M.D. KRAM SOFTWARE

[Jaguar]

Headhawg, I am familiar with Brainmaker and there is published data on the results achieved by individuals using that outfit's nets.

The results I saw didn't impress me. Those rather narrowly written templates are too limited in scope. For example, a handicapper using such a program selects the data he wants analyzed and then types it into rows and columns.

How many different data elements- or variables- is a guy going to type in for the net to make a model? A man could spend 3 weeks typing in data, hoping it's the significant criteria related to pp's, and then wind up with a rather limited model for a specific type of race at a certain track.

Meanwhile, all the trainer tricks and traps would have likely been left out, for simplicity and convenience.

Furthermore, our guy would wind up with a speed and pace disc- and when he's confronted with a new race, he would have to enter the data by hand.

I know quite a few computer handicappers, but I haven't seen anyone enter data by hand since MultiLine 3-D in the 1980's.

Moreover, the developers of dedicated horse racing neural net programs make their models using huge databases.

I know from the guys at PRI that it's not unusual for them to run 20,000 races to test an algorithm. RaceCom often tests data against 60,000 races on their server.

Basically, the problem- as I see it- for the individual handicapper, is the problem presented by vast amounts of data which needs to be sifted and analyzed really fast.

Fortunately, there are horse racing programs available, both neural net and expert systems disc, as well, which can accomplish this task for a reasonable amount of money.

All The Best,

Jaguar

[LOU M.]

Wasn't John Rancont involved with Winners Circle software. Have you tried this software. Or has anyone else for that matter? Haven"t heard to much about it. Price tag is pretty steep.

Thanks, Lou M.

[Derek2U]

Do you think MOST SoftWareSellers (SWS) are cheats? What %
are good guys who Cap shit etc etc ... Maybe the software code sucs or maybe the code is great but the info sucs?? If racing was
100% math or computing power it would have been solved a LONG time ago.... buts its the Horse not understood + shit data.

[Tom]

Originally posted by Derek2U
If racing was
100% math or computing power it would have been solved a LONG time ago.... buts its the Horse not understood + shit data.

What makes you think it hasn't been solved? <G>

[Jaguar]

Lou, John was a sincere guy who wanted to develop the best possible software.

In those days(early-mid-90's) everyone had accumulated some horse racing data-and there were some pretty good speed and pace programs on the market, most of which had very modest commercial success and were soon abandoned by the developers, who found that they could make more money driving a cab at night.

John counted on his loyal customer base for input and he was always tweaking algorithms and trying new ones. A stumbling block for him was that quite a few handicappers were reluctant to pitch in and help him upgrade the program, fearful that someone else would use their "excellent" insights and actually make money betting on horses.- John several times told me how frustrated he was.

As the big outfits- such as Frandsen and BRIS- appeared on the scene- it soon felt as though all of the oxygen had been sucked out of the room- as far as the little guys were concerned. Most of them quietly folded their tents.

This phenomenon was also reflected in the attrition in the ranks of
the Sartinistas, who have always been fanatically loyal to Howard's doctrine, but who have not been entirely able to resist the inevitability of progress, represented by the on-rushing tide of better handicapping methodology- such as, for example, an excellent program like Equisim, to name one example.

Handicapping has always been surrounded by a not-so-faint air of schizophrenia, the handicapper having a sense of community and wanting to contribute to the user group- while at the same time fearing that he will give away the farm and that others will cash on the same 6-5 shot that he is betting on.

John seemed to be an unhappy guy in his later years and I suspect that he wasn't a well man.

He would certainly have enjoyed seeing the brilliant modern handicapping programs available today.

All The Best,

Jaguar

[shanta]

Originally posted by Jaguar
Tom, beginning in 1986 I started buying horse handicapping programs, survived the Sartin wars and bought every program RaceCom and Frandsen issued.

In 1995 I worked part-time as a beta tester for 2 of the largest outfits and wound up getting interested in expert systems discs- and eventually in neural nets.

While I was doing some research for Happy Broadbent(he owns BRIS), and also for Frandsen, I had the opportunity to see some developing programs- some of which were eventually greatly improved. Others, such as Neurax and Multicaps, were marginally improved.

When Dave Schwartz produced his brilliant Thorobrain 4, around 1994, I realised that we were suddenly in a whole new world of handicapping power and accuracy. Dave used neural nets with great precision.

