Software Learning Curve

[traynor]

An issue that needs exploration is that of the so-called "learning curve" associated with race analysis software, and the relationship of that "learning curve" to software intended for recreational use as opposed to professional use.

For recreational use, "learning to use the software" is part of the activity--it creates a sense of investment (of time and effort) that is (hopefully) rewarded with increased accuracy (higher win percentage, higher ROI, fewer losers). Because the primary motivation is recreation, there is little or no downside to software that requires a "learning curve"--a euphemism for "subjective interpretation of output." Specifically, the software does not "make selections"--it presents data in various formats so the user can make subjective decisions on a race-by-race basis as to which bit of output is most important/significant/predictive in that particular race.

Having developed, and worked on team development of "professional-level" software applications, when "learning curve" is mentioned to a professional bettor or betting syndicate, the immediate response is "go back and work on it until it is complete, then show us." Specifically, the notion that subjective decisions made on a race-by-race basis can outperform computer algorithms and structured decision-making is dismissed as irrelevant nonsense. That is, the "learning curve" associated with race analysis software (that is highly desirable in software intended for recreational use) is almost certain to be a deal-breaker in the design of professional-level race analysis software.

If the software is incapable of "learning" on the fly, adapting to changing circumstances, and generating the most predictive algorithms on its own--it is incomplete. That deficiency is unrelated to the "game of racing"--it is related directly to developer competence in the design of decision-support software.

There is a place for both kinds of race analysis software. Unfortunately, the distinctions between the two types are often not made. That leads directly to disappointment (and often hostility) in new users who have spent upwards of a thousand dollars in software and data subscriptions fees before discovering that what they believed would/could be a professional-level (as in "profitable") software application was in fact designed for recreational use.

The simple litmus test for race analysis software is the "learning curve." The steeper and/or longer the "learning curve," the more indicative it is that the application is designed for recreational use. The so-called "black box" application is a reality of professional-level software. Serious bettors want applications that do the heavy lifting in implementing decision-making processes, and expect software to leverage the strong points of computer capabilities, rather than software that requires subjective interpretation of the output by the user.

Is such software prohibitively expensive? No. Is it impossibly difficult to develop, requiring an investment of hundreds of thousands of dollars for a crew of snarky quants? No. Is it necessary to have a bankroll of millions to "take advantage of the software's capabilities"? No. Are you likely to find someone who has such an application willing to sell/lease/loan it to you for some modest (or outrageous) sum of money? No.

For anyone interested in the potential profit available to astute bettors in the horse racing field--as opposed to purely recreational use--I strongly encourage you to pay attention to what Delta Lover is doing with his open-source data analysis project. He seems to be one of the few who realize the potential of software as an aid to decision-making, and the (obvious to serious bettors) potential of structured decision-making processes as opposed to "subjective interpretation of the output."

[traynor]

I assume everyone realizes that the injection of subjective interpretation into race analysis essentially destroys the value of regression analysis as a decision aid. Specifically, models based on past results use specific criteria to determine success or failure. Tweak or fiddle with that criteria with subjective interpretation of trivial details of individual races, and the model's value evaporates. The numbers only work in your favor when the software recognizes, "Aha! This is Race Scenario D-456. The best possible decision in this scenario is ..."

It is unwise to let one's ego and (largely illusionary) "handicapping experience" diminish or eliminate the potential of the software by reverting to seat-of-the-pants scrambling at the last moment. Humans are much too inclined to selectively remember "stuff that worked" and to ignore or forget "stuff that flopped." One of the primary advantages of using (well-designed) software is that it has no ego to defend.

Professional-level race analysis software should be as explicit in defining the best decision to make under each set of circumstances as professional-level blackjack strategies. Perhaps that is why blackjack professionals adapt to professional-level race analysis so easily--and why they have so little interest in software applications that "require" subjective interpretation of trivial details of individual races. They are quite willing to let the software do what it does best--select the best possible decision to make in a specific race scenario.

