Short-term vs long-term predictor models?

[jackad]

Should predictor models best be based on short-term (the most recent of one to two weeks of data) or long-term data (one month's to one year's data)? Why?

Jack

[ranchwest]

I'd say there is valuable information in both groups. Certainly hot jockeys, post positions, trainers and such can be very short-term.

I mostly go with long-term data, but I divide it out by years as well as log it combined so that if I discern a change in tendancies I can always roll out the older data.

[Jaguar]

Jackad, if a person takes the 24 best angles in horse racing(Dan Pope, among others, has compiled these) or, if he took the impact values Nunamaker rates in his definitive study, that handicapper would have an extremely powerful model, or template, that would apply to every racetrack in the country.

But, having said that, the bettor is only going to play those races where the model's numbers say:"Go for it!"

In other words, only when the algorithms find a consistent animal with great connections, and when the horse's speed- and form cycle- meet the criteria within the model, can we bet with confidence.

Interestingly, the sudden introduction of powerful and illegal stimulants- on a massive scale, which began 5 years ago, in the racing world, did not destroy the classic formulas which search for owner-trainer-jockey patterns.

The old algorithms held up, and they highlight the cheats, once the handicapping program discovers patterns at a track.

We can still win by picking our spots. Drugs did not destroy racing, but they made it things a bit tougher for the serious handicapper.

All the best,

Jaguar

[Rick]

Good handicapping models seem to work for years but you have to be aware of major changes in the popularity of various handicapping factors and availablity of information. Speed, class, form, pace and other factors have all had their popular periods. Pace is currently very popular, especially second call figures. I think form cycle analysis is the next trend. Good speed figures are easy to come by now and even the best don't have much value because they're so widely available. If DRF ever publishes pace figures, it'll probably kill the prices on pace horses altogether.

Another thing you have to be aware of is that many tracks make drastic changes to their surface between meets, so last year's pace model may not work this year. Most tracks seem to be going to a "fair" track these days which is pretty consistent but still favors early speed a little. My favorite tracks are those with very long seasons. If they don't have any time to dig them up, they can't change the characteristics very much.

I haven't found very short-term models (days to weeks) to be useful at all for my handicapping but I do know of some jockey/trainer factors that are most effective if measured over several months so I'm sure others will disagree. I've read a lot about short term track bias, but I've had exactly zero success in identifying anything like that. In fact, if anything, I'd say that yesterday's bias is more likely to disappear today that to stay the same. I attribute that to track maintenance actively working to eliminate any bias.

[Jaguar]

One thing I've learned by using the same 25 algorithms- or "strategies" since 1988-89- is that there are certain patterns which are valuable forecasting tools.

But, interestingly, the computer program I use has settled- in recent years- on models which don't change. They changed after the drugs swept in like a cyclone in the late 90's, but the models now appear to be set in cement. They have pointed out to me that the old form-cycle charts and the old layoff algorithms are obsolete, among other old handicapping shibboleths- which are also obsolete.

Curiously, in the last 5 years, the models dumped 7 out of the 25 algorithms the program employs. So the new models are made without those "dead 7".

The new models stress a horse's connection's patterns and his speed, primarily.

It seems the program is now focused on the owner-trainer-jockey moves as well as speed and pace. But I should add that pace has been downgraded in its statistical impact on the winning model.

Watching all this happen has been disconcerting to say the least.
Alot of traditional handicapping verities have turned to dust and been blown away by the sands of time.

It's almost a whole new world out there.

I might almost say that in the future we will be handicapping based only on 3 criteria: connections, speed, and consistency.

All the best,

Henry

[Rick]

Jaguar,

What is the best way to measure consistency according to your studies?

Your observations regarding various handicapping factors are pretty much consistent with my experience. The last major change I made involved connections and it surprised me how much it improved the ROI. But it surprised me more to find that what worked was considerably different that what I expected.

[GR1@HTR]

Looks like he was nice enuf to send me a personal letter. He says he can show me how to get 95% winners from my top 2 picks. I wonder if this has anything to do with waiting for a 2 horse match race...? He has peaked my curiosity and I shall sell my dog and send him $40 so I can earn $15,000/month income as he as stated. Does it get any better than this????

[ranchwest]

GR1,

Yeah, right, sell your dog. Uh huh.

(not gonna happen, folks) :)

[Jaguar]

Rick, you might enjoy Nunamaker's book, the definitive work on the subject- derived from a huge database and brilliantly done.

