+ betting choices am I supposed to find when all I see -- especially on weekdays -- are short fields of 7 horses or less? If I confine myself to only my "longshot" plays, then I run the risk of becoming lethargicly unprepared for the prime betting opportunities which pop up unannounced during the course of the day's play. To play my best, I must remain sharp during the course of the day...and passing race after race causes me to lose track of the "flow" that I have to be in, in order to play my "A" game. Consequently, I have created a rational style of play which INCLUDES betting on some favorites, not as a cure for my "boredom", but as a device for keeping me sharp and aware of the betting opportunities that may still lay ahead. My prime consideration as a horseplayer isn't to employ the most "long-term-profitable" handicapping and betting techniques, whatever those may be. My main horse-betting concern is to create a betting style which matches my temperament and personality...on which I can rely upon for a nice balance between "profit", and "performance". A few of my well-meaning friends have suggested to me that I might be losing "ROI points" by utilizing my looser betting style...but I know from personal experience that a higher ROI does not necessarily translate to a higher long-term profit. The vital thing is to discover what works best for us as individuals.

"Know thyself...and then you will know all important things" declared Socrates...who purportedly made a fortune wagering on the chariot races in ancient Athens. :)

[AltonKelsey]

I can personally attest to the fact that this is VERY false...:lol:

I agree, 100 is too small a sample. Maybe 500 or more plays over a full season might give you a pretty good idea, but nothing is written in stone, as the game is constantly morphing (as are payoffs) .

Depends on the kind of bets you're making.

[classhandicapper]

Another thing I'm testing, something that has always been a puzzle for handicappers is, do you play your top pick, or do you handicap contenders and then look for value? So say you handicap a race and come up with four horses ranked in order of preference. Regardless of what method you're using to handicap, I think you could make a case that you should never bet the favorite. So if your top pick is the favorite, you should either pass the race, or look over your other three ranked horses to see if there's any value there. But I'm not sure that you should ever bet the second choice, either. The first two choices win around 54% of the races, but the value is usually in the horses that win less often... it's sort of a paradox because you have to find the best value among the horses that are less likely to win the race.

There is theory and there is reality.

I theory, you should make a betting line and bet the best value regardless of where the horse ranks. The reality is that we are all working with incomplete information and understanding. So making odds lines is fraught with peril.

I've found that "in general" I am better off sticking to my top choices and passing the race when the horse is below 5/2. I am a little flexible on that bottom limit if I think the horse has a huge edge and get a little tighter if I have some extra concerns or the field seems deep.

I will occasionally bet someone further down my ranking, but only if the horse figures so close to my top choice it was a real struggle to separate them. Then I might even bet both of them if the odds are right.

I will also occasionally use a live longshot that ranks further down by trying to get him into the exacta, trifecta or super, but I won't key the horse on top. I'll use him in different spots.

If I totally hate the favorite I will also spread a little trying to beat him even if I don't have a strong opinion.

[Dave Schwartz]

I agree, 100 is too small a sample. Maybe 500 or more plays over a full season might give you a pretty good idea, but nothing is written in stone, as the game is constantly morphing (as are payoffs) .

Depends on the kind of bets you're making.

We have had players who did very large live-play tests. One in particular comes to mind who was very profitable at 1,500 races played and was slightly loser after 2,400.

That is a lot of races to bang in by hand without making any real bets.

[CincyHorseplayer]

I can personally attest to the fact that this is VERY false...:lol:

Not here. Usually a losing streak in there over 100 bets. And a run or two. My last 10 years of bets aren't any different than my last 100 bets. Or very little. I don't know how that could be for you!

[PaceAdvantage]

I have turned a superb ROI and win% over 100+ races into a pile of dogshit in no time flat...recently...

[CincyHorseplayer]

I have turned a superb ROI and win% over 100+ races into a pile of dogshit in no time flat...recently...

I find myself usually flat or plain total dogshit for about 20-30 bets with anywhere from 40-60% loss on everything I put into play. Things mysteriously change back again to normal as they did when going south.

Maybe it's the time of the year but I've been doggy paddling recently. I don't think it's any coincidence that it is at a time where sub 2-1 shots are crushing.

[thaskalos]

I have turned a superb ROI and win% over 100+ races into a pile of dogshit in no time flat...recently...

Don't feel bad. I've done this more than a few times.

[steveb]

P.S. to the above in light of overlays: Just like overlaid favorites, not every high priced-overlay is a good, bet-able overlay. The 15/1 in your assessment system going off as 25/1 doesn't necessarily makes it a good bet.

why?
if your probs are as correct as possible, and your 15/1 chances win at that percentage(or greater if you set your probs to less than 1) then there can be but one result.
if not, then that insinuates that your 15/1 horses should have been given longer prices.

[steveb]

why?
if your probs are as correct as possible, and your 15/1 chances win at that percentage(or greater if you set your probs to less than 1) then there can be but one result.
if not, then that insinuates that your 15/1 horses should have been given longer prices.

and i should have added.....if they win at a greater price than your assessed price.
it's no good if your 15/1 chances are winning at that rate, but at lesser average odds.
that would be negative expectation, but it's not what you stated.
25/1 about 15/1 pops is a GREAT bet, provided your probs are reasonable.

[classhandicapper]

I have turned a superb ROI and win% over 100+ races into a pile of dogshit in no time flat...recently...

One system I tested was profitable for 2015 and 2016 but fell apart in 2017.

I have another one that was hovering around break even for the latter part of 2014 (when testing began), all of 2015, all of 2016, and all 2017, but is getting killed so far in 2018. That's for about 500-600 horses per year that are a mix of favorites, 2nd choices, and 3rd choices, with an occasional longshot mixed in. I have no idea how 2018 will end up, but I think to really be certain you need some serious samples sizes.

[JerryBoyle]

One system I tested was profitable for 2015 and 2016 but fell apart in 2017.

I have another one that was hovering around break even for the latter part of 2014 (when testing began), all of 2015, all of 2016, and all 2017, but is getting killed so far in 2018. That's for about 500-600 horses per year that are a mix of favorites, 2nd choices, and 3rd choices, with an occasional longshot mixed in. I have no idea how 2018 will end up, but I think to really be certain you need some serious samples sizes.