[Dave Schwartz]

Having developed, and worked on team development of "professional-level" software applications, when "learning curve" is mentioned to a professional bettor or betting syndicate, the immediate response is "go back and work on it until it is complete, then show us." Specifically, the notion that subjective decisions made on a race-by-race basis can outperform computer algorithms and structured decision-making is dismissed as irrelevant nonsense. That is, the "learning curve" associated with race analysis software (that is highly desirable in software intended for recreational use) is almost certain to be a deal-breaker in the design of professional-level race analysis software.

Rarely do I agree with you as wholeheartedly as I do here.

But here's the thing: To many handicappers the ability to add subjectivity is essential. It is why they play.

It took me a long time to realize that.

For the developer, it is a case of "Give them what I think they need" versus "Give them what they want."

Many handicappers are looking for a good Handicapping Experience even more than Proven Profitability.

[traynor]

Originally Posted by Dave Schwartz

Rarely do I agree with you as wholeheartedly as I do here.

But here's the thing: To many handicappers the ability to add subjectivity is essential. It is why they play.

It took me a long time to realize that.

For the developer, it is a case of "Give them what I think they need" versus "Give them what they want."

Many handicappers are looking for a good Handicapping Experience even more than Proven Profitability.

Absolutely. That is why there is such a dichotomy between types of applications, and such frustration when users fail to understand the differences. I agree that the subjective element is why many handicappers enjoy handicapping--and software applications that emphasize (and facilitate) that aspect.

The recreational bettor who tries to out-think professional-level software (that is expressly designed to eliminate subjective interpretation) is going to be as disappointed as the bettor expecting profitability to be the primary driver of recreational software.

[AndyC]

Originally Posted by Dave Schwartz

Rarely do I agree with you as wholeheartedly as I do here.

But here's the thing: To many handicappers the ability to add subjectivity is essential. It is why they play.

It took me a long time to realize that.

For the developer, it is a case of "Give them what I think they need" versus "Give them what they want."

Many handicappers are looking for a good Handicapping Experience even more than Proven Profitability.

So let me see if I understand. Don't be subjective and let the software do it's job. What does it mean when you said "The most difficult of these two goals is clearly #1: winning. About 20-25% of our users are winning players, and several win at a "professional level. That still leaves 75-80% of our users that are not winning players." Does that mean that the software does it's job but that 80% of your clients don't use it properly?

[Dave Schwartz]

Andy,

The software is VERY programmable. No two people use the software the same way.

YOU are still responsible for figuring out how to win.

Would you hold the DRF or BRIS responsible for not winning with their products? Of course not. It is all about what you do with it.

That being said, in a world where 1% of the race going public wins, having 20% winners is a pretty good increase.

[thaskalos]

Originally Posted by Dave Schwartz

Andy,

The software is VERY programmable. No two people use the software the same way.

YOU are still responsible for figuring out how to win.

Would you hold the DRF or BRIS responsible for not winning with their products? Of course not. It is all about what you do with it.

That being said, in a world where 1% of the race going public wins, having 20% winners is a pretty good increase.

Dave, forgive me if this has been asked of you before...but does your software offer the option of doing its own handicapping without any extensive programming on the part of the user?

[Dave Schwartz]

There are several pre-designed ways to handicap right out of the box, so that answer is "Yes."

There is an installation video that begins with: "20 minutes from now you will be handicapping your first race, counting installation."

Also, not sure that "programming" as really the operative word. More like "arranging."

There are probably a dozen different handicapping paradigms... handicapping objects, Impact Values, Monte Carlo simulations, several forms of AI, dynamic research tools (i.e. build systems according to your specifications on a per-race basis), and a lot more.

Most can be mixed together. You choose the factors from around 4,000 factors per horse.

[AndyC]

Originally Posted by Dave Schwartz

Andy,

The software is VERY programmable. No two people use the software the same way.

YOU are still responsible for figuring out how to win.

Would you hold the DRF or BRIS responsible for not winning with their products? Of course not. It is all about what you do with it.

That being said, in a world where 1% of the race going public wins, having 20% winners is a pretty good increase.

So wouldn't programming the software be subjective? Either you are subjective using the software or you aren't. The narrative of this thread has been that you should not be subjective using the software. What am I missing?

