Computational Intelligence - November 2012 - 11

roughly 1.5), the best performance so
far. For 13x13 kill-all Go, in Tainan in
2011, MoGoTW could win as White
with H8 against Ping-Chiang Chou
(5P), which is seemingly the best performance so far.
❏ 19x19: Humans are still stronger
than computers, in particular for the
big 19x19 board where strategic elements matter a lot. 19x19 Go is the
main version of the game of Go. The
thirteen 19x19 games' results show
that the performance so far is a win
with H4 when Zen was against a 9P
player. The player (Chun-Hsun
Chou) said that he miscounted the
compensation and so he believed
that the game was an easy win. When
he understood his mistake, it was too
late for trying to win. On the other
hand, Joanne Missingham (6P) was in
a very difficult situation in her H4
game against Zen; she played a move
for cheating Zen (a very suboptimal
move, aimed at making the situation
difficult) and Zen er roneously
resigned, whereas the situation was
admittedly a win for Zen. There are
still 4 handicap stones, giving to the
program an advantage in each corner, so professional Go players mentioned that switching to handicap 3
is a huge challenge. In particular,
computers still have weaknesses for
combining multiple local fights.
Remarkably, Many Faces of Go won
as Black with komi 3.5, i.e. roughly a

handicap 0.5, against a 6D player
(Sheng-Shu Chang).
❏ Level assessment in Go: Parts
above were dedicated to the assessment of programs strength, and to
physiological measurement. We here
consider a different goal, for Go programs: to evaluate the level of the
human opponents. Simple tools for
level assessment in Go consist in
playing several games, increasing the
strength of the computer when the
computer loses, and decreasing it
when the computer wins. Then, the
level of the human is estimated using
a classical Bayes analysis using the
ELO model [7]. We performed such
tests on players from kyu level to 6D,
using MoGoTW playing on many
computers in Taiwan in May 2012.
We have then seen that in 5 games,
we can estimate the strength of play-

ers somehow efficiently. This has
direct applications for eLearning
because evaluating Go players (by
playing games) is similar to evaluating students (by proposing exercises).
This time, in June 2012, each player
also played only 5 games and we
tried to compare three 6D players, a
difficult task because they have a very
similar strength. The games' results
show that MoGoTW could nonetheless find out that Shang-Rong Tsai
(6D) is a particularly strong 6D player. This shows that our artificial player is also able to analyze the strength
of strong players. Importantly, along
with this ability to evaluate the
strength of humans, the computer
can adaptively adjust its strength to
the opponent, in order to increase
entertainment, which is helpful for
motivating children to learn.

Hussein Abbass (background, right 2) and Chang-Shing Lee (background, left 2) with Shi-Jim
Yen (foreground, left 1) measuring the physiological signals.

Participants attending the opening ceremony, including Gary Fogel, Hisao Ishibuchi, Hussein Abbass, and Marios Polycarpou (foreground, left 1,
left 5, left 7, and right 5, respectively).

november 2012 | Ieee ComputatIonal IntellIgenCe magazIne

Table of Contents for the Digital Edition of Computational Intelligence - November 2012