I've been struggling with this as of late. I've attached some charts of cumulative profit for a model I backtested. First image is a cumulative profit chart that I was proud of, which covers ~3000 races from 20160201 to 20170501. Utilized kelly betting with a fixed $1k bankroll at the start of each race (no reinvestment). Second attachment is the same model from 20170901 to 20180301. I agree that 500 bets may not be enough...

I think it comes down to having a good way to evaluate how much better our predictions are doing than the publics. No strategy is going to last forever, as we're only winning when we're doing something that no one else is. It has taught me that having a way to i.d when to turn a strategy off is as valuable as i.d'ing the strategy in the first place.

[lansdale]

I've been struggling with this as of late. I've attached some charts of cumulative profit for a model I backtested. First image is a cumulative profit chart that I was proud of, which covers ~3000 races from 20160201 to 20170501. Utilized kelly betting with a fixed $1k bankroll at the start of each race (no reinvestment). Second attachment is the same model from 20170901 to 20180301. I agree that 500 bets may not be enough...

I think it comes down to having a good way to evaluate how much better our predictions are doing than the publics. No strategy is going to last forever, as we're only winning when we're doing something that no one else is. It has taught me that having a way to i.d when to turn a strategy off is as valuable as i.d'ing the strategy in the first place.

Your performance in the first chart is spectacular, and the contrast with the second suggests that you were either playing different tracks or different types of races. But, I would give more weight to the first since it's more than twice the size of the second and more than twice the length of time -- the second chart appears to cover only six months of play. One explanation that strikes me is that the second chart covers mainly winter racing, when handicapping opportunities are much worse in many places. If you're simply playing on the same tracks which are less congenial to your model during winter months, you either need to play elsewhere or simply not play.

Your instinct that 500 races is too small a sample to be meaningful is correct -- read Dave. S.'s post above -- he's right. Also, very much worth reading Benter's article on building statistical models for horseracing -- he also claims he saw very little variance in the performance of his model after 2,500 races. As I've said before, this appears to be the minimum 'long-run' of horseracing -- for blackjack and poker the relevant number of trials is much higher.

Most important thought, is what you mean when you say you're Kelly betting. How are you determining your Kelly bet size? As you probably know, Kelly betting can often produce a roller-coaster-like ride, which many gamblers prefer to avoid. That also could explain the huge swing you experienced here. I could be wrong, but you might be better off with flat-betting -- which can also be profitable. But, bottom line, I would take that first chart very seriously -- I think you really have something there.

Cheers,

lansdale

Anohttps://www.channel4.com/news/data-democracy-and-dirty-tricks-cambridge-analytica-uncovered-investigation-expose

[pandy]

Thaskalos mentioned the mental aspect of only betting horses 6-1 or higher. I don't disagree when you play a mixture of odds and include favorites, you cash a lot more tickets, and although the ROI may be lower, if you can still grind out a profit, albeit smaller profit, psychologically, that's a lot less stressful. My point about ROI was that it tends to be higher when you avoid favorites. From a purely mathematical viewpoint, over the long run, favorites bring down your ROI. I also think it's tougher than ever to find overlays that are below 5-2 odds. And even if you do, the profit margin is slim. On win bets, if you can generate a consistent 1% profit on horses below 5-2, you're a master handicapper. Every test I've ever run has shown that it's virtually impossible to generate a long-term profit betting favorites to win.

[steveb]

Thaskalos mentioned the mental aspect of only betting horses 6-1 or higher. I don't disagree when you play a mixture of odds and include favorites, you cash a lot more tickets, and although the ROI may be lower, if you can still grind out a profit, albeit smaller profit, psychologically, that's a lot less stressful. My point about ROI was that it tends to be higher when you avoid favorites. From a purely mathematical viewpoint, over the long run, favorites bring down your ROI. I also think it's tougher than ever to find overlays that are below 5-2 odds. And even if you do, the profit margin is slim. On win bets, if you can generate a consistent 1% profit on horses below 5-2, you're a master handicapper. Every test I've ever run has shown that it's virtually impossible to generate a long-term profit betting favorites to win.

i can't see how that is correct.

favourites would likely have you losing less overall than longer prices would.
the favouriites would tend to be closer to their 'true' odds, and the roughies would tend to salute far less often than their public probs suggest they should.

but a player that had a sound methodology, would trust that over whatever the public says, so that the public odds would play no(or minor) part in his/her thinking.

the public may well be right overall, but they are also very very often miles out.
the variance is huge.

[pandy]

i can't see how that is correct.

favourites would likely have you losing less overall than longer prices would.
the favouriites would tend to be closer to their 'true' odds, and the roughies would tend to salute far less often than their public probs suggest they should.

but a player that had a sound methodology, would trust that over whatever the public says, so that the public odds would play no(or minor) part in his/her thinking.

the public may well be right overall, but they are also very very often miles out.
the variance is huge.

Sometimes you'll see stats comparing the ROI on thousands of favorites to horses at various odds, and the very high odds have a very low ROI. But that's random, using every horse. Longshots that have very little chance to win aren't hard to eliminate. Those type of stats based on "all" horses at certain odds are not relevant.

Again, from a mathematical viewpoint, the takeout is just too high to overcome when betting favorites to win, and the rebates are too low on win bets, so even if you bet through a rebate house, it doesn't help enough.

And I'm not recommending bets on horses that are 40-1. 6-1 to 20-1 is a good range, or even 5-2 to 20-1, or 4-1 to 20-1. Once you go below 5-2 odds, on win bets, it's almost impossible to show a profit on those bets.

[PaceAdvantage]

Your performance in the first chart is spectacular, and the contrast with the second suggests that you were either playing different tracks or different types of races.Some of the drawdowns in that first chart would have sent the less hearty among us scurrying for the exits long before we gave it a chance to pan out over the next billion races...:lol:

[GBL]

Sometimes you'll see stats comparing the ROI on thousands of favorites to horses at various odds, and the very high odds have a very low ROI. But that's random, using every horse. Longshots that have very little chance to win aren't hard to eliminate. Those type of stats based on "all" horses at certain odds are not relevant.

