Are certain computer models now useless

[TrifectaMike]

Has anyone come to the conclusion that computer models described by Bolton and Benter have been rendered usless, because of the increased psuedo R2 for the public model?

What do I mean by this. Is it now nearly impossible to get a postive delta R2 between the combined public and fundamental vs the public model.

Mike

[DeltaLover]

Has anyone come to the conclusion that computer models described by Bolton and Benter have been rendered usless, because of the increased psuedo R2 for the public model?

What do I mean by this. Is it now nearly impossible to get a postive delta R2 between the combined public and fundamental vs the public model.

Mike

In theory, somebody can always claim that he has identified a model resulting to a positive delta R2 although without providing the related evidence he cannot convince us about the validity of his statement.

My own research has convinced me that something like this is either impossible or that it requires more data than what I currently have access to,

Of course I need your confirmation since I am not even close to you when it comes to statistical analysis, but I believe that one of the major holes of Bolton / Benter approach, lays in the assumption that the final effort of a starter follows the normal distribution.

In contrary, the more I am dealing with the problem, the closer I am getting to the conclusion, that as you have repeated many times before, Bayesian networks is the most accurate approach to the betting game...

[sjk]

I don't know how to calculate the R2 for a line so I cannot answer the question directly. It is not necessary to have a line better than the public if it is nearly as accurate and different.

The process of making a line and betting overlays is still viable although there seem far fewer profitable opportunities than 10-15 years ago. There are fewer overlay opportunities because of all the smart whale money and the short fields.

It is also not as interesting and action packed as in years past. I used to bet 60-80 races on a weekend and I see only 29 bets last weekend. Meanwhile win percentage has gone down because the low and mid-priced horses are most often fairly priced.

It is not as much fun as it used to be and if making a profit were significant to my lifestyle the trends would give me reason to be concerned.

[Magister Ludi]

Has anyone come to the conclusion that computer models described by Bolton and Benter have been rendered usless, because of the increased psuedo R2 for the public model?

What do I mean by this. Is it now nearly impossible to get a postive delta R2 between the combined public and fundamental vs the public model.

Mike

Public pseudo R^2 in Hong Kong 1988-1993 was .1237, according to Mr. Benter. Now it's about .18.

[TrifectaMike]

Public pseudo R^2 in Hong Kong 1988-1993 was .1237, according to Mr. Benter. Now it's about .18.

Using log odds, that is very similar to NA tracks(major and minor) of approximately .17

Mr. Benter is honest and speaks the truth.

Mike

[Ray2000]

If interested in seeing the math

CXWong in his book "Precision and Mathematical Methods in Horse Racing" shows the method used for calculating a 'pseudo' R Squared number.

He gives years:
96-01....0.1237
03-04....0.1673
05-08....0.1741

The method in Excel spreadsheet form is at

http://maxabet.com/precision/ file...5.4

[classhandicapper]

As I suggested in the other thread, IMO, the route to profits is specialization. If you insist on building a model, then the model should at least be different for each type of race because the dominant factors are different depending on the type of race.

If you are playing 2yo maiden races at Saratoga you should probably be looking at the trainer's record with 2yos (especially first and second time starters at Saratoga) sales information, sales workouts, pedigrees, recent workouts, comments from people that have seen horses work in company etc...

If you are handicapping stakes, you should be looking at an entirely different set of information.

As a non model player, the approach I like to take is to isolate "situations" where I know the public often makes mistakes. That typically involves having information much of the public does not have or where the public tends to evaluate the same information poorly. That means I am betting a small subset of races because IMO most races are bet fairly efficiently.

[Robert Goren]

Has anyone come to the conclusion that computer models described by Bolton and Benter have been rendered usless, because of the increased psuedo R2 for the public model?

What do I mean by this. Is it now nearly impossible to get a postive delta R2 between the combined public and fundamental vs the public model.

MikeI think the answer to your question is yes. The home PC and the advent of whale betting put an end to the profits of these models. It still may be a possible niche when looking at certain "types" of runners. I am thinking maybe with runners that would generally considered the third or four most likely winners by examining the PPs. There may be increase in the "psuedo R2" that is large enough where bets on some those types of runners are profitable. When I get back to my home PC, I will resume looking into that possiblity and report it on this site if it is. I think the days of looking at a race as whole and expecting to find a decent "psuedo R2" ended 15-20 years ago.