Also, any good A.I program will make strong models, tell you the percentages, etc. All-Ways, or TB5 would be an excellent choice.

Prof. Evans did great work on impact values, in case you run across a copy of his work, but the values- particularly layoff and pace values- have changed somewhat in recent years.

All the best,

Jaguar

[Rick]

Jaguar,

Yeah, I have Nunamaker's book but all that really tells me is that consistency measured the traditional way doesn't have much value now (or in 1994). I remember a time when simple in-the-money consistency was very profitable. I knew a guy from Kentucky who made his living on a method based on that for about a year. I wouldn't really call him a pro though because he eventually had to go to work painting houses because the method fell apart. I swear I checked out his method on over 600 races and it showed more than 20% profit, but it just didn't hold up. These things happen sometimes. Go figure!

[jackad]

Jaguar,
I believe you mentioned that you only bet on a horse if your program tells you that horse has a 77% chance of winning. I'm curious as to the frequency with which you find such horses, and their average final odds. It would seem that their odds would be pretty low, so that a losing streak might do considerable damage to your bankroll.
Thanks.
Jack

Anyone else: Do you require any minimum probability of winning before betting on your top selection? If so, what is it?
Thanks.

[ranchwest]

The percentage going in should be reflected in the percentage coming out. 77% is a mighty high percentage.

I could hit for a much higher percentage than I do, but, for me, the higher the percentage the less the action and the lower the payoffs. I prefer catching a longshot every now and then.

[so.cal.fan]

Anyone else: Do you require any minimum probability of winning before betting on your top selection? If so, what is it?
Thanks. Jackad.

That's a great question. I thought about it, and while I do not make my own morning line, I think I would have to think my horse has a 50% or better chance to win.
Having said that, I will take a chance (bet the come line) on an unknown chance, I guess if I have a hunch on it being live, but only if it is a really good price.....like better than 7/1.

[Rick]

I bet on my top choice if it is NOT the morning line favorite and meets some other guidlines designed to increase both it's win % and overlay potential. I don't need to look at actual odds at all because it seems to offer very little improvement in my ROI and requires a lot more effort.

[ranchwest]

Originally posted by jackad
Anyone else: Do you require any minimum probability of winning before betting on your top selection?

I keep track of my winning percentage and my "balance". If those numbers remain satisfactory, I don't need anything else. I play my choice. Sometimes my choice is odds on, but I recently hit one at over 40-1. Most of the time I don't even have access to an odds board. It works for me.

[Rick]

ranchwest,

That approach seems to work fine as long as you're good at identifying overlays. If the horse is bet down more than expected it just wins at a higher rate. I had an $81.80 winner last week that won by 5 lengths though so sometimes you get more than you expect. I think more of my horses go down in odds than up but I don't have any exact numbers. You've got to pick winners that the other guy doesn't pick and public handicappers usually reflect the current popular handicapping methods used by the crowd. Or is it the other way around? Has there ever been a race run where there was no published handicapper's selection or morning line odds? Maybe in an Arab country. You could probably make a lot of money before you were beheaded.

[so.cal.fan]

"I bet on my top choice if it is NOT the morning line favorite "-Rick.

This is usually my guide as well, Rick. It would seem if we beat the obvious favorite, provided we think we CAN beat it, would ASSURE us an OVERLAY?

[Aussieplayer]

It would seem that way SCF. The way I figure it, if you eliminate enough percentage points, random betting should provide a profit. Or, just betting the second fav. should provide the highest hit rate.
But....(I don't think) we should assume that this is a task as easy as "A,B,C." You have to be correct at eliminating favs. so that these favs. return at a very poor hit rate & ROI. Poor ROI fav. situations can still return a high hit rate.
And....I'm not 100% certain that this is as easy as it sounds.

Cheers
AP

[ranchwest]

Originally posted by Aussieplayer
It would seem that way SCF. The way I figure it, if you eliminate enough percentage points, random betting should provide a profit. Or, just betting the second fav. should provide the highest hit rate.
But....(I don't think) we should assume that this is a task as easy as "A,B,C." You have to be correct at eliminating favs. so that these favs. return at a very poor hit rate & ROI. Poor ROI fav. situations can still return a high hit rate.
And....I'm not 100% certain that this is as easy as it sounds.

