Mark Gritter (markgritter) wrote,
Mark Gritter

Last Round Poker Play as Linear Programming

(I wish I remembered more from my graduate algorithms class. Unfortunately we didn't have a textbook or even a course reader, just whatever Serge Plotkin felt like talking about that day.)

I've been trying to think about how to solve last-round play in 2-7 lowball when both players draw one. (I peeked ahead in Chen and Ankenman and they do solve the [0, 1] game for a single street and multiple raises but I didn't find anything about applying the results to a discontinuous game yet. Still, it will be interesting working through to that point.)

It occurred to me that finding the optimal play against a particular opponent strategy is just a linear programming problem. Not that I expect this to be an original insight or anything.

Player A's strategy is defined as, for each card he draws, a probability distribution of each of his possible lines: check/fold, check/call, check/raise/fold, etc. (If you are interested in this sort of thing, check out my strategy calculator allowing only a single line per card.) This can be expressed as a set of variables xc,l where c is the rank of the card drawn, l is the particular line.

The constraint is, of course, that for each c the sum over l of xc,l = 1. This can be normalized by giving a "default" line which is used whenever one of the other lines is not, so that the constraints are all of the form

x2,KF + x2,KC + ... + x2,BRC <= 1

And note that all the x's are >= 0.

Player B's strategy is defined the same way but his preferences must be treated as constants to yield a linear system.

The function to be maximized is, of course, A's expected value. This is just the sum over all terms of the form P(c && d) * xc,l * yd,m * V(c,d,l,m) --- the probablity that A will get card rank c and B will get card rank d, times the probability that A will have picked strategy l and B will have picked strategy m, times the outcome from A's perspective of those choices. One of the xl's is actually 1 minus the sum of all the other lines due to our normalization above but that's still a linear term. If we treat the y's as constants then this is linear. Easy!

Unfortunately it's not clear to me how to solve for both A and B simultaneously; that appears to be inherently nonlinear. Maybe after I finish The Mathematics of Poker.
Tags: poker, theory
  • Post a new comment


    default userpic

    Your reply will be screened

    Your IP address will be recorded 

    When you submit the form an invisible reCAPTCHA check will be performed.
    You must follow the Privacy Policy and Google Terms of use.