Team analysis

[JJMartin]

Been thinking of testing an idea involving pooling various handicappers with different techniques who have an edge in flat win betting. Comparing picks on past races first and then testing forward. The goal is to develop the highest strike rate/roi from determining the best bets based on the collective judgment from within the team when selections are commonly arrived at. Ideally, your picks will be computer generated/assisted. Got a couple people on board, would like a databaser perspective/input...PM me.

[traynor]

Been thinking of testing an idea involving pooling various handicappers with different techniques who have an edge in flat win betting. Comparing picks on past races first and then testing forward. The goal is to develop the highest strike rate/roi from determining the best bets based on the collective judgment from within the team when selections are commonly arrived at. Ideally, your picks will be computer generated/assisted. Got a couple people on board, would like a databaser perspective/input...PM me.

When you get to the decision making part in real time, you might consider the Delphi model (of decision making). What you propose has been tried a number of times, and usually fails because it devolves into an 800 lb. gorilla surrounded by a gaggle of sychophants. In short, the decisions made going forward are not discrete, but rather rely on individual perceptions of "what the group wants" (which becomes--very quickly--"what the 800 lb. gorilla wants").

It is a great idea, but one difficult to make work in the real world (going forward, rather than in retrospect). One of the first things you will likely discover (ask me how I found this out) is that the majority of the "various handicappers with different techniques who have an edge in flat win betting" very quickly "lose" that edge under the pressure of actual performance in real time (rather than in records of past races).

Several currently successful groups are using group decision-making for pre-race analysis. The most successful (of those I am personally acquainted with) use the Delphi model. How do I know this? I helped write the software.

[JJMartin]

When you get to the decision making part in real time, you might consider the Delphi model (of decision making). What you propose has been tried a number of times, and usually fails because it devolves into an 800 lb. gorilla surrounded by a gaggle of sychophants. In short, the decisions made going forward are not discrete, but rather rely on individual perceptions of "what the group wants" (which becomes--very quickly--"what the 800 lb. gorilla wants").

It is a great idea, but one difficult to make work in the real world (going forward, rather than in retrospect). One of the first things you will likely discover (ask me how I found this out) is that the majority of the "various handicappers with different techniques who have an edge in flat win betting" very quickly "lose" that edge under the pressure of actual performance in real time (rather than in records of past races).

Several currently successful groups are using group decision-making for pre-race analysis. The most successful (of those I am personally acquainted with) use the Delphi model. How do I know this? I helped write the software.

That sounds interesting, I am not acquainted with the Delphi model, will look into it. The main thing I had in mind with this project was to simply compare selections from a predetermined sample of races and then determine whether a 'group' selection would outperform an individual's selection over time. The premise is that more perspectives from diverse angles would be more conducive to accuracy then an individual's alone. I am not really expecting a disclosure of a player's method details but when each instance arises where the majority eye the same selection (at good odds) it warrants a closer examination of those horses to see if there is a predictive quality that can be detected and isolated that was previously elusive. There is no doubt that this is a difficult task but perhaps we can enhance our knowledge to some degree through an intelligent collaboration.

[Dave Schwartz]

various handicappers with different techniques who have an edge in flat win betting.

I suggest that if you have this, you already have all you need.

[traynor]

That sounds interesting, I am not acquainted with the Delphi model, will look into it. The main thing I had in mind with this project was to simply compare selections from a predetermined sample of races and then determine whether a 'group' selection would outperform an individual's selection over time. The premise is that more perspectives from diverse angles would be more conducive to accuracy then an individual's alone. I am not really expecting a disclosure of a player's method details but when each instance arises where the majority eye the same selection (at good odds) it warrants a closer examination of those horses to see if there is a predictive quality that can be detected and isolated that was previously elusive. There is no doubt that this is a difficult task but perhaps we can enhance our knowledge to some degree through an intelligent collaboration.

Understood. The Delphi model factors out a lot of semantic noise from the process. Basically, none of the decision makers see anyone else's decisions until after the fact. Ideally, there should be no disclosure of what an individual did or did not do correctly (in this particular, isolated incident--one of a very long string of events) or incorrectly. All it takes is one healthy, "The bleeping 4? Why would anyone in their right mind bet on the bleeping 4?" to totally mess up group decision making (and any useful result from group decision making). Activity degenerates into minimizing "risky" selections (to avoid post-race snickering--sometimes even worse than pre-race snickering) and dedicated CYA. As in, "Bet the chalk. At least if I'm wrong, I will have a lot of company."