Again, from a mathematical viewpoint, the takeout is just too high to overcome when betting favorites to win, and the rebates are too low on win bets, so even if you bet through a rebate house, it doesn't help enough.

And I'm not recommending bets on horses that are 40-1. 6-1 to 20-1 is a good range, or even 5-2 to 20-1, or 4-1 to 20-1. Once you go below 5-2 odds, on win bets, it's almost impossible to show a profit on those bets.


As you might've understood already, I am a believer of mixed-odds with an overall good hit ratio. They both complement my ROI purposes. One thing I certainly don't do is bet odds-on no matter how strong they look. The lowest I can go is probably 8/5. But that's not the point.

The problem with any ROI win-statistics is they're based on $2 flat win bet regardless of odds. That's good for statistical purposes but hardly practical.

In my case, if I like 2/1 favorite (which I think should've been 7/5), I bet HEAVY. Actually, much heavier. For example, if I generally play $20 win on a 12/1, I probably would go 10 times more on a 5/2. Similarly, if I really love an 8/1, I don't restrict myself to any fixed amount rather try to go as heavy as I think feasible.

Isn't it that's how we all play?

[CincyHorseplayer]

I think what Pandy is saying is that odds are merely a point of departure not an end in itself. A certain odds range puts you in the neighborhood of profitable expectation. Being tote board dependent and a prisoner of the true odds concept is not remotely the same thing. And not only that with win betting you are not only at the mercy of the tote board but a single outcome. They win or lose. I use the odds as merely a gauge for a potential play. And payoffs as validation or not. Heavy favorites or even 5/2 good things can still drag down all the payoffs attached to them. But there are plenty of 2 horses races that have 5/2 and 2-1 shots that aren't being hammered in exotics that are playable. In my own play since December the return on my bets is a touch above 7.50-1. 2/3rds of my play is exotics to 1/3rd win. Win bet profits can evaporate quickly by even a mildly bad run. Being married to a single outcome and also when favorites at multiple tracks go on a rampage it is unavoidable. Having up to 6 outcomes that yield a higher odds return is less volatile. If you aren't getting payoff validation you pass. Pretty basic IMO.

[pandy]

As you might've understood already, I am a believer of mixed-odds with an overall good hit ratio. They both complement my ROI purposes. One thing I certainly don't do is bet odds-on no matter how strong they look. The lowest I can go is probably 8/5. But that's not the point.

The problem with any ROI win-statistics is they're based on $2 flat win bet regardless of odds. That's good for statistical purposes but hardly practical.

In my case, if I like 2/1 favorite (which I think should've been 7/5), I bet HEAVY. Actually, much heavier. For example, if I generally play $20 win on a 12/1, I probably would go 10 times more on a 5/2. Similarly, if I really love an 8/1, I don't restrict myself to any fixed amount rather try to go as heavy as I think feasible.

Isn't it that's how we all play?

I usually bet heavier on longshots. First of all, I agree with you that you should have some sort of cut off point. You say that you prefer at least 8-5 odds. I would say that making win bets on thoroughbreds who are below 8-5 odds is a losing proposition, so why bother? But horses that are 8-5 to 2-1, also almost impossible to generate a profit on a win bet, and not easy in exactas, either.

I bet heavier on the higher odds horses because I can't get that excited about the lower odds horses. And, in my case, I know that longshots generate my profit, so naturally, I bet more on them.

[PaceAdvantage]

Shouldn't you be betting more on your largest overlays regardless of what their tote odds are?

In fact, the lower the odds, the more you can bet on overlays in that range.

[GBL]

I usually bet heavier on longshots. First of all, I agree with you that you should have some sort of cut off point. You say that you prefer at least 8-5 odds. I would say that making win bets on thoroughbreds who are below 8-5 odds is a losing proposition, so why bother? But horses that are 8-5 to 2-1, also almost impossible to generate a profit on a win bet, and not easy in exactas, either.

I bet heavier on the higher odds horses because I can't get that excited about the lower odds horses. And, in my case, I know that longshots generate my profit, so naturally, I bet more on them.

Isn't it interesting how we all reach our goals taking very different routes and using different methodologies?

There's never a single and only way to success :-)

[GBL]

Shouldn't you be betting more on your largest overlays regardless of what their tote odds are?

In fact, the lower the odds, the more you can bet on overlays in that range.

I would say there's no hard and fast rule here. It all depends upon how strongly I feel. Adaptability, open mind, and flexibility has done me good, so far.

[classhandicapper]

i can't see how that is correct.

favourites would likely have you losing less overall than longer prices would.
the favouriites would tend to be closer to their 'true' odds, and the roughies would tend to salute far less often than their public probs suggest they should.

but a player that had a sound methodology, would trust that over whatever the public says, so that the public odds would play no(or minor) part in his/her thinking.

the public may well be right overall, but they are also very very often miles out.
the variance is huge.

This is an example of the theory vs. reality problem I brought up earlier.

You would think it would be easier to find value among the group of horses that's losing less per dollar wagered than a group that's losing a lot more (favorites vs. longshots). The problem is that there are apparently more individual large errors among mid priced horse and perhaps longshots than there are among favorites. Even if you are pretty good at identifying the occasional very bad favorite, the remainder is a tough nut to crack. I'm sure there are people that can do it if they are selective, but the ROI on favorites is not going to be impressive unless perhaps you get ridiculously selective.

Another way of thinking about it is this.

If a horse is already below 5/2, what are the chances you know something "positive enough" about it that everyone else doesn't already know to overcome the take.

[pandy]

This is an example of theory vs. reality that I brought up earlier.

You would think it would be easier to find value among the group of horses that's losing less per dollar wagered than a group that's losing a lot more (favorites vs. longshots). The problem is that there are apparently more individual large errors among mid priced horse and perhaps longshots than there are among favorites. Even if you are pretty good at identifying the occasional very bad favorite, the remainder is a tough nut to crack. I'm sure there are people that can do it if they are selective, but the ROI on favorites is not going to be impressive unless perhaps you get ridiculously selective.