Around 1998, RaceCom, which had only produced one half-way decent handicapping program to date(most of the RaceCom products were unusable), made a tremendous leap forward. RaceCom began using modern neural nets and started making models on their server.

Today, I use RaceCom's Analog 2.0 because it's so fast and user-friendly- although it is out-of-date, an antique. However, I don't make models with it, I only use the speed and pace feature, which is superb.

I make my models and my betting decisions using Multi-Strats last issue, Titan.

While there have been dozens upon dozens of handicapping discs produced since the party started in 1978, there are only a handful of really good horse programs on the market today.

Unfortunately, there is no clearing house for software, and many handicappers are persuaded that speed and pace is the be-all and end-all of of horse racing. So, developers make money selling pace programs dressed up with a few features and the resulting models are often less than the best.

The really comprehensive horse programs make the best models.

All The Best,

Jaguar

Jaguar, if you feel that tb5 and nspace are without a doubt the 2 most powerfull programs today with their modeling capabilities why would you be using analog 2.0 which by your own words is an antique and multistrats titan which is a dos based program that is no longer available? the nspace probably is very expensive (don't know) but tb5 is available free of charge with a 3 month data subscription to hdw for 300 bucks. thats a hundred beans a month. a couple of solid bets on these "superb model plays" would certainly take care of the cost of that.i have seen 6 days of multi track readouts for tb5 and i think it is a very good program . i really would like to hear your reasoning on this if you would be willing to share. thanx,
Richard Pizzicara

p.s. btw i use sartins val2 which as i am sure you know is an ancient dos based program with no modeling or database capabilities.kind of out of its league with these Battleship programs right?

[Jaguar]

Richard, one of the hurdles I think all computer handicappers- as is true in the case of all TV, Computer, Automobile, and Stereo owners and users- is that as technology changes we are compelled to give up what we have come to know and enjoy, and are pushed and pulled toward new technology.

Sometimes, as in my case, where I have databases that have cost alot of time and money to develop, it's a major psychological leap to abandon the comfort and familiarity of models which are so reliable that when I do find a "profile" horse, I can bet with confidence.

For example, the speed and pace disc I use, Analog 2.0, has a clever earnings analysis feature, which indicates which type of bet to make for maximum profitability in a given race.

As well, the betting screen in Titan tells me not only that a horse is a key horse, but that the horse fits the model of a winner for that type of race, indicating his statistical chance of winning.

These models are invaluable to me and keep me off losers.

It's interesting that on PA some folks are focused on a given handicapping program's gross performance, total number of winners for a meet- for example.

To me that's not a meaningful indicator. I'm more interested in the effectiveness of a horse program's models.

The example I would use would be Equisim, a program which has become familiar to many because of its excellent website- which has an overall win rate of between probably 25-27%. So, a handicapper- looking at that stat might say, "unimpressive".

However, I look at the program's track-specific modeling feature, which Nathan keeps improving- by the way- which quickly and clearly indicates when to bet- and when to stay in your seat at OTB and call your girlfriend on your cell phone(use a "Tracfone" so that your wife doesn't see the bill on your regular cell phone).

All of a sudden, Equisim looks good, very good, by that standard, because horse racing is a "spot" game.

Once my insurance company and my lawyer get through vacuuming money out of my savings account because of an automotive mishap, as soon as I can catch my breath, I plan to jump into the latest version of "The Virtual Handicapper".

The ouput is sensational and- among other things- the program indicates when a horse has an edge, and how much of an edge.
Also, unlimited daily access to every track in America, no data fee, and no modeling(Yippee!) All the models are made on the server. I'll get my life back- and not a moment too soon.

As well, the seminars will be on-line- no more traveling. Now when I go to Vegas I'll be full-time in the casinos and not listening to some guy drone on about par times and about how he won a $500 exacta in 1982 by making an error when he bought his mutuel ticket.

Change and upheavel invade my quiet, comfy little handicapping world.

All The Best,

Jaguar

[Tom]

HTR has track/class/age/sex specific modelling. It has had modeling and export capability for years. When I used the old DOS sartin programs, I always modeled specific tracks/distance/age groups. MPH has that same capability.
Mutlicaps does too.
This isn't state of the art technology. It is commonsense.

