Well, I've always thought of hyper regs as of the biggest gamblers among poker regs
And actually breakeven stretches are significantly winning ones due to FPPs... a future good run (we don't know if it will happen - remember, the future doesn't depend on the past - what we can do is play well) will be just a bonus but the life won't threatened in almost any case.
Quote:
Originally Posted by streityboy
I appreciate that English may not be your first language (heh even with my surname may German isn't hot at all!) but I don't think you understand the point I was trying to make.
It's a basic maths formula. The formula is correct. I wasn't suggesting the calculation was incorrect at all.
First of all, ICM is only a model. Even if it gives correct finishing position probabilities in the all-in shootout setting (which I'm not convinced of, except for the first place), it doesn't evaluate them precisely in the real setting because players' strategies differ.
Secondly, no modern calculator can compute true ICM equities as such (the 'true' ICM is based on imagining a 'tourney' where chips in players' stacks are removed one by one in a random order) remotely fast. Instead, different approximations are used.
And there's a dispute on which approximation works the best. E.g. there's a
discussion in STTF right now about benefits of Ben Roberts' algorithm over the standard Malmuth-Harville (the one you likely use, that assumes that the conditional probab. of players B and C finishing 2nd if A wins are proportional to their stacks etc.) and the Malmuth-Weitzman (assuming that the probab. of busting out next are inverse proportional to players' stacks and distributing the busted player's stack evenly among the remaining ones).
Quote:
Originally Posted by streityboy
What I was saying, and it may not have come across very well, was that the use and application of it is troublesome.
1. You cannot do an ICM calc at the table
Spot on. 'Post-mortem' Nash equilibrium calcs to learn from mistakes in session reviews are possible for NLHE, but there are so many starting hands in O8 that the brute force algorithm (accounting equities of all possible starting hands against each other) isn't feasible for the Nash equilibrium. Most likely, if an ICM Nash calc for O8 is ever written (which is one of my dreams
), there will be a small number of 'key' starting hands (like key frames in digital videos) and an actual hand will be replaced by the closest (in terms of equity distributions vs ranges) key one.
It's easy to do ICM EV calcs vs concrete hands shown down (which HM2/PT4 do) and vs user-defined ranges (which can be done by hand post-mortem because scoop and split probabilities are all computable), but it's hard to figure out 'unexploitable ranges' of all players, even HU.
Quote:
Originally Posted by streityboy
2. The accuracy of it's calculation requires very specific knowledge of the range villain is pushing. Not easy versus randoms.
3. It cannot take into account blinds levels. e.g you have the same marginal $cev decision at 15/30 and your M is say 15 versus the same situation at 75/150 and your M is less than 3. ICM would tell you fold in both instances. If I recall, MaCros made an interesting post on this a couple of years ago from a thread by Bakya.
4. It cannot take into account the speed of the blind levels. You have a marginal $cev decision with a blind structure that goes at 10 minute blind levels as opposed to 2 minute blind levels. ICM would dictate you fold in both instances.
You see, what I am trying to say is that it's application in real time is difficult and that there are very significant variables that are omitted in it's calculation. It forms an important part of a lot decisions, but it isn't the definitive answer.
All these points are valid (flipya4dinna was wrong about accounting the blinds). NLHE calculators have the 'future game simulation' feature (FGS) that sees 1-5 hands ahead, and hence somewhat accounts for posting blinds and antes and their size, but it delivers a significant improvement only in the case of extremely short stacks and rather flat payout structures (DONs). Of course FGS for O8 isn't achievable on modern PCs - even for NLHE, FGS seeing 4-5 3-max hands ahead takes ages.
Bottom line: don't worry about remotely exact ICM Nash ranges; maybe study toy examples once in a while, but not more than that.