I haven't gotten to read through your paper as carefully as I'd like yet,
but I do have a few observations that may be of benefit to other readers on
the list... Mostly observation of assumptions used in the paper.
1. A max tree is used instead of minimax
2. Rewards can be more than just 1 or 0 (
> It is clear that in
> professional play 2 handicap stones is overwhelming.
Kageyama mentioned a student who had been playing him at a small handicap
and winning. The student didn't think he could lose a game and nine
stones. So they played a nine stone game; Kageyama kicked his butt and
says the
On Tue, 24 Apr 2007, Jacques BasaldĂșa wrote:
Christoph Birk wrote:
I am sure that Daniel is wrong here ... 2 kyu difference is more like
80% likelyhood of win.
That depends on strength. Between a 20 and 22 kyu, it is even lower. But in
professional
play Daniel should be right. Note that 2 steps
Hi,
I have been encouraged to post a message to this list because we wrote a paper
on the analysis of UCT and other bandit-based methods for tree search, and
this might be useful for computer-go.
The paper "Bandit Algorithms for Tree Search" is available as an INRIA
technical report at: http:
Christoph Birk wrote:
> I am sure that Daniel is wrong here ... 2 kyu difference is more like
> 80% likelyhood of win.
That depends on strength. Between a 20 and 22 kyu, it is even lower. But
in professional
play Daniel should be right. Note that 2 steps means 2 stones handicap.
It is clear th