Don, please help me understand

If an index number indicates a deviation from basic strategy, why would a simulation using only this one deviation from BS ever result in a negative expectation? For example, using SBA for my simulations, 4.5/6, DAS,leave at -2, spread 1-10, SCORE is 18.76. Adding insurance only, SCORE is 22.77, etc. But when adding only A,8 vs 6 at +1, SCORE falls to 16.61. Similar results for a few other plays. I have SBA set to run 900 million hands so I assume the results are fairly accurate. Either I am doing something wrong or I just don't understand (probably both). Please advise.

I'm sorry, but this makes no sense. John Auston did some runs for me, adding one I18 change at a time, but, unfortunately, as you know, this isn't one of them. General principle: being smarter can't cost you money!

However, as you add a double to your repertoire, have you redone the unit size, for optimal betting? To get the SCOREs, you aren't using the same units every time, are you, as you change plays?

Don

...is an idea I had and Wattenberger did find some similar results, back then. I had in mind the possibility that this can happen but never found a way to predict when an index might be more wrong then right. This will NEVER happen however with an index that has a correlation of over 50% to the precise effects of removal in a given game. But most of the time indexes of less than 50% (note also that I am speaking of the absolute value of the CC here too) will still show expected value in their use and WILL NOT result in such losses.

I am guessing as well here then, but you would not lose much by tossing some indexes in the ranges 0 to 50% CC, and tossing any index that shows this behavior.

You are at the frontiers here.

You are doing something else wrong too!

Thanks all for you replys and advice. I have just noticed that my original post showed a SCORE of 16.61 for A,8 v 6. Should have been 18.61. Sorry for the confusion.

Don, my bet spread is +1 one unit, +2 two units, +3 four units, +4 six units, +5 ten units. I did my simulations using this spread for all plays, including doubles. Obviously this is wrong. I'll make an adjustment and start over. Thanks for your reply.

If I understand you correctly, the SCORE is going down from 18.76 to 18.61. That's a pretty minor difference. I'm also assuming you're set to S17. SCORE is weighted for variance. Is the risk averse index perhaps a little higher than +1?

ETF

"SCORE is weighted for variance. Is the risk averse index perhaps a little higher than +1?"

The EV-maximizing index is actually 0.7, according to Friedman. At a 2% bet, the r-a index jumps to 2.0, but realize that the index is only +1, so we don't have that large a bet out, usually.

If SCORE is done properly, and r-a indices are not used, it adjusts the bet (unit) sizes to make up for the increased variance. So, SCORE has to be higher for using any kind of index, but with soft doubling, the play occurs so infrequently that the difference might not show up until a huge sim is run.

Don

Even with a typo from the indexes used. You shouldn't get this much even making a totally wrong play for that decision. Just look at the Griffin TOB tables for importance of decisions. Even a totally poison index would not be this extreme.