I'm home from the WSOP now, so I figured I'd post a couple of thoughts about this thread.
1) Pokerstove is a valuable tool for calculating EVs for exploitive play, but when trying to do calculation of exploitation-proof strategies I think it is generally not very useful. If you want to do this kind of thing seriously you need tools that are more flexibly automated, which means some programming language. My recommendation is Python, which is fairly easy to use, and for which a fast suite of poker evaluation tools has already been ported.
2) I did some work on the effect of card removal (from folded hands) on jam-or-fold strategy between the blinds a while ago. I found one surprising thing: that the problem of finding the hand distributions of the blinds given that the first n players have folded a specified set of distributions is NP-hard, which means that simulations such as the ones Barry has done are as good a method as any. When I did this, I found that there was very little change to JoF strategies under 10 BB, which is what I was concerned with at the time. If someone really desires it, I can probably dig up the old work and have a look-see at larger stacks.
3) I agree with the following points made throughout the thread, or which I made up:
- If you are going to fold K8 here, you should jam instead.
- You probably should not open-jam many hands at this stack size, and K8 is probably not one of them.
- If you are going to raise a small amount, it may be cleaner to assume your opponent always shoves, because then you can use pokerstove to figure out your equity. However, in real life, your opponent should call a lot and play the pot with 19 BB stacks in position. This is harder to analyze without making up a bunch of assumptions; the clean certainty is removed because there are all these streets and flops and so on.
4) Re: Barry's summary:
Using math to analyze postflop play is hard and pokerstove doesn't help. So he is right that in these situations where there is a choice between a certain mathematical alternative (JAM, because we are 100% sure it's +EV), and a uncertain one (raise to 3x, because we think we are making money from the combination of preflop and postflop play), it is often correct that the uncertain one has higher EV, and so the first fact isn't that useful. So Barry is right in that sense.
But the kind of analysis presented here (you can jam for 22 BB with K8 because it's +EV) isn't exactly the state of the art, and as we keep playing poker, the set of situations which are mathematically "solved" is going to grow until a lot of what Barry calls "poker considerations" are mere inputs or slight adjustments to known strategies. So it's a dangerous idea to think that "judgment and poker considerations" are going to provide a durable, persistent edge for players playing in tough games, just as it was dangerous to think that "fit or fold" was a reasonable way to play shorthanded limit holdem back in the day - even though it probably worked. Since
AFAIK all of the mathematical published NL analysis basically concerns preflop play, this state of affairs may last for a little while, but I wouldn't count on it remaining that way.