[Cratos]

Has anyone come to the conclusion that computer models described by Bolton and Benter have been rendered usless, because of the increased psuedo R2 for the public model?

What do I mean by this. Is it now nearly impossible to get a postive delta R2 between the combined public and fundamental vs the public model.

Mike
First of I like to compliment you on the wording of your question with the use of the word
' certain.'. By qualifying it that way my answer would be yes.

How coefficient of determinants are still is play and it is my understanding that Benter model has gone from 16 variables to 120 which says modeling is evolving to the next level.

[TrifectaMike]

Public pseudo R^2 in Hong Kong 1988-1993 was .1237, according to Mr. Benter. Now it's about .18.

Magister, the higher the public pseudo R2 the better for me. I can always do better. I can do it with six covariates. And none of the covariates are handicapping factors. I know it sounds impossible, but it is true.

Mike

[Capper Al]

As I suggested in the other thread, IMO, the route to profits is specialization. If you insist on building a model, then the model should at least be different for each type of race because the dominant factors are different depending on the type of race.

If you are playing 2yo maiden races at Saratoga you should probably be looking at the trainer's record with 2yos (especially first and second time starters at Saratoga) sales information, sales workouts, pedigrees, recent workouts, comments from people that have seen horses work in company etc...

If you are handicapping stakes, you should be looking at an entirely different set of information.

As a non model player, the approach I like to take is to isolate "situations" where I know the public often makes mistakes. That typically involves having information much of the public does not have or where the public tends to evaluate the same information poorly. That means I am betting a small subset of races because IMO most races are bet fairly efficiently.

:ThmbUp:

[barn32]

What is R2?

[Dark Target]

Magister, the higher the public pseudo R2 the better for me. I can always do better. I can do it with six covariates. And none of the covariates are handicapping factors. I know it sounds impossible, but it is true.

Mike

How exactly do you mean they are not handicapping factors, if they assist in obtaining probabilities more accurate than the public then they are handicapping factors, are they not?

[Capper Al]

http://en.m.wikipedia.org/wiki/Coefficient_of_determinationQUOTE=barn32]What is R2?[/QUOTE]

I'm assuming it is the above. In easy to understand terms, it's how well the points on a graph map on a line. The higher the number (correlation) to 100% the better the fit.

[DeltaLover]

http://en.m.wikipedia.org/wiki/Coefficient_of_determinationQUOTE=barn32]What is R2?



link does not work

[Cratos]

What is R2?
A symbolic way of stating curve fitting

[Cratos]

Magister, the higher the public pseudo R2 the better for me. I can always do better. I can do it with six covariates. And none of the covariates are handicapping factors. I know it sounds impossible, but it is true.

Mike

Mike ,
Your answer appears to be a contradiction or maybe I am misunderstanding it because I would. think that the covariants would be handicapping factors; please elaborate in layman's terms.

[Capper Al]

R2 link (http://en.m.wikipedia.org/wiki/Coefficient_of_determination)

If this doesn't work again go to wikipedia and search on r2 coefficient of determination.

[Magister Ludi]

Magister, the higher the public pseudo R2 the better for me. I can always do better. I can do it with six covariates. And none of the covariates are handicapping factors. I know it sounds impossible, but it is true.

Mike

If you would have stated that the six covariates were handicapping factors, I would have found it hard to believe. Certain covariates are as useless as certain computer models. It takes a creative intellect such as yours to succeed in this business.

[GameTheory]

If you would have stated that the six covariates were handicapping factors, I would have found it hard to believe. Certain covariates are as useless as certain computer models. It takes a creative intellect such as yours to succeed in this business.Yes, but what one wonders is if they are not handicapping factors, then what could they possibly be? And if they create a model with which one uses to make profitable bets, then they are handicapping factors by definition. It makes no sense, you see.

When he says they are not handicapping factors, it sounds like he is saying he rolls dice or takes his dog's temperature or some other nonsense to determine his bets, i.e. things that to other than the supreme being maybe have no bearing whatsoever on the results of horse races...

[Magister Ludi]

Yes, but what one wonders is if they are not handicapping factors, then what could they possibly be? And if they create a model with which one uses to make profitable bets, then they are handicapping factors by definition. It makes no sense, you see.

When he says they are not handicapping factors, it sounds like he is saying he rolls dice or takes his dog's temperature or some other nonsense to determine his bets, i.e. things that to other than the supreme being maybe have no bearing whatsoever on the results of horse races...