I agree. When I bet favorites, I would sometimes get on a roll and click off four or five winners in a row, and maybe 8 winners in 11 bets. Wow, easy money! But invariably, the law of averages caught up and suddenly I lost four bets in a row and 8 of my next 11 bets on horses that looked so strong on paper, and there goes the profit. With double digit horses, there's more leverage and the big hits you can put together make up for a lot of losers.

[storyline]

Your performance in the first chart is spectacular, and the contrast with the second suggests that you were either playing different tracks or different types of races. But, I would give more weight to the first since it's more than twice the size of the second and more than twice the length of time -- the second chart appears to cover only six months of play. One explanation that strikes me is that the second chart covers mainly winter racing, when handicapping opportunities are much worse in many places. If you're simply playing on the same tracks which are less congenial to your model during winter months, you either need to play elsewhere or simply not play.

Your instinct that 500 races is too small a sample to be meaningful is correct -- read Dave. S.'s post above -- he's right. Also, very much worth reading Benter's article on building statistical models for horseracing -- he also claims he saw very little variance in the performance of his model after 2,500 races. As I've said before, this appears to be the minimum 'long-run' of horseracing -- for blackjack and poker the relevant number of trials is much higher.

Most important thought, is what you mean when you say you're Kelly betting. How are you determining your Kelly bet size? As you probably know, Kelly betting can often produce a roller-coaster-like ride, which many gamblers prefer to avoid. That also could explain the huge swing you experienced here. I could be wrong, but you might be better off with flat-betting -- which can also be profitable. But, bottom line, I would take that first chart very seriously -- I think you really have something there.

Cheers,

lansdale

Anohttps://www.channel4.com/news/data-democracy-and-dirty-tricks-cambridge-analytica-uncovered-investigation-expose


Almost never have I heard anyone speak or even mention this very real variable (spring, summers vs winter racing) when manipulating databases for modeling.

Do most handicappers ask the right questions when determining which horse to play?

I think if you are basing wagering decisions on past performance of horse and trainer you'll miss cashing tickets when price-horses run new tops which happens everyday at a track near you.

Perhaps handicappers could spend time determining why some horses are ready to run a big race under today's conditions or not.

[JerryBoyle]

Some of the drawdowns in that first chart would have sent the less hearty among us scurrying for the exits long before we gave it a chance to pan out over the next billion races...:lol:


Haha, yep. I wouldn't bet this directly, as there are some simplifications made to allow quicker generation of the backtest. It's also full kelly rather than some fraction. But yeah, that 100%+ drawdown in the beginning is a bit worrisome...

[JerryBoyle]

Almost never have I heard anyone speak or even mention this very real variable (spring, summers vs winter racing) when manipulating databases for modeling.

Do most handicappers ask the right questions when determining which horse to play?

I think if you are basing wagering decisions on past performance of horse and trainer you'll miss cashing tickets when price-horses run new tops which happens everyday at a track near you.

Perhaps handicappers could spend time determining why some horses are ready to run a big race under today's conditions or not.

My (lazy) attempt at handling seasonality is to roll my model recalibration as new races come in. Thinking is that if my model utilizes enough factors, whose relative impact varies by season, recalibrating as we enter the season will shift the weights as appropriately. That backtest was done using a rolling recalibration, so it may not be the best strategy :)

[storyline]

My (lazy) attempt at handling seasonality is to roll my model recalibration as new races come in. Thinking is that if my model utilizes enough factors, whose relative impact varies by season, recalibrating as we enter the season will shift the weights as appropriately. That backtest was done using a rolling recalibration, so it may not be the best strategy :)



If I had to choose between "more filtering" or "more testing" I'd pick more filtering everytime.

Garbage in - garbage out.

Not all factors have the same weight/significance for the same horse at all times throughout the year.

Data mining is a slippery slope if someone is entering all the different variables and then trying to gleam meaningful information imo

[Dave Schwartz]

Almost never have I heard anyone speak or even mention this very real variable (spring, summers vs winter racing) when manipulating databases for modeling.

Do most handicappers ask the right questions when determining which horse to play?

I think if you are basing wagering decisions on past performance of horse and trainer you'll miss cashing tickets when price-horses run new tops which happens everyday at a track near you.

Perhaps handicappers could spend time determining why some horses are ready to run a big race under today's conditions or not.

Generally, I do not play during the winter. Last November, just before Thanksgiving, I did a live play session (i.e. with an audience) and the racing was just horrible. My handicapping was not so hot, either.

I dedicated several weeks to figuring out just how different winter racing really is.

For over a decade I have used month-of-year in gathering races from the database to build a model of "races like this one." Specifically, +/- 2 months.

But this was different.

The first thing I determined was that racing is "different" from the week before Thanksgiving to around Jan. 21. The races are just far less predictable and/or do not match my handicapping approaches.

Ironically, many of the usual factor values - like recent speed ratings, for example - actually perform BETTER! But somehow the puzzle just doesn't fit together properly; the picture of the winner's circle just does not match the puzzle box.

I THINK... but have no proof of this... that it stems from the trainer patterns changing for the holidays. I THINK that the best trainers simply take time off from racing and hand the reigns over to assistants for about two months. This causes differences.

Very open to other ideas. BTW, it is not just weather because things change even in the warmer climates. (Not as much, but still different.)


Dave

[CincyHorseplayer]

Generally, I do not play during the winter. Last November, just before Thanksgiving, I did a live play session (i.e. with an audience) and the racing was just horrible. My handicapping was not so hot, either.

I dedicated several weeks to figuring out just how different winter racing really is.

For over a decade I have used month-of-year in gathering races from the database to build a model of "races like this one." Specifically, +/- 2 months.

But this was different.

The first thing I determined was that racing is "different" from the week before Thanksgiving to around Jan. 21. The races are just far less predictable and/or do not match my handicapping approaches.

Ironically, many of the usual factor values - like recent speed ratings, for example - actually perform BETTER! But somehow the puzzle just doesn't fit together properly; the picture of the winner's circle just does not match the puzzle box.