[Jaguar]

Tom, the difference being that your posts- over time- reveal you to be an experienced computer handicapper.

In contrast, I got a pm from a lurker in California the other day asking "What's a "model" and how do you make them, what should go into them?"

For a while, I was getting 3 pm's and e-mails per week, and I figured it was now dying out. Turned out not to be the case. I am suddenly getting 5 or 6 pm's and e-mails per week from folks I never heard of, asking questions about horse racing software.

The only conclusion I can come to is that handicappers want information and because there is no clearing house for horse programs- let alone a publication which even prints honest reviews of software- these guys feel compelled to contact PA posters who yak about the subject.

There are a bunch of guys on this board, Dave Schwartz, Game Theory, Larry Hamilton, Nathan, and others, who are smarter than me and could probably give very insightful responses to software inquirers- and quite possibly they are doing so.

This entire software "information vacuum" is likely related to 2 things:

1. Alot of folks don't trust developer's websites as a source of information about the true capabilities of a software package.

2. Mail order catalogs, such as RPM, don't sell the better horse programs, and furthermore- the catalogs don't even review software, in order to give the subscriber a candid picture of what those horse programs can really do.

Hence, we have alot of lurkers on PA, and unfortunately- some of them are too intimidated to contribute.

Let me say to those guys, "Jump in, it's not so bad. You can make friends here. The more the merrier!"

All The Best,

Jaguar

[delayjf]

JAG,

As Tom pointed out HTR includes models in it's program. How are HTR's models different then those in HSH,Equisim, tb5 etc?? I assume that you feel they are not in the same league with the A.I. programs because you don't mention them in the same catagory as the A.I. programs

[Jaguar]

Delayjf, you mentioned strong, sophisticated handicapping programs, each one having unique insights into solving the handicapping puzzle. A guy can make nice money with each of these programs.

Among the modern expert systems discs, certainly for me Equisim stands out. Nathan has done such an intelligent job in his attack on the mutuels, and he is improving Equisim, even as we type these posts on PA.

Nathan's indicators for when to jump on a horse, and a certain race, are excellent. This is what happens when a smart man really focuses on the task at hand. -And what an enormous task. My hat is off to him.

The problem of statistical modeling is a complex one, because only certain information- discrete data elements- will be included, in spite of the fact that horse racing is an incredibly complicated activity.

Handicappers are confronted not only by the burden of correctly evaluating an animal's true performance potential, but are also faced with sorting out any patterns or "trainer moves" which can possibly be measured.

At the same time, figuring the weight- or statistical impact- of trainer angles on the total rating for a horse in today's race is really a tough proposition. One that bedevils public handicappers even today. For every Mott/Bailey stickout on a card, there are dozens of hard to figure trainer/jockey combinations.

So, the handicapper is looking at a race as an an "organic" phenomenon which must somehow be analyzed by being broken down into small pieces of information(speed, distance, class, trainer, jockey, track layout{Pimlico's tight turns which favor the 4 inside post positions}, weather/track condition, layoff's, workouts, etc., etc.)

Traditionally, software developer's have approached this problem by dedicating a screen to each major aspect of a horse's running record. Example being the first edition of All-Ways- in the early 90's- with a zillion screens.

Today, All-Ways is a much more integrated, more rational horse program.

Often, in the past, a handicapper looking at software output would be frustrated by seeing a trainer/jockey stat screen which was separate from the main handicapping screen. In other words, the trainer/jockey impact was not an integral component of the program's final rating.

Expert systems discs use algebraic linear regression to discover the statistical "weight", or impact value of different elements of data. This method is highly effective and is one reason why Equisim, and other well-written handicapping programs, are so effective.

For example, the program I use- Multi-Strats- is an expert systems approach, using 25 algorithms to evaluate the past performances of each horse in a race. By looking at only 25 criteria, imagine how much information is not examined.

-Yet, Multi-Strats produces some highly predictive race track models and the program is often able to correctly evaluate- and
to highlight, a horse which is "well meant for today"- that is, a horse which fits the profile of a winner for this type of race.

Then we come to the true Artificial Intelligence programs, the much-vaunted Neural Net handicapping programs.

The new neural nets are incredibly powerful and fast and have the advantage of not "over-training", which was sometimes a problem in the older NN methods.