It can work. It can be useful. It can be profitable. Understand that group decision making is a process, like many other things, and benefits from a thorough understanding of its strong points and weak points.

A situation to avoid might be the typical "groupthink" of a number of people arguing among themselves about the best bets in an upcoming race. ("Groupthink" is not a malapropism. I understand Janus and groupthink quite well. In fact, I did a graduate level term paper on Morton-Thiokol and the Challenger.)

[JJMartin]

I suggest that if you have this, you already have all you need.

I want to expand and optimize and keep learning. The goal is to develop the most consistent optimal bets possible defined as the highest strike rate and lowest losing streaks possible (the perfect balance in theory), one that hopefully exceeds an individual effort (without the need to disclose any personal components of methodology). Add to that the team motivation to alleviate falling into stagnation and disinterest. Never tried this before so I am curious to see what results.

[traynor]

I want to expand and optimize and keep learning. The goal is to develop the most consistent optimal bets possible defined as the highest strike rate and lowest losing streaks possible (the perfect balance in theory), one that hopefully exceeds an individual effort (without the need to disclose any personal components of methodology). Add to that the team motivation to alleviate falling into stagnation and disinterest. Never tried this before so I am curious to see what results.

I think if you make it what most on forums tend to shun--impersonal--your chances for success will dramatically increase. Meaning, those whose primary motivation is seeking strokes and recognition make lousy long-term bettors. That category includes many who claim (or sincerely believe) that they are "winning bettors." In the real world, there is a big, big difference between a regression that shows a profit and actually betting it.

If it is "impersonal" (the essence of the Delphi decision model) the objective is "everybody paddles (pedals, pushes, pulls, or whatever), nobody rides for free." When people can see that the effort they exert is to their benefit--as well as to the benefit of others--motivation to achieve increases. Its a bit tricky to pull off, but when it works, it works well.

There were lots of "Sartin groups" in the 1980s that thought group efforts would be synergistic. Problem was that most of those who believed they had "a 70% strike rate betting two horses with an average mutuel of $10.78" soon discovered their analytical skills did not translate well to betting in the real world.

[JJMartin]

...There were lots of "Sartin groups" in the 1980s that thought group efforts would be synergistic. Problem was that most of those who believed they had "a 70% strike rate betting two horses with an average mutuel of $10.78" soon discovered their analytical skills did not translate well to betting in the real world.

This is why I want to test past races to start and then test going forward, tracking everyone's individual results for some period of time. A reasonable foundation in skill evidenced through results is a critical part of the procedure. I am not really interested in sharing handicapping ideas per se within this project, I am interested in expediting a process that would offer a quick assessment on a race, one that would probably involve using a custom Excel worksheet for efficiency. Now, I haven't worked out the exact details on every step but I have an idea of how it should work and any input is appreciated.

[Robert Fischer]

It's not a bad idea, and even if it doesn't prove fruitful it should be a process that you can learn from.

I think you will find the highest strike rates will simply start with the low-odds spectrum of the consensus. From there, the most indicative factors will be unlikely to better those of your best individual player within that odds range.

Profitable flat win bettors are highly skilled. It's difficult to do a consensus project like that, and attempt to improve on experts who are dominating their field, often in a specialized approach.

good luck

[traynor]

This is why I want to test past races to start and then test going forward, tracking everyone's individual results for some period of time. A reasonable foundation in skill evidenced through results is a critical part of the procedure. I am not really interested in sharing handicapping ideas per se within this project, I am interested in expediting a process that would offer a quick assessment on a race, one that would probably involve using a custom Excel worksheet for efficiency. Now, I haven't worked out the exact details on every step but I have an idea of how it should work and any input is appreciated.

Yes and no. It is a lot like handicapping a race from last month, and discovering that one selected the winner. Unless one bet on it, it really doesn't matter. That is not meant to be facetious--lots of people handicap backwards, and believe they are doing wonders. Some are even able to convince others that they are doing wonders. I have seen enough betting groups to understand that no one really cares what who did yesterday--the only thing that matters is what are they going to do today?