I THINK... but have no proof of this... that it stems from the trainer patterns changing for the holidays. I THINK that the best trainers simply take time off from racing and hand the reigns over to assistants for about two months. This causes differences.

Very open to other ideas. BTW, it is not just weather because things change even in the warmer climates. (Not as much, but still different.)


Dave

I believe bad weather affects the racing. I still might bet Mahoning or Aqueduct in the winter on a fast track but I long ago stopped betting any northern racetracks after the Breeders Cup.

What you mention about trainers I feel is true but for me it's January where I see the most chalk and nonsense results over the years. Everywhere. But I admit that hasn't been true for me the last 2 years betting Fair Grounds, Tampa, and GP.

Other than January for me the worst months are September and November. The pools close in this city and Saratoga is over after Labor day. And unless I am elsewhere the summer is officially over feel leaves me uninspired and I don't do much good handicapping work. Then after the Breeder's Cup and before the winter meets kick off it's much of the same unless I am absolutely itching for action. December 1st is my horseplaying new year so I usually sit this month out in anticipation anyway!

Anyway just piggybacking random thoughts of your post because it reminded me of these things!:)

[storyline]

Generally, I do not play during the winter. Last November, just before Thanksgiving, I did a live play session (i.e. with an audience) and the racing was just horrible. My handicapping was not so hot, either.

I dedicated several weeks to figuring out just how different winter racing really is.

For over a decade I have used month-of-year in gathering races from the database to build a model of "races like this one." Specifically, +/- 2 months.

But this was different.

The first thing I determined was that racing is "different" from the week before Thanksgiving to around Jan. 21. The races are just far less predictable and/or do not match my handicapping approaches.

Ironically, many of the usual factor values - like recent speed ratings, for example - actually perform BETTER! But somehow the puzzle just doesn't fit together properly; the picture of the winner's circle just does not match the puzzle box.

I THINK... but have no proof of this... that it stems from the trainer patterns changing for the holidays. I THINK that the best trainers simply take time off from racing and hand the reigns over to assistants for about two months. This causes differences.

Very open to other ideas. BTW, it is not just weather because things change even in the warmer climates. (Not as much, but still different.)


Dave

Lots of things here worth discussing Dave.

A small list of things not captured by simply inputting data

Adjusting for maturity, lasix, barn changes, breeding, jockey changes, equipment changes, quality of competition, how figures were earned, class levels, form, trainer intent, race shape, speed of track, determining false favorites (I'll stop here)

Example to a question that any handicapper might ask - #2 Noble Indy is racing for Pletcher in the La Derby Saturday and is getting blinkers. Statistically Pletcher horses do well in those situations but we're talking about a trainer that does pretty well in all situations (25% win Trainer).

The question that needs to be answered is will Noble Indy move forward, regress or stay about the same because of the equipment change? Pretty important question if you're betting the race imo.

[storyline]

Lots of things here worth discussing Dave.

A small list of things not captured by simply inputting data

Adjusting for maturity, lasix, barn changes, breeding, jockey changes, equipment changes, quality of competition, how figures were earned, class levels, form, trainer intent, race shape, speed of track, determining false favorites (I'll stop here)

Example to a question that any handicapper might ask - #2 Noble Indy is racing for Pletcher in the La Derby Saturday and is getting blinkers. Statistically Pletcher horses do well in those situations but we're talking about a trainer that does pretty well in all situations (25% win Trainer).

The question that needs to be answered is will Noble Indy move forward, regress or stay about the same because of the equipment change? Pretty important question if you're betting the race imo.


Now here's the trick, if you can accurately assess that Noble Indy will move forward the next question you might ask is by "how much". Is it enough to dominate this field?

That answer could significantly influence your wagering decision and amount invested.

[Denny]

What's great about 25%?

That means he loses 75%.

Statistics are misused by most.

[storyline]

What's great about 25%?

That means he loses 75%.

Statistics are misused by most.



That's what you got from my posts? Amazing.

[Denny]

Thaskalos is right. Knowing ones own psychological makeup is critical.

Not only in dealing with winning and losing, knowing one's own self when handicapping and betting is essential. Unless you're a computer, your decision making is influenced by your present state of mind and beliefs.

[classhandicapper]

Almost never have I heard anyone speak or even mention this very real variable (spring, summers vs winter racing) when manipulating databases for modeling.

Do most handicappers ask the right questions when determining which horse to play?

I think if you are basing wagering decisions on past performance of horse and trainer you'll miss cashing tickets when price-horses run new tops which happens everyday at a track near you.

Perhaps handicappers could spend time determining why some horses are ready to run a big race under today's conditions or not.

I think handicappers sort of understand that the game changes seasonally.

For example, the winter in NY is probably the best time to pay attention to trainers that do well off the claim because that's the time of year they can make the most money. There are more opportunities and the trainers with quality stock leave.

At this time of year handicappers in NY are getting ready to deal with a lot of horses shipping up from Florida and possibly other warm areas. It probably makes sense to pay attention to which barns are sending their horses up ready for a peak effort against the weaker NY horses. That's something you won't have to deal with at other meets.

They also know there are going to be a lot more turf horses coming off layoffs because they were turned out and rested during the winter months. The same is even true of many stakes horses that were given 2-3 months off specifically for a freshening.

Personally, I have never tried to change my methods seasonally, but I realize aspects of the game change seasonally and certain strengths or weakness in my own game could be accented by what's important at that time of year.

[Denny]

Lots of things here worth discussing Dave.

Example to a question that any handicapper might ask - #2 Noble Indy is racing for Pletcher in the La Derby Saturday and is getting blinkers. Statistically Pletcher horses do well in those situations but we're talking about a trainer that does pretty well in all situations (25% win Trainer).

The question that needs to be answered is will Noble Indy move forward, regress or stay about the same because of the equipment change? Pretty important question if you're betting the race imo.


It was the question you asked. Most of the rest was useless bs to me.

Statistics is a large component of this thread. Though most of them are misinterpreted by most.

That's whats truly "AMAZING"!

[storyline]

It was the question you asked. Most of the rest was useless bs to me.

