Originally Posted by
Justin7
Ahhh. You fell right into my trap!
If you bet a game at +3, and it moves to +3.5, you need to re-evaluate whether your edge was truly 2%. In general, if the market disagrees with you, you are screwed. I would argue that you shouldn't bet more at +3.5. If I had a fair way to get rid of my +3 bet at that point, I would.
Same thing applies if you bet at 2.0, and the odds move to 2.05. The market just told you that you were wrong...
(You're treading dangerously close to begging the question here ...)
In quantitative modeling and programming one often hears the phrase "Garbage in, garbage out." This refers to the fact if you provide a model with shoddy input data, your conclusions will be similarly unreliable.
Similarly, this methodology makes no assumptions whatsoever about expectations, rather it takes as input operator specified return estimates. This, after all, is a a risk/return optimization model and not a return forecast model.
I wish you had initially phrased the problem as you did in your last post. For one thing, it would have saved me the effort of demonstrating the calculations. And for another, the problem as you respecified, which demonstrates an overbet, is actually just the sort of problem designed to be solved by the attached spreadsheet. (Although I do still need to include the facility for handling correlations < |1| for it to be strictly applicable. This is just a UI issue.)
The idea is quiet simple. This is the general case:
A player makes a bet and at some time later time determines that he bet too much. The rationale for coming to this determination is completely irrelevant. It could just be because he accidentally added an extra zero to his bet size, because the line moves in his favor, because the line moves against him, because the line doesn't move at all, because it starts raining, because it doesn't start raining, or because his second-cousin-once-removed tells him he thinks the bet is definitely going to lose. The point is that return estimates are 100% exogenous to the model.
Once the determination is made based upon current return estimates (whether or not they've changed since the bet was initially placed) that the player has bet more than the optimal quantity, the spreadsheet can then be used to determine how much (if any) of the bet would be laid off at the current price.
So that's how it works. If you provide accurate return and/or probability estimates you'll get accurate results. If your estimates are inaccurate, your results will be inaccurate.
Garbage in, garbage out.