While applying specific methodologies to past races is nice for research, it proves little. I suggest you might progress faster if you start with tomorrow's races, rather than yesterday's. Especially if you are dealing with handicappers using computer apps--it doesn't take much to backfit a chunk of races to "discover" something that looks profitable.

[raybo]

I think Dave is probably right. If you already have a profitable, real time, flat bet method, you're probably better off concentrating your efforts on that. Entering into a process, as suggested, will necessarily take time away from what you are already involved in doing.

While there is the possibility of being more successful within a group, that is not a given and might actually be a negative outcome. Most flat bet winning players are going for prices, not winners, so the idea that there will be some sort of consensus between several of these kinds of players is probably not likely.

I'm still considering this process, but the more I think about it, the more I think I would be better served with my individual approach.

I hope you find a couple of net profit players to join in, but as of now, I doubt I will be one of them.

[JJMartin]

I think Dave is probably right. If you already have a profitable, real time, flat bet method, you're probably better off concentrating your efforts on that. Entering into a process, as suggested, will necessarily take time away from what you are already involved in doing.

While there is the possibility of being more successful within a group, that is not a given and might actually be a negative outcome. Most flat bet winning players are going for prices, not winners, so the idea that there will be some sort of consensus between several of these kinds of players is probably not likely.

I'm still considering this process, but the more I think about it, the more I think I would be better served with my individual approach.

I hope you find a couple of net profit players to join in, but as of now, I doubt I will be one of them.

Yes, it seems people don't have the time to commit, which is understandable.
As an alternative to this, I had another idea that may require a programmer unless I can figure out how to do it myself, let me present the concept:


Am embedded interactive Excel spreadsheet on a web site that would allow anonymous bet selections from anyone. A user would log on with a user name and password. They would view the day's current races and make just 1 horse selection to Win on any race they wish to contribute to, from as many as they want to none. Only after making a selection are they able to view the same race's cumulative score. The individual's betting history would be tracked and their current ROI% would be shown next to each selection in subsequent races. In addition, other people selecting the same horse would have their own current % averaged in. You would never see a person's name, only percentages of ROI. By using ROI rather than strike rate, the emphasis is on price rather than win rate. The premise would be to harness the community's best selections based on the best ROI's. In order to maintain integrity, any ROI's that are about average with random selection or below would be weeded out. You would always be able to view your own personal history privately. This format makes it easy to participate with no pressure. May sound complicated but its not. I could put up a sample spreadsheet pic if anyone is interested.

[raybo]

Yes, it seems people don't have the time to commit, which is understandable.
As an alternative to this, I had another idea that may require a programmer unless I can figure out how to do it myself, let me present the concept:


Am embedded interactive Excel spreadsheet on a web site that would allow anonymous bet selections from anyone. A user would log on with a user name and password. They would view the day's current races and make just 1 horse selection to Win on any race they wish to contribute to, from as many as they want to none. Only after making a selection are they able to view the same race's cumulative score. The individual's betting history would be tracked and their current roi% would be shown next to each selection in subsequent races. In addition, other people selecting the same horse would have their own current % averaged in. You would never see a person's name, only percentages of ROI. The premise would be to harness the community's best selections based on the best ROI's. In order to maintain integrity, any ROI's that are about average with random selection or below would be weeded out. You would always be able to view your own personal history privately. May sound complicated but its not. I could put up a sample spreadsheet pic if anyone is interested.

That sounds a little better, as each player wouldn't be required to play races or tracks he doesn't normally play. I would suggest a minimum odds requirement alongside the pick(s), also (value is a major consideration for most profitable players). Also, maybe up to 3 picks (with the minimum odds requirement), rather than 1, might fit some players' methods better.

[JJMartin]

That sounds a little better, as each player wouldn't be required to play races or tracks he doesn't normally play. I would suggest a minimum odds requirement alongside the pick(s), also (value is a major consideration for most profitable players). Also, maybe up to 3 picks (with the minimum odds requirement), rather than 1, might fit some players' methods better.

That could easily be implemented by designating categories such as "primary pick", "secondary" and so forth. Could make a column for each one and each category would have its own % tracked.

[raybo]

That could easily be implemented by designating categories such as "primary pick", "secondary" and so forth. Could make a column for each one and each category would have its own % tracked.

There you go!