Gentlemen,
My real-life money management job has kept me busy recently, so I apologize for the lack of posts. Let me try to address some of the recent posts with a summary:
You can't measure 10 events of anything and use that as a basis for logical regression. 100 events or 1000 would be a better number, and the formula has worked over two seasons now. I know that may be contrary to the way gamblers think, but my approach is like investing, and you have to be prepared for the long run. You don't stop simply because probabilities catch up with you.
And, as I've said repeatedly, we ARE using current year data. There is some data from last year in the database also, but of course that factored into our successes from April to September. The question to ask is whether there is a place for previous season data early in the year, but at some point it is no longer relevant, or whether there is no place for it, or whether it should have a diminishing weight in the db. My experience last season gave no indication that the presence of previous season data was problematic.
To those wondering about the bankroll, I've said several times that the bankroll doubled early in the season, but that was due to betting an disproportionate percentage of the total. It was undue risk that I decided to back off of in terms of unit %, which got me to the 2% I've mentioned previously.
Also, I've bet on alot more games than those I posted here, plus NFL games, and I never adjusted my unit size upwards as the bankroll grew. Finally, I've been fortunate enough to be able to add to the bankroll with additional contributions (the goal being quitting the day job by MLB season end). Yes, a few of the larger bets I have had to spread across multiple books. I suspect many of you have not limited your bets to the MLB underdogs I've posted either.
Obviously those who started significantly later in the season would have a different experience than someone who started on day 1, as is the case for anyone who invests at different times and durations. The guy who invested with me 10 years ago has done alot better than the guy who started last month. That's just how it works.
Also, openbahr and others of you who still think this is a progressive 'double' system still don't understand-go back and read again. I don't mind anyone critiquing the methods, but hopefully we can stick with constructive ideas grounded in accuracy and objectivity.
I have run enough numbers now that I think for next year I will abandon the progressive approach altogether, and simply bet more games with a straight unit approach. Looking back at this season there are many days where a 2nd or 3rd game could be bet with reasonable accuracy, and this provides the opportunity to maintain the profit level while avoiding the risk of the progressive approach. It could also be that the progressive approach should be scaled to a certain unit size or certain number of games, an approach that might take advantage of the high incidence of two game losses without getting too big on unit size commitment.
Additionally, some of you have emailed me offline asking why the formula couldn't identify favorites with a reasonable ML, and I've reconisdered that approach and think it is possible, although I haven't backtested it. The idea would be to identify favorites who have a 60% probability of winning and who are not suffering from a severe ML disadvantage (as is so often the case). If you can't pick favs at 60% or better, I don't think it's a good value proposition.
Mark
Hot Forum Topics