Cheers
AP

Races with poor ROI high percentage favorites are sometimes pass races. It is too difficult to win with or without them. You need to have a horse with enough positives to think that it can actually win, not just a shot in the dark, to beat those types.

[so.cal.fan]

You are absolutely right, Ranchwest.
If you check out the favorite percentages in the free section of the todaysracingdigest.com, you will see some races where only 15% of the favorites win, and they go off a LONG database there.
These races are not only hard for everyone to select a winner, they are hard for experts.
This poses an interesting question.
Is is better to beat a heavy favorite in a race type that has a very high percentage, or try to select a winner in a type of race that is tough for EVERYONE to pick?

[ranchwest]

Originally posted by so.cal.fan
This poses an interesting question.
Is is better to beat a heavy favorite in a race type that has a very high percentage, or try to select a winner in a type of race that is tough for EVERYONE to pick?

IMHO, neither. The high percentage money burner leaves you starting at low percentage, not a good situation. The race that is tough for everyone is, well, tough for everyone. I try to pass both types and concentrate on the types that I have had success with in the past.

A few days ago, I had a good one that paid around $15 and beat an even money favorite. Let's call those two A & B, I had an exacta with A,B/A,B,C and win and place on the $15 horse. They came A/B and the exacta paid almost $50. Nice profit on those type propositions, a lot of heavy favorites finish second.

[Rick]

My experience is that closely contested races (according to my ratings) are the most profitable. In fact, those are the best races to bet more than one horse to win.

The reason I go against morning line favorites is that they're usually (but not always) overbet. It's a simple (lazy man's) way to eliminate a big group of losers. Eventually I plan to use alternate handicapping methods to find overlays in those races. Right now, if I wanted to have action in every race, I could play exactas and pick threes including those horses. Also, I think my second choice to place or in a backwheeled exacta would do pretty well.

[ranchwest]

Originally posted by Rick
My experience is that closely contested races (according to my ratings) are the most profitable. In fact, those are the best races to bet more than one horse to win.

That certainly is one approach. If I'm understanding, you anticipate a deep stretch situation of five horses across and you go for the one(s) that will pay enough to be profitable. I don't like those races.

I prefer looking for an advantage horse and going with it. If I can find one with maybe a little back class and/or maybe a disguised positive pattern, then I might catch a winner and sometimes at a nice price.

[Rick]

ranchwest,

Well, not really. Most races don't really turn out as close as you predict them to be, especially if you emphasize early speed. Say there are two closely matched horses with good early speed and class. Usually one of the two will win and nobody else has much of a chance. And the odds are better because of the confusion.

[ranchwest]

Rick,

Sounds like our bottom line is very similar, though I would guess that our approach to getting there is probably different.

[Rick]

ranchwest,

Yeah, it's fascinating how many different ways there are to win. What's true for one person may be completely false for another. For example, my current method is very good at exactas. The method I was playing in the 80's, however, was positively awful at them. The top rated horses either won or ran out. That's pretty common with methods that have a heavy emphasis on early speed. As little as five years ago I probably wouldn't have believed things that I now know to be true because I hadn't experienced it before myself.

[cj]

Originally posted by jackad
Jaguar,
Anyone else: Do you require any minimum probability of winning before betting on your top selection? If so, what is it?
Thanks.

I require the horse to have 2 times his natural percentage to be considered a play. So if a horse is in a 10 horse field, his natural % is 10, I require 20.

CJ

[Rick]

CJ,

An interesting approach that takes into account field size, which is important to consider. Another way I like is to compare the rank of your selection according to your method relative to its rank in actual odds. Two ranks better should be an overlay if your method is a good one. It might sound less accurate than calculating odds and comparing them, but in practice it works better and is easier at the same time.

[ranchwest]

Originally posted by cjmilkowski


I require the horse to have 2 times his natural percentage to be considered a play. So if a horse is in a 10 horse field, his natural % is 10, I require 20.

CJ

Do you compare or adjust your expectation based on historical results?

[Triple Trio]

Originally posted by Jaguar
Jackad, if a person takes the 24 best angles in horse racing(Dan Pope, among others, has compiled these) or, if he took the impact values Nunamaker rates in his definitive study, that handicapper would have an extremely powerful model, or template, that would apply to every racetrack in the country.

Pardon my ignorance but I've never heard of Dan Pope. Can you give me a reference where I can read about the 24 angles?