Moreover, these handicapping programs are able to look at vast amounts of data and are able to sift that data and determine what types of information in a race horse's past performance history is really significant.

The effectiveness of the newer neural net handicapping programs really comes across when you benchmark the track models they make against the models made by less sophisticated horse programs.

Example, a race with a stickout horse and 2 other competitive runners, as well as 5 other entrants which are weaker animals.

I handicap the race using Multi-Strats, which indicates a standout horse, 2 lesser horses, and 5 other horses- any one of which- might jump out and run in the exacta.

I now handicap the same race using a powerful neural net program. This method indicates a stickout horse, a likely second-place finisher, a horse which may- at best- run 3rd, and a bunch of weak sisters which don't even belong in the race.

Summing up, the neural net had the exact order of finish, 1,2,3.

Whereas, Multi-Strats picked the winner, and one of the program's 2 other choices ran second, while the 3rd-place finisher didn't figure at all- according to the program.

It's the incredible precision and accuracy of the newer neural net methods which is so impressive. I believe that it points the way for the future of handicapping methodology.

All The Best,

Jaguar

[delayjf]

Jag,
Thanks for the indept answer. As I read through it, a few observations come to mind. 1) It seems to me that most of the analysis done using large databases looking at long term patterns. Can I assume then that short term patterns are missed??
2) If these programs really take off, everybody will have the same horse.

[Jaguar]

Delayjf, short term patterns are not over-looked because the algorithms give credit for recency- both in the case of a horse's racing finishes and also in the matter of trainer/jockey stats.

The reason the country's mutuels aren't being ground down to nothing(this has already occurred at Santa Anita, where there is a modest cadre of talented horses running against a whole bunch of "never-will-win" animals...which has been the case at Golden Gate Fields for years), is because there are relatively so few computer handicappers and because public handicappers cling to outmoded methods- and thank goodness they do so.

Very few people will spend $400 or more for a racing program and there are only a handful of horse programs which can show sales of 1,500 units.

Horse racing programs which sell for $1,000 or more hardly ever have more than 400 users in this country.

As well, most bettors bet on impulse and on newspaper picks. Also, many of the races which the unwary bet their hard-earned money on are unplayable "chaos" races, which should have been avoided.

The bettors want action. At the track or at OTB, they bet every race. As a consequence of making mis-informed wagers, the bettors soon fall behind.

Then, in a desperate effort to recoup their lost money, the public starts hitting every animal in the late races, skewing the pools, making true 4-1 shots 9-5 favorites, and creating overlays for the sharpies to scoop up.

I have many times seen late doubles at Aqueduct- which should have paid $18- pay as much as $30 or more, because of this "late race panic" phenomenon.

As long as the public continues to lose 66% of their bets, a stat which hasn't changed in 40 years, the hard-working horse handicapper will be able to pick up a car payment or a mortgage payment at the track or at OTB.

Woe unto us if the betting public ever smartens up. We'll be back at Burger King, saying "Would you like fries with that?"

All The Best,

Jaguar

[raybo]

Are you saying that you don't further manipulate the output of your handicapping software before you place a wager? Do you only require that the output fit the model? Do these software programs really figure current condition, or do they just look for particular, specific trainer/jockey moves as condition indicators? Do they compare all the relevant numbers in the horse's PP's in order to ascertain his relative position regarding his form cycle? Do they select a particular PP to use for grading purposes or do you have to tell it what to do?

My opinion is that, although there may well be some worthy software out there that learns from itself, that is very powerful and thrashes thousands of race cards very quickly, they will never replace the human brain as an analytical tool. If you know of a particular piece of software that can tell me both a horse's capability and also how far from that capability his current condition has limited him to, please let me know, so I can get out of the handicapping/wagering game, because my edge over the betting public will be gone.

[Tom]

I am riding a positive betting cycle right now- 34% winners with a postive roi and I can honeslty say that of the last 50 winners I have bet, I have absolutley no idea the trainer or jockey were. Never looked, never cared. the horse does the running, I have his history laid out in front of me in a multitude of screens, and that is all I need.
Not to say I don't look at trainer/jockeys sometines, but right now I am not and right now I am winning. My worst cycle in a long time was this winter when I was really following trainer patterns over pace/speed/form.
If you glean the truth from trainer moves, great--cash in on it.
But that is not the only doorway to profits.