I cannot speak for Trifecta Mike but his reference to "handicapping factors" I interpreted to mean factors which directly measure the performance of a horse as opposed to factors which measure the performance of the horses' handicappers.

[Capper Al]

I cannot speak for Trifecta Mike but his reference to "handicapping factors" I interpreted to mean factors which directly measure the performance of a horse as opposed to factors which measure the performance of the horses' handicappers.

I agree. It would be nice if he spoke up. But he did pretty much say this in another thread.

[Cratos]

I cannot speak for Trifecta Mike but his reference to "handicapping factors" I interpreted to mean factors which directly measure the performance of a horse as opposed to factors which measure the performance of the horses' handicappers.


Magister Ludi, what you have added is another layer of ambiguity with your response; I agree with Game Theory in his post #20

[Jeff P]

Educated guess (never having seen his model.)

He's likely referring to metadata (http://en.wikipedia.org/wiki/Metadata) ... or data about the data.

In my opinion, metadata about starters pointed out by your model is worth knowing during live play.


-jp

.

[Sapio]

I cannot speak for Trifecta Mike but his reference to "handicapping factors" I interpreted to mean factors which directly measure the performance of a horse as opposed to factors which measure the performance of the horses' handicappers.

Hi Magister Ludi

As someone who has recently read most of TM's posts, I agree with you. In several of his posts he alludes to modeling the performance of the horses' handicappers.

Thomas Sapio

[Sapio]

Educated guess (never having seen his model.)

He's likely referring to metadata (http://en.wikipedia.org/wiki/Metadata) ... or data about the data.

In my opinion, metadata about starters pointed out by your model is worth knowing during live play.


-jp

.

Hi Jeff P

There are two things which I have found to be consistent within a majority of TM's posts:

1. His Bayes based posts are all relevant to horse racing, although not well recieved by many board members (possibly due to the level of math involved)

2. In several of his posts he alludes to understanding the "data on the data".

Thomas Sapio

[clocker7]

Link

http://en.m.wikipedia.org/wiki/Coefficient_of_determination

(Edited: nevermind, Al provided the link above)

[dnlgfnk]

From Jeff's link...

..."the main purpose of metadata is to facilitate the discovery of relevant info"...

What's the relevant info? The same for the high-tech modeler as it is for the creative, intuitive player...In a typical 8 horse N.A. field for instance, what manner of looking at the same data differently than the public allows me to confidently select from the solid 2nd to 4th ranked horses who win 50% of the time, while sprinkling in the interesting longshot(s) from the bottom 4 rankings, who collectively win as often as the favorite, while defending the exotic play by using the favorite when deemed necessary, perhaps by emphasizing the latter combinations?

Aren't we all "handicapping the handicappers" by this time?

[raybo]

From Jeff's link...

..."the main purpose of metadata is to facilitate the discovery of relevant info"...

What's the relevant info? The same for the high-tech modeler as it is for the creative, intuitive player...In a typical 8 horse N.A. field for instance, what manner of looking at the same data differently than the public allows me to confidently select from the solid 2nd to 4th ranked horses who win 50% of the time, while sprinkling in the interesting longshot(s) from the bottom 4 rankings, who collectively win as often as the favorite, while defending the exotic play by using the favorite when deemed necessary, perhaps by emphasizing the latter combinations?

Aren't we all "handicapping the handicappers" by this time?

If our goal is to bet value, looking at the long term, we logically must be handicapping the consensus of the players in the pool. If we deem the consensus of players in the pool are correct, then we either accept the odds offered or attempt to leverage racing's variance by demanding higher odds on horses less likely. If we deem that the consensus of players is incorrect, then our decision is much easier to make, we simply bet the most likely.

Either way, we are handicapping the public consensus. We either are good enough at this analysis, or we are not, long term.

[TrifectaMike]

When he says they are not handicapping factors, it sounds like he is saying he rolls dice or takes his dog's temperature or some other nonsense to determine his bets, i.e. things that to other than the supreme being maybe have no bearing whatsoever on the results of horse races...

Actually the number of chickens exported to the EU countries on a daily basis... I expected better

Mike

[TrifectaMike]

I cannot speak for Trifecta Mike but his reference to "handicapping factors" I interpreted to mean factors which directly measure the performance of a horse as opposed to factors which measure the performance of the horses' handicappers.