Statistics is a large component of this thread. Though most of them are misinterpreted by most.

That's whats truly "AMAZING"!


That's precisely the point, I'm imploring players to measure the horse more as opposed to measuring statistics more.

[storyline]

Getting the average of previous figures is not useful for handicapping and betting reasons. In contrary, in most of the cases such an approach will suppress your chances of catching career best performances, which is where most of the betting interest lies in.

Also using constraints like “same surface” or “same distance” make your process more conservative focusing to an easier to attack problem that is also solved pretty well by the betting crowd. Focusing on more challenging problems, like surface, distance or circuit switches is way more attractive as this is where most of the crowd miscalculations occur.

As far as the second part of your posting, I think you are asking the wrong question. Your best betting selection should not be decided by its ranking (meaning whether is your top selection of the fourth one) but from the possibility of the crowd to misestimate its angles regardless of how its final price will be shaped in the market. The crowd is really good in ABC handicapping so trying to beat it by forming a superior ranking of horses (meaning where your top picks will win more often than the pubic favorite for example) is simply a dream pipe that is impossible to accomplish. The best you can do is to detect the way the public is forming the pools (something that changes constantly through time and also differs from track to track) and try to “correct” any potential mistake before the crowd realize it and encapsulate it to its betting.


Not sure how I missed reading this post which imho is entirely correct, well done Dlover :ThmbUp:

[JerryBoyle]

My (lazy) attempt at handling seasonality is to roll my model recalibration as new races come in. Thinking is that if my model utilizes enough factors, whose relative impact varies by season, recalibrating as we enter the season will shift the weights as appropriately. That backtest was done using a rolling recalibration, so it may not be the best strategy :)

Made some modifications and improved my model a bit. The attached chart includes the testing and validation ranges. Clearly the validation range is much better than the previous I posted. Results in general are also much better w.r.t draw downs.

There are man simplifying assumptions made in the backtest which would cause real results to likely be significantly worse. Not sure how many people on the forum are doing much modeling, but I'd be interesting in starting a chat about going from test results to actual live betting.

[acorn54]

to determine if a model exceeds randomness, it must exceed the critical z score of the null hypothesis for the model you are working on.
when you have such a model, assuming you ever do get to that point of development, you use it in real time going forward to see if it continues to exceed randomness, and of course if it produces a profit commensurate with the risk you are taking using said model. that is how you "test overlays"

[JerryBoyle]

to determine if a model exceeds randomness, it must exceed the critical z score of the null hypothesis for the model you are working on.
when you have such a model, assuming you ever do get to that point of development, you use it in real time going forward to see if it continues to exceed randomness, and of course if it produces a profit commensurate with the risk you are taking using said model. that is how you "test overlays"

What metric do you use and over what time horizon? E.g weekly roi, daily profit, etc?

[acorn54]

What metric do you use and over what time horizon? E.g weekly roi, daily profit, etc?


there are statistical tables that give the critical score for various confidence levels. i highly suggest a statistical textbook that you can pick up at your local college for elaboration. as to the time horizon used and sample size. that is the purpose of using statistical tables for the certainty of whether what you are getting in your results exceed random occurance and the certainty (confidence level) that going forward the results will be repeated. as a caveat, there is no absolute certainty that whatever model you use, even if it exceeds randomness, is going to be a "sure thing", there will ALWAYS be chance of failure. the best that statisticians can achieve is MAYBE 90 percent certainty.

[JerryBoyle]

there are statistical tables that give the critical score for various confidence levels. i highly suggest a statistical textbook that you can pick up at your local college for elaboration. as to the time horizon used and sample size. that is the purpose of using statistical tables for the certainty of whether what you are getting in your results exceed random occurance and the certainty (confidence level) that going forward the results will be repeated. as a caveat, there is no absolute certainty that whatever model you use, even if it exceeds randomness, is going to be a "sure thing", there will ALWAYS be chance of failure. the best that statisticians can achieve is MAYBE 90 percent certainty.

Sorry, didn't mean metric as far as z score. I meant what metric are you measuring the distribution of? E.g. profit, roi, etc?

[acorn54]

Sorry, didn't mean metric as far as z score. I meant what metric are you measuring the distribution of? E.g. profit, roi, etc?

i understood what you meant. i keep the factors that i use to myself, however i am sure you will be able to find your own unique factors.

[lansdale]

Made some modifications and improved my model a bit. The attached chart includes the testing and validation ranges. Clearly the validation range is much better than the previous I posted. Results in general are also much better w.r.t draw downs.

There are man simplifying assumptions made in the backtest which would cause real results to likely be significantly worse. Not sure how many people on the forum are doing much modeling, but I'd be interesting in starting a chat about going from test results to actual live betting.


Not sure what you're doing here or if you've tested vs. holdout samples, but your growth chart here very much resembles that of a professional blackjack player, or, as I said, Bill Benter -- a random walk with an upward drift. If so, you're on your way to multi-millionaire status.

[JerryBoyle]

Not sure what you're doing here or if you've tested vs. holdout samples, but your growth chart here very much resembles that of a professional blackjack player, or, as I said, Bill Benter -- a random walk with an upward drift. If so, you're on your way to multi-millionaire status.

That is my hold out sample, though rather than having one single training period with one single hold out period, I use a walk forward backtest. https://www.amibroker.com/guide/h_walkforward.html.

Unfortunately, there are some optimizations applied to speed up the process, so results in the backtest aren't entirely reproducible. For example, the effect of my bet size isn't accounted for in the final payoff. Using a relatively small bankroll and bet size partially accounts for this but still means real profits will be less than tested

[lansdale]

That is my hold out sample, though rather than having one single training period with one single hold out period, I use a walk forward backtest. https://www.amibroker.com/guide/h_walkforward.html.

