[BJennet]

Originally Posted by Ted Craven

J2EE Developer,

Some interesting back and forth! I agree with many of your points, and at the same time, I cannot understand some of your assertions at all.

I acknowledge it is my responsibility as proprietor/developer of RDSS to provide not only the software but the learning tools, and I am sorry you found not all the tools you had hoped for, for example: automated paceline/contender selection, more integrated modeling tools, search for race/betting opportunities among all races today matching your observed situational strengths (i.e. a 'portfolio' of types of races you're good at). These specific tools do not exist yet, though as I mentioned, I am working on each of them, and more, for RDSS 2.0.

Other users are currently making use of 'good-enough-for-now' versions of paceline selection (the Perceptor ranking in concert with Total Energy and Primary Line Score on the Primary Screen), and factor modeling (including mutuels and tote info) via the Export to Excel feature. With some practice, you should expect be able to select consistent pacelines and winnow them down to 5 contenders in just a few minutes per race (some races demand a bit more investigation); exporting to an external model keeper and collating the output - a few minutes more.

If you thought you might be getting the foregoing tools and improvements on teh Methodology you knew - and didn't - I accept your disappointment, though I note that Sartin, Bradshaw and others always advised that if you don't have all the data (or tools) you need, use what you've got...

I believe you are evaluating tools from a specific point of view of how optimally you might 'ramp up the volume' so as to get enough bets at your current hit rate to make it worthwhile without breaking the pools (and your psychological wager limits). Also, since you can see Synergism 6 as a baseline for at least modest profit, you are seeking to improve from there. Completely reasonable goal.

Some things I don't understand at all: you say RDSS (or the modern Methodology of circa late 1990's forward) offers no improvements to figures you are able to compute by hand entry to your TI calculator, or which dozens of other software provide equally well. If you have read all the Follow Ups 3 times, including the later ones, you will know that Sartin considered his later figures (adjusted by TrackMaster DTV and ITV variants) far superior to earlier efforts. Perhaps you either disbelieve him, or you have succeeded in reverse engineering those formulae and have them already embedded in your calculator (though I wonder at the source of the inter-track variants). If RDSS is no better at compounded incremental energy calculations than dozens of other software (or your own manual efforts) - I am alarmed that the sometimes generous mutuels which various folks post screenshots pointing to - often by selecting lines from the most recent 3 races - continue to persist, given that users of dozens of other software must be seeing (and if so, why not betting more on) the same horses. I submit that RDSS has a few tricks up its sleeve, which enable a practitioner to gain insight into more races, or more formerly obscure races, than some of its peers (or predecessors). If this last assertion is true, perhaps you could gain a few more playable races per track than presently, or perhaps a few more ROI points than before. If you think it's not true, or at present too much work - well, just keep doing what you're doing.

I also don't understand your assertion that a practiced user can do no better currently than zigging and zagging in their paceline selection, employing largely intuition, certainly not applying their method consistently enough that their demonstrations, or screenshots, or analysis of approach is mostly useless to a new user. Sure, in a post-mortem of a lost wager, one might frequently see that another line selection, or contender decision would have obtained a better bet - but that's not what the mature analyst does! The mature analyst creates a consistent line selection method and contender winnowing process (by reading published materials, observing others' strategies, their own trial end error and 'bootstrapping') then applies it rigourously - winning where they win and losing where they lose (and with equanimity), always focused on their records which tell whether that method is producing profit over reasonable session lengths (e.g. 20 cycles). If it is - one the one hand, who cares of you lose 50% of your bets (though post-mortem zigzags may point to a winner in every race): that is a fantasy which must be surmounted to get a method producing a base of profit from which one can fine tune. I assert than you can develop and apply currently a rigourous enough approach that it could later be captured as rules in software. And the proof is in one's own betting records. Different people may have differing line selection and wagering approaches, and win and lose different races with that. Doesn't mean one can't develop a rigourous approach.