Insightful

Mike

[TrifectaMike]

Educated guess (never having seen his model.)

He's likely referring to metadata (http://en.wikipedia.org/wiki/Metadata) ... or data about the data.

In my opinion, metadata about starters pointed out by your model is worth knowing during live play.


-jp

.

Clever

Mike

[Capper Al]

Link

http://en.m.wikipedia.org/wiki/Coefficient_of_determination

(Edited: nevermind, Al provided the link above)

Scary Clocker. Look at post 18.

[whodoyoulike]

Actually the number of chickens exported to the EU countries on a daily basis... I expected better

Mike

I don't get it.

I think GameTheory was responding to your post in #10.

[flatstats]

Maybe they are useless because they have not adapted to changes.

In the UK many swore by Adrian Massey's ratings. Those ratings ultimately failed though and many said this was because 'they became overbet', 'the bookies knew about it' and other such guff.

IMO Massey failed because his ratings did not adapt to changes.

There were many changes in British racing over the past few years. Some were subtle, like changes in race conditions and some were major like the renumbering of post positions for right handed courses. Other changes included changes to whip usage, race classes, watering practices, even things like changing different AW surfaces.

The point is that if you don't adapt to these apparent changes then you are at the mercy of randomness and in racing randomness is what you don't want to be the mercy of.

[raybo]

Maybe they are useless because they have not adapted to changes.

In the UK many swore by Adrian Massey's ratings. Those ratings ultimately failed though and many said this was because 'they became overbet', 'the bookies knew about it' and other such guff.

IMO Massey failed because his ratings did not adapt to changes.

There were many changes in British racing over the past few years. Some were subtle, like changes in race conditions and some were major like the renumbering of post positions for right handed courses. Other changes included changes to whip usage, race classes, watering practices, even things like changing different AW surfaces.

The point is that if you don't adapt to these apparent changes then you are at the mercy of randomness and in racing randomness is what you don't want to be the mercy of.

Well done! This is probably the main argument for smaller, more recent databases, versus larger, older databases. If one intends for the database (or method) to evolve over time, a larger/older database will take much longer to significantly reflect relatively recent changes, in whatever form they may take. There is probably an optimum sample size, and that is, IMO, where the majority of our database activities should be concentrated. Everyone that has been using databases for a lengthy period of time has all the data in them that is needed, already. But, of what value is that data if it does not reflect current "environments"? The watering down of current data due to the emphasis on larger, older sample sizes, without adequate emphasis on recency within that/those sample(s), tends to cause the resulting methodologies to become antiquated/stale, IMO, and thus, less effective in the near future.

[whodoyoulike]

Has anyone come to the conclusion that computer models described by Bolton and Benter have been rendered usless, because of the increased psuedo R2 for the public model?

What do I mean by this. Is it now nearly impossible to get a postive delta R2 between the combined public and fundamental vs the public model.

Mike

Is this correct? I've never seen any detailed description of Bill Benter's computer model but, I know he was very successful with it over a number of years. It sounds as if your model has started to provide poor results. So, now you're stating all similar computer models must be providing similar results.

How about describing what your model does (or did) and what it's currently doing? It could be it never really worked as planned. Were you back fitting results? Maybe, you had a design flaw or a math error(s)? Or, you just had one or several logic errors?

I've found that there several dozen if not hundreds of members who are always willing to help solve problems of others on here

Magister, the higher the public pseudo R2 the better for me. I can always do better. I can do it with six covariates. And none of the covariates are handicapping factors. I know it sounds impossible, but it is true.

Mike

Actually the number of chickens exported to the EU countries on a daily basis...

Mike

This could be part of your problem.
.

[Native Texan III]

Is this correct? I've never seen any detailed description of Bill Benter's computer model but, I know he was very successful with it over a number of years. It sounds as if your model has started to provide poor results. So, now you're stating all similar computer models must be providing similar results.

How about describing what your model does (or did) and what it's currently doing? It could be it never really worked as planned. Were you back fitting results? Maybe, you had a design flaw or a math error(s)? Or, you just had one or several logic errors?

I've found that there several dozen if not hundreds of members who are always willing to help solve problems of others on here





This could be part of your problem.
.