Unfortunately, there are some optimizations applied to speed up the process, so results in the backtest aren't entirely reproducible. For example, the effect of my bet size isn't accounted for in the final payoff. Using a relatively small bankroll and bet size partially accounts for this but still means real profits will be less than tested

Since you're using what looks like financial models for this, I would assume you had some background in the areas of STEM or quantitative finance (neither of which I have), but what you've written is so confusing it suggests that you don't. It sounds like what you might be doing is using a machine learning/SVD program to create a handicapping model, which is fine. And you seem to realize that you shouldn't have started to simulate betting before you completed testing the model, which is true. But it also sounds like you don't completely understand the difference between a training (or in-sample) model and a hold-out (or out-of-sample) model. The hold-out model is 'always' less powerful (or has higher RSME) than the training (or optimization) model -- this is a given. The holdout model's results are the ones that count . So if you had really already validated your training model, you would be much more confident in your results. As a side comment your bet size or bank size are irrelevant.

I won't write more until I get a better idea of what you're doing, but I would say this much-- if the results you post are accurate, and the dataset does include ca. 3600 races, which works out to ca. 240 races per mo., your model has to be quite robust. In addition, although it was a mistake to start simulating betting, if you were using fully Kelly based on whatever edge your model was providing, that in itself can be considered a kind of validation -- the betting and handiicapping act as a check on each other. Put another way, if your model really sucked, full Kelly would have wiped out your bank in a flash.

I'm not averse to responding if you can elaborate further, but it would help if you could answer some of these questions. Last but not least, don't discuss the parameters, inputs, factors, variables (or whatever you want to call them) that you're using, with anyone, in case this model does turn out to be valid. I know people who have lost a lot of money that way.

[JerryBoyle]

Since you're using what looks like financial models for this, I would assume you had some background in the areas of STEM or quantitative finance (neither of which I have), but what you've written is so confusing it suggests that you don't. It sounds like what you might be doing is using a machine learning/SVD program to create a handicapping model, which is fine. And you seem to realize that you shouldn't have started to simulate betting before you completed testing the model, which is true. But it also sounds like you don't completely understand the difference between a training (or in-sample) model and a hold-out (or out-of-sample) model. The hold-out model is 'always' less powerful (or has higher RSME) than the training (or optimization) model -- this is a given. The holdout model's results are the ones that count . So if you had really already validated your training model, you would be much more confident in your results. As a side comment your bet size or bank size are irrelevant.

I won't write more until I get a better idea of what you're doing, but I would say this much-- if the results you post are accurate, and the dataset does include ca. 3600 races, which works out to ca. 240 races per mo., your model has to be quite robust. In addition, although it was a mistake to start simulating betting, if you were using fully Kelly based on whatever edge your model was providing, that in itself can be considered a kind of validation -- the betting and handiicapping act as a check on each other. Put another way, if your model really sucked, full Kelly would have wiped out your bank in a flash.

I'm not averse to responding if you can elaborate further, but it would help if you could answer some of these questions. Last but not least, don't discuss the parameters, inputs, factors, variables (or whatever you want to call them) that you're using, with anyone, in case this model does turn out to be valid. I know people who have lost a lot of money that way.

Apologies for the confusion. I will try to explain the design of the backtest I run a little more clearly. I believe I'm reasonably comfortable with the ideas of in sample and out of sample forecasting, so it might just be difference in applying terminology which is causing our confusion.

Let's say my data goes from 2016-01-01 to 2018-01-01, covering 2 full years and let's saying 10,000 races after all filtering is done. So our total dataset is 10k races. The most familiar backtest, and what I think you're alluding to, would be to fit a model on one subset, for example the first 2k races, and then run a betting simulation on the hold out sample of 8k. I agree that performance in the hold out sample (whether performance is measured via a betting simulation or some measurement of prediction error) will be worse than if we measured performance on the in sample data.

Rather than doing the above, I apply a walk forward backtest. This is done as follows. Given our sample space of 10k races, I take say the first 2k races and train my model on those 2k. I then run a betting simulation on a relatively small out of sample portion following the 2k (let's say 500 races from race 2001 to 2500). I stash the results from those 500. I then shift the 2k training races forward by 500 races, dropping the earliest 500 and adding the original out of sample 500, so the next training window is on races 500 to 2500. I then run a betting simulation on the next 500 races (in this case races 2501 to 3000). And so on and so on until I'm out of data. The result is that I end up training many models with many corresponding 500 count, non overlapping out of sample tests that I then stitch together to form the graph I attached earlier. The motivation for doing this is that I don't believe the model coefficients for horse racing remain stable throughout a year or over years.

*A note about the betting simulation I run on the out of sample races: I do use full kelly; however, I do not simulate reinvesting profits or losses after each race. Rather, at the start of each race I assume a $1000 bankroll. So the cumulative profit graph will never flatline, but if it dips far below $1000, I would have effectively gone broke.

[lansdale]

Apologies for the confusion. I will try to explain the design of the backtest I run a little more clearly. I believe I'm reasonably comfortable with the ideas of in sample and out of sample forecasting, so it might just be difference in applying terminology which is causing our confusion.

Let's say my data goes from 2016-01-01 to 2018-01-01, covering 2 full years and let's saying 10,000 races after all filtering is done. So our total dataset is 10k races. The most familiar backtest, and what I think you're alluding to, would be to fit a model on one subset, for example the first 2k races, and then run a betting simulation on the hold out sample of 8k. I agree that performance in the hold out sample (whether performance is measured via a betting simulation or some measurement of prediction error) will be worse than if we measured performance on the in sample data.

Rather than doing the above, I apply a walk forward backtest. This is done as follows. Given our sample space of 10k races, I take say the first 2k races and train my model on those 2k. I then run a betting simulation on a relatively small out of sample portion following the 2k (let's say 500 races from race 2001 to 2500). I stash the results from those 500. I then shift the 2k training races forward by 500 races, dropping the earliest 500 and adding the original out of sample 500, so the next training window is on races 500 to 2500. I then run a betting simulation on the next 500 races (in this case races 2501 to 3000). And so on and so on until I'm out of data. The result is that I end up training many models with many corresponding 500 count, non overlapping out of sample tests that I then stitch together to form the graph I attached earlier. The motivation for doing this is that I don't believe the model coefficients for horse racing remain stable throughout a year or over years.

