CBedo
08-26-2009, 01:43 AM
I was asked what percentage of my total plays are model plays and I thought a more thorough discussion might be interesting.
Personally, for the last few years, the percentage of my plays that were model plays has for the most part always been greater than zero. This might be a little loosely defined. Let me explain. For a good chunk of that time, as I worked on more automated approaches (most of which failed miserably), I have at least had models that produced "for further study" lists. These could be as simple as a spot play that I liked which an automated screener would let me know I need to look at a certain race more closely. I have never been able to handicap more than 2-3 tracks per day manually, as I tended (still do) to overdo it and really dig into races spending forever on them. So the "for further study" lists were easy ways to hopefully pick up a few more bets on race types that I was doing well with at the time, or at least spending my time where there was a higher probability of generating a bet.
That progressed to me trying to develop more comprehensive models (which again, most have failed miserably). I always wanted to spend more time on them, but between data costs being high, simulcasting and ADWs being early and tough to use, and having a real job managing equity money, I never spent that much time. All of those things have changed, and I really enjoy the intellectual challenge (I'm always been a blowhard big thinker--executing on some grandiose plans has been the bigger challenge). Lately, I have been trying to spend more time getting some of the "what ifs" out of my head and either into a spreadsheet, database, or search server. As I have worked more on the research side, it has spurred more model building, and the current culmination of it is the model picks that you guys have seen me post over the last couple months. The first iteration was the model that I posted picks for for a week. Since then, I have made a few tweeks, but it is still basically the same model. It has continued to be profitable (but has had some terribly painful losing streaks), and I have tried to automate the whole process more so I could bet it and use my time more efficiently to work on improving it (I see lots of areas of work; many of you have helped me and given me your insights). The first part of that was automating the creation of the bets. Once I did that, I was betting it (small size) every day at a couple of ADWs or at the OTB. For the last couple of weeks, I have been trying to take that one step further, trying to not only automate the model, but the whole process from downloading of data to uploading of bets--almost there.
Sorry that I got a little off track, but it was somewhat necessary to explain the answer to the original question. As I started working more on this model, the percentage of my total bets has gone from approaching zero to approaching 100%. This was not so much because I thought it was better than what I was doing, but more that it was (and is) time consuming and I look at it as (hopefully) a good investment. As such, the percentage of model bets has gone up but my total handle has dropped precipitously. That has started to change a bit as I have more of the automation done and can get back to doing some other handicapping instead of spending time creating bets.
So I guess that's the long way of saying that my model percentage of plays is too high right now due to development, but I would hope that as it goes down somewhat as a percentage of plays, the percentage of handle will rise to a higher percentage. I guess that all depends on profitability, since that's what drives the the thing.
Enough of my rambling, how about the rest of you model builders? I'd love to hear from the rest of you about how much trust you have in your automated modeling techniques. What percentage of your wagering is model driven?
Personally, for the last few years, the percentage of my plays that were model plays has for the most part always been greater than zero. This might be a little loosely defined. Let me explain. For a good chunk of that time, as I worked on more automated approaches (most of which failed miserably), I have at least had models that produced "for further study" lists. These could be as simple as a spot play that I liked which an automated screener would let me know I need to look at a certain race more closely. I have never been able to handicap more than 2-3 tracks per day manually, as I tended (still do) to overdo it and really dig into races spending forever on them. So the "for further study" lists were easy ways to hopefully pick up a few more bets on race types that I was doing well with at the time, or at least spending my time where there was a higher probability of generating a bet.
That progressed to me trying to develop more comprehensive models (which again, most have failed miserably). I always wanted to spend more time on them, but between data costs being high, simulcasting and ADWs being early and tough to use, and having a real job managing equity money, I never spent that much time. All of those things have changed, and I really enjoy the intellectual challenge (I'm always been a blowhard big thinker--executing on some grandiose plans has been the bigger challenge). Lately, I have been trying to spend more time getting some of the "what ifs" out of my head and either into a spreadsheet, database, or search server. As I have worked more on the research side, it has spurred more model building, and the current culmination of it is the model picks that you guys have seen me post over the last couple months. The first iteration was the model that I posted picks for for a week. Since then, I have made a few tweeks, but it is still basically the same model. It has continued to be profitable (but has had some terribly painful losing streaks), and I have tried to automate the whole process more so I could bet it and use my time more efficiently to work on improving it (I see lots of areas of work; many of you have helped me and given me your insights). The first part of that was automating the creation of the bets. Once I did that, I was betting it (small size) every day at a couple of ADWs or at the OTB. For the last couple of weeks, I have been trying to take that one step further, trying to not only automate the model, but the whole process from downloading of data to uploading of bets--almost there.
Sorry that I got a little off track, but it was somewhat necessary to explain the answer to the original question. As I started working more on this model, the percentage of my total bets has gone from approaching zero to approaching 100%. This was not so much because I thought it was better than what I was doing, but more that it was (and is) time consuming and I look at it as (hopefully) a good investment. As such, the percentage of model bets has gone up but my total handle has dropped precipitously. That has started to change a bit as I have more of the automation done and can get back to doing some other handicapping instead of spending time creating bets.
So I guess that's the long way of saying that my model percentage of plays is too high right now due to development, but I would hope that as it goes down somewhat as a percentage of plays, the percentage of handle will rise to a higher percentage. I guess that all depends on profitability, since that's what drives the the thing.
Enough of my rambling, how about the rest of you model builders? I'd love to hear from the rest of you about how much trust you have in your automated modeling techniques. What percentage of your wagering is model driven?