The original model did not work well at all and it took years of modifications to get somewhere near profitable. Even then it did not do well without incorporating the public odds. Once it got going it made very good profits ($HK not $US) as it was the only model game in town betting on long term value. Once others copied it, the value disappeared.

The essential problem with general models is that horseracing is won or lost by a fraction and economics rather than engineering modelling which is essentially an averaging process does not deal with the minute differences and race specific insights needed to make a reliable profit. All the economic model does is to provide a consistent way of estimating long term value. Once others do the same, the value, once available, is removed. Folks have moved on.

[Cratos]

The original model did not work well at all and it took years of modifications to get somewhere near profitable. Even then it did not do well without incorporating the public odds. Once it got going it made very good profits ($HK not $US) as it was the only model game in town betting on long term value. Once others copied it, the value disappeared.

The essential problem with general models is that horseracing is won or lost by a fraction and economics rather than engineering modelling which is essentially an averaging process does not deal with the minute differences and race specific insights needed to make a reliable profit. All the economic model does is to provide a consistent way of estimating long term value. Once others do the same, the value, once available, is removed. Folks have moved on.

Very good summation; did you ever study econometrics?

[flatstats]

The original model did not work well at all and it took years of modifications to get somewhere near profitable. Even then it did not do well without incorporating the public odds. Once it got going it made very good profits ($HK not $US) as it was the only model game in town betting on long term value. Once others copied it, the value disappeared.

The essential problem with general models is that horseracing is won or lost by a fraction and economics rather than engineering modelling which is essentially an averaging process does not deal with the minute differences and race specific insights needed to make a reliable profit. All the economic model does is to provide a consistent way of estimating long term value. Once others do the same, the value, once available, is removed. Folks have moved on.

I don't buy into this "everyone does it so the value has gone".

Nothing, ever, IMO has influenced prices such that a method becomes overbet. Only in the 70s/80s you could say Beyer ratings did but I don't believe there is anything now that is overbet because of duplication of a successful method.

How could anyone duplicate Benter anyway? Did they clone his program? Use the same people he did? Place the same wagers in the same betting outlet?

If a (sound) method ultimately fails then it fails because it did not adapt to changes. It does not fail because everybody else did it too.

[barn32]

wjxX6ql4Eso

[GameTheory]

How could anyone duplicate Benter anyway? Did they clone his program? Use the same people he did? Place the same wagers in the same betting outlet?
Actually, yes, all of the above.

[Seabiscuit@AR]

Benter says openly in the paper which appears in Efficiency of Racetrack Betting Markets that his method might not work in another place or even in HK at another time. He says if the opposition get smarter then his method might no longer work. He was quite aware that the window for exploiting his method might be small

The models of one of the big HK syndicates broke a few years back after years of success with it. They claimed to have moved on to a new model that was working last I heard. The general agreement from HK is that it was easier to win in HK in 1994 than in 2014

[flatstats]

The differences between 1994 and 2014:

Computers 100 times faster. The post race analysis and the pre race processing could be done more instantly.

Wagering links faster. Didn't they place bets via old 33.6K modems? Surely price updates were slower then too so they could not work out the optimum stake in a timely manner. With faster pools updates now isn't that an advantage?

More mug and / or more smart money? As the population increases surely there is more mug money even if more smart punters are using technology like this.

Corruption. Are there more bent races now or were there more then? Sidenote: I am of the opinion that bent races go as much for you as much against you so it balances out for the average punter who is not in the know (of the bentness)

More Authoritative Experts. Who did Benter learn from? Did he have influence from authoritative sources at the time? Now days we have diverse experts from the field of racing and betting that we can call up in an instant. Now we have contrary views from people like Michael Mauboussin. We can be influenced by this where as in 1994 system creators may not.

Just Got Lucky. On that last point about Mauboussin. Maybe Benter just got lucky? Maybe he was in the right place at the right time? Maybe he did amass $100M but did he lose $99M before that and just hit a lucky streak? There are whales about today that maybe 'just got lucky'. They come and go and will always be replaced by someone who just got lucky.

[Seabiscuit@AR]

flatstats

You forgot rebates. Rebates in HK are big at 10% for big bets in the win pool and some other pools. This cuts the takeout in half for the rebate players and by itself will drive the odds much closer to efficiency. All the syndicates will be playing the rebates

[GameTheory]

The differences between 1994 and 2014:

More mug and / or more smart money? As the population increases surely there is more mug money even if more smart punters are using technology like this.Are you talking about the USA? If you haven't noticed, racing has been in freefall for the last 20 years (or more) -- less people playing less races and ratio of dumb/smart money MUCH more lopsided towards the smart side these days...