[Dave Schwartz]

Andy,

Exactly! The design of your system is subjective but the actual handicapping of a given race is not.

This is what most of our users strive for.

So much so, that they can create a rule-based system for betting the races and fire off the bets with a single click.

(This creates a bet file that can be uploaded to any ADW that uses the industry "standard" approach to the bet file.)

[traynor]

Originally Posted by AndyC

So wouldn't programming the software be subjective? Either you are subjective using the software or you aren't. The narrative of this thread has been that you should not be subjective using the software. What am I missing?

It is not that one should or should not make subjective decisions with his or her software application. It is that software designed for recreational use encourages (or requires) tweaking and fiddling by the user, as part of the "handicapping" process. There is a "learning curve" associated, as the user learns to make the various adjustments, filters, and weightings associated with such software. As Dave pointed out, that is what most recreational handicappers want. ROI is either a distant second or non-existent as a motivating factor.

Conversely, professional-level software is focused exclusively on maximizing the capabilities of computers to automate the process of building AND implementing appropriately predictive wagering models--with little or no "subjective" decision-making beyond the choice of track and day (and even that may be automated).

The basic reality is that "subjective opinion" and user tweaking is a hallmark of recreational software, and diminishes both the accuracy and usefulness of professional-level software.

[traynor]

I think the major issue (and primary cause of user dissatisfaction and frustration) is confusing software designed primarily for recreational use (with ROI as almost irrelevant) with software designed for professional-level use (with ROI as the primary goal). They are two distinct types, with two very distinct (and separate) goals. Each has its place. One should clearly understand both the intent of the software and one's own motivation before making software selection decisions.

[raybo]

Originally Posted by Dave Schwartz

Rarely do I agree with you as wholeheartedly as I do here.

But here's the thing: To many handicappers the ability to add subjectivity is essential. It is why they play.

It took me a long time to realize that.

For the developer, it is a case of "Give them what I think they need" versus "Give them what they want."

Many handicappers are looking for a good Handicapping Experience even more than Proven Profitability.

I agree! I have clients who purchased my black box and then found out that they couldn't, or wouldn't, do what the output told them to do. They have just destroyed the true ability of the method.

Most players would rather keep doing what they've always done (pick apart races subjectively), and keep losing, than to let the software do what it does best (make the decisions for them), and improve their bottom line.

[thaskalos]

Originally Posted by traynor

It is not that one should or should not make subjective decisions with his or her software application. It is that software designed for recreational use encourages (or requires) tweaking and fiddling by the user, as part of the "handicapping" process. There is a "learning curve" associated, as the user learns to make the various adjustments, filters, and weightings associated with such software. As Dave pointed out, that is what most recreational handicappers want. ROI is either a distant second or non-existent as a motivating factor.

Conversely, professional-level software is focused exclusively on maximizing the capabilities of computers to automate the process of building AND implementing appropriately predictive wagering models--with little or no "subjective" decision-making beyond the choice of track and day (and even that may be automated).

The basic reality is that "subjective opinion" and user tweaking is a hallmark of recreational software, and diminishes both the accuracy and usefulness of professional-level software.

Traynor...may I ask a simple question?

Let's say a horseplayer is overwhelmed with the other issues of life...and he cannot give this game the time and effort that it takes to become a profitable player. He is a man of considerable means, and is willing to invest any amount necessary in order to acquire a piece of software which will do all the handicapping that he himself is unable to do. He is not looking for fun or excitement; he is 100% profit-motivated. But he is also not interested in doing any "assembling", or "programming" with the newly-acquired software. All he wants to do is the betting. And he is not looking for a discount. He wants to pay what is asked...provided the product is as advertised.

Where does this horseplayer turn for the help that he needs?

This idea that the software user must "figure out his own way to win" reminds me of the Sartin groups of old...who used to blame the losses not on the method...but on the improper pace-line selection.

[Dave Schwartz]

Dave is waving his hand furiously because he would like to answer Thaskalos' question but will wait fro Traynor.

It is very difficult.

Dave may need therapy.

(Especially with this talking-in-the-third-person-thing.)