Idea for SNG software based on Gigabet's post

[sergsz]

One of the ideas coming out of gigabets's somewhat mystifying post was that sometimes it makes sense to take -ICM_EV play if the upside puts you in a position where you can use your stack to significantly increase your $EV and the downside doesn't hurt your $EV significantly. Of course, evaluating whether a a situation fits this criteria is very player-dependent: if you are awesome at wielding a big stack it makes more sense to take risks to build that stack than if you are somone who can't take advantage of a big stack as much. People may have a general idea of how well they play in different situations (big stack, short stack, etc.), but it would help to get an idea at what your actual equity of the prize pool is in different situations, based not on generic ICM, but rather an empirical reflection of your OWN past results in such situations.

Based on this, i think it would be useful to have a program that would read your pokertracker database, allow you to enter a certain filter (e.g.you are the big stack and have at least 10% edge over 2nd stack on lvl5 with <6 players left), and give you your average prize pool equity for all your SNGs that meet these criteria. Sample size would obviously be an issue, but for players with lots of SNGs in their database, i think this analysis would have a couple very useful implications:

1) You could look at marginal decisions in terms what your tournament equity would be if you win and what it would be if you lose, and compare it to folding. Similar to ICM, but your equity in various situations would be based on your history of playing in these situations, rather than generic ICM equity.
2) You could compare your historical equity in certains situations with the equity predicted by ICM, to see where your strengths and leaks may be.

Any thoughts on whether this would actually be useful?

I don't have programming experience unfortunately, but if people think this would be useful, i'm sure we could find someone on the forums who could write the program.

[eastbay]

There's been about 57,000 posts about hashing out a way to do this. Take a look for the issues involved. It isn't simple.

eastbay