[flatstats]

Are you talking about the USA? If you haven't noticed, racing has been in freefall for the last 20 years (or more) -- less people playing less races and ratio of dumb/smart money MUCH more lopsided towards the smart side these days...

What about the rise in online betting? Can you still not access betting exchanges in the US?

In the UK sites like Betfair have opened up great opportunities for exploiting mug bets. Not just on horseracing but other sports.

OK, you still can't bet in Benter quantities or make anything like he did via an exchange but the odds are so much more favourable than the Tote or with bookmakers.

[GameTheory]

What about the rise in online betting? Can you still not access betting exchanges in the US?Nope, no exchanges. They've only been taking away online opportunities to bet by passing new laws, not adding to them...

[Native Texan III]

I don't buy into this "everyone does it so the value has gone".

Nothing, ever, IMO has influenced prices such that a method becomes overbet. Only in the 70s/80s you could say Beyer ratings did but I don't believe there is anything now that is overbet because of duplication of a successful method.

How could anyone duplicate Benter anyway? Did they clone his program? Use the same people he did? Place the same wagers in the same betting outlet?

If a (sound) method ultimately fails then it fails because it did not adapt to changes. It does not fail because everybody else did it too.

In Hong Kong, which is what we are talking about, the syndicates have copied and improved on Benter's pathway methods. No need to clone - just need to know it can be done and pay Chinese maths stars to sort it. Once there was one, now there are dozens all fighting each other for scraps.

The method was eventually sound in identifying long term value. It still identifies prices which were long term value but are no longer value because others on the Tote are betting the same combinations as the models all use the same Hong Kong data, that is widely available.

Where you may be correct is if someone has a little system that is profitable and keeps it to himself. If he publicises it only a very few take it up. However once a major profitable methodology is given away people flock to copy it and the value definitely disappears.

There are changes in racing but a well formed model keeps itself up to date dynamically and automatically. That won't happen if you are using database stats from a decade ago within a static method.

[Native Texan III]

Very good summation; did you ever study econometrics?

I have studied some econometrics and many other subjects that might shine light on things. I find many economists never really try to understand the mechanisms and the whys and wherefores - they just use unproven theories that are simplistic in the extreme and get weird causation theories from correlations that don't ever apply in reality. Hence economists are the world's worst forecasters.

Horseracing is simpler in that the mechanics of getting to the line before the other horses is repeatable and checkable.

[gm10]

Using log odds, that is very similar to NA tracks(major and minor) of approximately .17

Mr. Benter is honest and speaks the truth.

Mike

0.17 sounds on the high side ... is this for all races?

[gm10]

Actually, yes, all of the above.

How did they manage to clone his program?

[GameTheory]

How did they manage to clone his program?They worked for him, and took a copy with them when they left. Several groups ended up with it. That's what I've heard anyway, I wasn't there.

[Robert Goren]

They worked for him, and took a copy with them when they left. Several groups ended up with it. That's what I've heard anyway, I wasn't there.Reverse Engineering?

[Cratos]

I have studied some econometrics and many other subjects that might shine light on things. I find many economists never really try to understand the mechanisms and the whys and wherefores - they just use unproven theories that are simplistic in the extreme and get weird causation theories from correlations that don't ever apply in reality. Hence economists are the world's worst forecasters.

Horseracing is simpler in that the mechanics of getting to the line before the other horses is repeatable and checkable.


I realize that this is a horseracing forum and I want to stay on that subject, but I disagree with you about your assessment of economists.

Also many occurrences are “repeatable and checkable’” but it is the successful predictability of the occurrence that pays the rent and horseracing in terms of predicting winners (the occurrence) is very difficult in terms of achieving long term profitability at a high dollar return level.

[GameTheory]

Reverse Engineering?No, I think they had the actual code, they probably helped write it. Then split off and formed their own group.

[flatstats]

At a guess the code then is totally useless now?

Like finding a cheat sheet for the Rubiks Cube.

[gm10]

They worked for him, and took a copy with them when they left. Several groups ended up with it. That's what I've heard anyway, I wasn't there.

You're right. I remember it now, there was an acrimonious split between Benter and his partner and they both took the model to start on their own.