[Jaguar]

CJMILKOWSKI,

in order for me to be excited enough about a horse's prospects to bet on it, the computer program I use has to indicate that the race's stability- or integrity, if you will- is at least 77%.

This means that:

1.The horses entered have measurable past performance lines, and that the entries have raced enough to present a reasonable picture of their athletic ability.

2. In addition to the race's statistical integrity, the best horse must be highly rated by the "connections model". This indicates that the trainer really brought his horse up close to the peak of the horse's form cycle, and that the trainer is going for the tag today, not just stabbing and hoping.

3. The computer's top pick must show a significant speed advantage(at 6f I am very partial to animals which are projected to finish out at 57fps, or thereabouts). A horse which has the stamina to finish at that speed is going to take some catching, particularly if the jockeys on the other mounts went fast early- in which case they'll be dueling for 2nd or 3rd at the end.

4. The best horse must fit the model. Old Multi-Strats users refer to such an entrant as a "profile" horse. That is, the horse has the statistical profile of a winner, according to the 18 strategies- or criteria(algorithms built in to the program which measure a horse's racing performance, connections, class level, form cycle, etc.){ I have to mention here that the program uses 25 strategies out of the box, left unadjusted, but that the model only uses 18 out of the 25 because after analyzing thousands of races, 7 strategies were revealed to be less effective than the remaining 18 strategies}.

5. The computer's top pick must be even money- or above- at post time.

Then, once all is said and done, and all the horses are in the gate,
you have to be lucky. The race has to unfold in the tactics, speed, pace, and performance paradigm which the model projected it would.- But then, that's the uncertainty which blends spice into the horse racing bouillabase, and makes us thrilled when our horse measures up and wins.

Can there be any greater thrill than cashing a ticket on a winning horse? I don't know, but Norman Dewis- old Jaguar hand- relates that at Le Mans, which he won in a D-Type Jaguar- he passed the Mercedes 300SLR in the Mulsanne straight at night at over 172 miles an hour. He noted that the mighty Mercedes engine was blasting flames from it's short, side-mounted exhaust pipes.

That would be something to tell the girlfriend about.

All the best,

Jaguar

[aaron]

Hi, Jaguar-Your computer program sounds interesting.Which one are you using?I've used computer generated numbers,but I have generally put too much emphasis on my own trainer patterns and in some cases disregarded the numbers.Your program seems to take all this into consideration.

[cj]

Originally posted by ranchwest


Do you compare or adjust your expectation based on historical results?

RW,

I keep records on all the betting lines I make, and I've found that horses who are overlays on my line, but do not meet the criteria I posted earlier, perform poorly ROI and win% wise. Of course, these horses are almost always the longest prices anyway, so I would think that explains it.

CJ

[ranchwest]

Originally posted by cjmilkowski


RW,

I keep records on all the betting lines I make, and I've found that horses who are overlays on my line, but do not meet the criteria I posted earlier, perform poorly ROI and win% wise. Of course, these horses are almost always the longest prices anyway, so I would think that explains it.

CJ

I understand that you know it works. I can tell by your posts that you are probably meeting with success.

My question, one of curiosity really, is whether you compare your 20% anticipation to actual win percentages.

[Jaguar]

Aaron,

I use an awkward old program which is outmoded, called Multi-Strats. You might want to consider a state-of-the-art program such as: Fastcapper, All-Ways, Thorobrain5, Joe Zambuto's new program, Handicapper Magic(when the upgrade is issued), or, if you have alot of money- RaceCom.

Be sure to go with a disc which either makes models or has models- or patterns- built in. You want to know exactly when to commit your hard-earned money, and when to step back.

There's no rush to bet, plenty of good races come up almost every day.

All the best,

Jaguar

[cj]

Originally posted by ranchwest


I understand that you know it works. I can tell by your posts that you are probably meeting with success.

My question, one of curiosity really, is whether you compare your 20% anticipation to actual win percentages.

RW,

I do not specifically track certain ranges as you are asking. I probably would find the lines I make aren't as accurate as I would like, but the errors probably even out and it works for me.

I track things I think are unique to my methods. It mostly involves running styles and pace figures and helps me to eliminate many short priced losers. I play mostly races where I can confidently eliminate one or two short price horses, almost guarenteeing overlays on the horses I bet. The horses I bet against lose upwards of 80% of the time and are always the 1st or 2nd choice.

Hope this explains!

CJ

[ranchwest]

Yes, thanks.