KOTS:
I think it is a mistake to work on just one axis like that, simply based on your perceived ROI. I think it is necessary to add a "reliability" axis to your trigger-pulling.
For example, suppose you have a really good concept of how often a Pick'em team holds on to win when they hold 7 pt lead after the first quarter. For the sake of this example (i.e. <u>WARNING: intentionally bogus numbers used for this example, do not reuse them they are worthless</u>), you know that teams leading by a TD hold on to win 74% of the time, but the interactive market consistently prices it 65-70 allowing you to buy at $70 and make a solid profit. In this case, your 74% estimate is based on hundreds, even thousands of samples and you are very, very certain. However, using a proportional system like you describe, you might only get one unit on this play.
In contrast, when you are using your simulation to model the scenario in the other thread, where you value TB +2.5 at $87 when they are receiving the kickoff ahead by 3 points, you would be buying TB into a 70-75 market. In this scenario, I (and the rest of the jury in the thread) think you are way off and would be making a mistake by purchasing TB for $75. And with the enormous price gap, it is likely that you are going to get to your 3-"pop" maximum at one point or another on this one, resulting in greater bets on the a questionable play.
 |