If you don't have access to RDSS, you can't answer the question: would I have done as well over a card of races as whatever poster did (and there are several almost full card workups posted recently and ongoing for new users to model). But you could if you wanted to. Maybe you don't bet 'those kinds of races': no maidens, no 2 years olds, no 3 year olds, no Turf, no artificial, no routes, no Mountaineer, no Penn National, no Tampa - no exactas...whatever filters one has previously identified as requirements to success using your method so far. Well and good. Though, consider expanding your horizons...

As for the Law of Small Numbers, and Messrs. Kahneman and Tversky, I only understood enough of it to suspect that neither were horse-race bettors, or at least, were discussing measurements from longitudinal enough samples and domains which did not require very recent, small sample intel from the very current time frame in which we need to be making a wager. Headline: "statistician drowns in average lake depth of 18 inches!" I want to know how deep the water is 2 feet in front of me (cause I need to make a wager now, not at any average point over a race meet or track's entire 4 year history). I don't need to know that a small local sample doesn't necessarily conform to other small local samples taken from hundreds of locations over the lake's surface (or a given handicapping factor's 6 month or 4 year history). I'm sure that analogy has been answered a thousand times in Statistics 101 classes (which I never took). But I believe it is pertinent. Anyway, some people haven't yet discovered that they can be making more money than they currently are if they only used a factor model.

I really enjoy an informed (and articulate) skeptic's point of view, and hope that you find that some views of others shared so far (some contrary to your own) are offered not only from a need for self-justification or as knee-jerk protectionism, but also from a helpful stance. I almost want to offer you a long-term software and data account so that you can challenge RDSS as it evolves, and be challenged to disprove, that the toolset offers distinctive abilities compared to the dozens, and is both in the spirit of Sartin's prior work but an evolution of it - and capable of professional level use.

Wishing you well in your further explorations.

Ted

Hi Ted,

Thanks for having the courage to make public a complaint which has produced a very interesting thread. First, I don't think you should feel too much of a need to defend the Spec 160/RDSS software - most of the members here, including myself will attest to its high value. And I don't think this poster was being very critical of the program per se - but was simply looking for functions that are currently not part of it.

Also, although I usually agree with most of your posts, I think your comments on Kahneman and Tversky really represent a misunderstanding of the application of their work to handicapping. Bear with me if I'm repeating familiar information, but the basic nature of their work deals with the difficulty the human mind has in grasping the nature of randomness. In particular, they uncovered a misperception of reality based on an overemphasis on the predictive power of recent events, as characterstic of the human mind.

As one of many examples, I'll mention their exposure of the 'hot hand' fallacy, in which an study of the Boston Celtics shooting percentages found that the likelihood of a given player to be 'hot' - to hit a number of shots in a row - was well within the range of the normal distribution, rather than an aberration.

Thus the poster is reasonably asking for more statistical validation of, say, the Brohamer Model, which is unavailable. The usual response is - 'Well it just works.' But, as with almost all of the Sartin material, the response demonstrates a lack of awareness of the difference between correlation and causation.

Sartin, in the later Follow Ups, also mentions that both he and Brohamer had abandoned the model, feeling that it was less predictive than the BL/BL, and that he himself modelled nothing.

You say that Kahneman and Tversky are irrelevant, as non-horseplayers, but it's worth keeping in mind that their frequent colleague in the world of behavioral finance, Richard Thaler, has collaborated on articles on the nature of handicapping with William Ziemba, who needs no introduction here. Ziemba was also an advisor to William Benter, who employed long-term statistical studies in the creation of his enormously successful handicapping model. And it's safe to say that all of these people understand the nature of stochastic processes, of which a horse race is an example, better than we do.

I've touched on some of these issues before, and I don't want to seem to be haranguing anybody, but given their crucial importance to the nature of gambling, I think it's worth continued discussion.

Thanks for your efforts, as always.

Cheers,

BC