> Both.
> Its probably not so difficult to make a simple bot. But it is also not
> difficult to make a simple UCT player. But I am sure, that reaching the
> level of Polaris is more difficult than writing the best Go-programm. I have
> the feeling, that Polaris is a very serious project. Its certainly not
> possible to beat it "out from nothing" like Crazy Stone and MoGo have beaten
> the Go programms. There is also a lot of work in these 2 programms too and
> it is not really "out of nothing". But its nevertheless not comparable to
> the work the Billings-group has done. There is also a very large gap between
> Polaris and the "rest". Without Polaris, everybody would say: Oh, its as
> difficult as Go, the programms are in relation to humans at about the same
> level. And now Polaris is strong and the argument is: This is because Poker
> is much easier. No, they have done a better job.
i think that you might be confusing two important things:
i) the difficulty of a problem.
ii) the amount and kind of effort that has gone toward solving a problem.
people have been playing go for (depending upon how you judge the gaps)
a few thousand years, and yet some of the biggest advances in opening
theory have happened in the last fifty years. probably there are many more
significant advances that can be made in the opening and the middle game.
can the same be said for poker? aside from the (arguably interesting, but
perhaps not complicated) fact that your opponent is allowed to misrepresent
his situation, a computer program really just has a few simple inputs to deal
with -- those cards that it can see, and those bets/folds that people have made.
the total number of complete games that a poker program might be expected
to play is based upon the number of different cards that it can be expected
to see, the maximum number of choices that it may have to make, the
number of different bets (or categories of differently-sized bets) that its
opponents
can make, and perhaps the total number of different opponents that it might be
expected to play and where each of those players are seated at the table.
i think that it's clear that the size of the problem is smaller (and that our
ability to measure "being good at the game" is less clearly defined) than go.
imperfect information does not necessarily mean that a problem is harder.
(just as perfect information does not necessarily make a problem hard).
if you (say) flip a coin ten thousand times and keep track of the number
of heads and tails that you get, i can guess that number to reasonable accuracy
even though i have absolutely no information about the actual value other than
the process that you used to generate it.
if i were to place bets after each and every flip, i could lose all kinds of
money
playing this game, but that wouldn't mean that i didn't have a perfect strategy.
one confusing thing about measuring the ability of a computer poker player play
is that even if it loses ten times in a row, it might be the best player at the
table.
the very best poker players in the world do not consistently win championships,
because (as i believe someone else (jacques?) said) the variance is so large.
imagine trying to set up ELO rankings for poker players and what would happen
to everyone's ranking after each tournament. you could probably establish some
broad categories (poor, good, better, best), but it would be difficult to
establish
exact rankings inside these categories. this doesn't mean that the game is
harder, it means that our ability to determine skill is very impaired. go, on
the
other hand, has, arguably, over 35 independent skill levels, and determining
which of these your program is at is a quite simple task. so it's quite easy to
measure how successful modern computer go players are. how would we do
the same for computer poker players? what's a good measure? what would
perfect play look like, and what would the variance in win/loss rate look like
against human players? for that matter, what would the variance be among a
table composed entirely of "perfect" computer players?
s.
____________________________________________________________________________________
Shape Yahoo! in your own image. Join our Network Research Panel today!
http://surveylink.yahoo.com/gmrs/yahoo_panel_invite.asp?a=7
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