*A note about the betting simulation I run on the out of sample races: I do use full kelly; however, I do not simulate reinvesting profits or losses after each race. Rather, at the start of each race I assume a $1000 bankroll. So the cumulative profit graph will never flatline, but if it dips far below $1000, I would have effectively gone broke.

Okay -- I see my first instinct was right -- you do have some background in probablility, statistics, sampling, etc., which I do not, aside from what I've picked up informally, as a blackjack player, from those far more knowledgeable. So, I'm somewhat mystified by what seems to be your skepticism about your model. I'm not clear how you arrived the initial parameter estimation you used to derive Kelly bets -- possibly there was no intermediate step testing vs. final results, and instead you began using the simulated bets as part of your out-of-sample model right away. Whatever you did, it's obviously working very well.

What's most amazing, and I'm not sure you mentioned this before, is that you did not re-invest in the bank, therefore no bet-resizing, as would be normal. Even so you made 4,0000% in a little over a year -- what if you had reinvested and re-sized? This is an issue with blackjack teams --whether to double the bet-size when the bank doubles or whether to halve it, given a diminished bank.

The training model/validation model split seems to be consistent with common practice, and your sample size is more than large enough. If you're looking for further testing I know some people who use the K-fold for cross-validation, but not for racing data -- but your take on this surely better than mine.

You are the first person I've ever seen mention using dynamic coefficients, which I understand are used to model turbulence, but possibly this is what's really making the difference for your model. People talk about racing as being complex -- dynamic coefficients may be a solution to this problem.

So, if I were you, I'd start betting this with money you can afford to lose. If it continues to go well and you start betting more serious coin, I would think about contacting some of the people - like Benter - who have done this successfully. As far as this site goes, sjk is a very smart guy who has built a successful model who might be responsive.

[JerryBoyle]

Okay -- I see my first instinct was right -- you do have some background in probablility, statistics, sampling, etc., which I do not, aside from what I've picked up informally, as a blackjack player, from those far more knowledgeable. So, I'm somewhat mystified by what seems to be your skepticism about your model. I'm not clear how you arrived the initial parameter estimation you used to derive Kelly bets -- possibly there was no intermediate step testing vs. final results, and instead you began using the simulated bets as part of your out-of-sample model right away. Whatever you did, it's obviously working very well.

What's most amazing, and I'm not sure you mentioned this before, is that you did not re-invest in the bank, therefore no bet-resizing, as would be normal. Even so you made 4,0000% in a little over a year -- what if you had reinvested and re-sized? This is an issue with blackjack teams --whether to double the bet-size when the bank doubles or whether to halve it, given a diminished bank.

The training model/validation model split seems to be consistent with common practice, and your sample size is more than large enough. If you're looking for further testing I know some people who use the K-fold for cross-validation, but not for racing data -- but your take on this surely better than mine.

You are the first person I've ever seen mention using dynamic coefficients, which I understand are used to model turbulence, but possibly this is what's really making the difference for your model. People talk about racing as being complex -- dynamic coefficients may be a solution to this problem.

So, if I were you, I'd start betting this with money you can afford to lose. If it continues to go well and you start betting more serious coin, I would think about contacting some of the people - like Benter - who have done this successfully. As far as this site goes, sjk is a very smart guy who has built a successful model who might be responsive.

Thanks for your thoughts landsdale. I've started betting it for very small amounts of money. So far so good, but I don't expect it to perform nearly as well as it has in the backtest.

Regarding the re investment. I suspect that the profit chart would exponentially increase until my wagers are restricted by pool size which would happen fairly quickly on smaller tracks

[lansdale]

Thanks for your thoughts landsdale. I've started betting it for very small amounts of money. So far so good, but I don't expect it to perform nearly as well as it has in the backtest.

Regarding the re investment. I suspect that the profit chart would exponentially increase until my wagers are restricted by pool size which would happen fairly quickly on smaller tracks


Cool. Re pool-size-limited betting, you may well already have an algorithm for this, but if not, you might want to take a look at Benter's comments on this subject, which will, hopefuly become increasingly relevant. He doesn't state the algorithm, but does refer to it in the endnotes to his well-known article.

BTW, just to correct a typo in my last post -- growth of bank per chart, ca. 4k% not 40k%. If you only get half of your model's projection going forward, which is common among those I know who do this, still would be very nice.

Best of luck.

lansdale

[mikesal57]

Generally, I do not play during the winter. Last November, just before Thanksgiving, I did a live play session (i.e. with an audience) and the racing was just horrible. My handicapping was not so hot, either.

I dedicated several weeks to figuring out just how different winter racing really is.

For over a decade I have used month-of-year in gathering races from the database to build a model of "races like this one." Specifically, +/- 2 months.

But this was different.

The first thing I determined was that racing is "different" from the week before Thanksgiving to around Jan. 21. The races are just far less predictable and/or do not match my handicapping approaches.

Ironically, many of the usual factor values - like recent speed ratings, for example - actually perform BETTER! But somehow the puzzle just doesn't fit together properly; the picture of the winner's circle just does not match the puzzle box.

I THINK... but have no proof of this... that it stems from the trainer patterns changing for the holidays. I THINK that the best trainers simply take time off from racing and hand the reigns over to assistants for about two months. This causes differences.

Very open to other ideas. BTW, it is not just weather because things change even in the warmer climates. (Not as much, but still different.)


Dave

I like to expand on this some more.....

I usually create a power number from the past year worth of data....
I make it with ALL tracks , not track specific.....
Someone asked me why dont I use the current last month or 2...I tested it with Gulfstream and found using the past year came out better....

My questions , Dave is:

Do you find the +/- 2 is better that whole year .....
or maybe +2/-1 ...?
track specific or all ?
class ?

I find that a handfull of factors usually prevail year after year , but the weights sometimes changes..

comments?

Thxs
Mike

[BCOURTNEY]

Thanks for your thoughts landsdale. I've started betting it for very small amounts of money. So far so good, but I don't expect it to perform nearly as well as it has in the backtest.

Regarding the re investment. I suspect that the profit chart would exponentially increase until my wagers are restricted by pool size which would happen fairly quickly on smaller tracks

.. next stop Hong Kong .. :ThmbUp: