1. Mura - Logi 16 - 48
2. Logi - Mura 48 - 16
3. Mura - Logi 26 - 38
So Logistello is leading by 3:0 after half of the games.
Ingo Althoefer.
In the last few days a six game Othello match was played in Princeton between
human World Champion Takeshi Murakani and the top computer program Logistello
( author Michael Buro ). Never before such a match was played in Othello.
Each side had 120 minutes for each whole game.
Logistello was running on a Pentium Pro 233 MHz. The results of the games are
( the more points a player has the better for him; for the match score the
points do not play a role - in each game the one with more points is winner
of that game ):
1. Mura - Logi 16 - 48
2. Logi - Mura 48 - 16
3. Mura - Logi 26 - 38
4. Logi - Mura 55 - 9
5. Mura - Logi 26 - 38
6. Logi - Mura 37 - 27
So Logistello became the winner with a 100 percent score. 6:0 !
Congratulations to Mic Buro !
And Mic, please make your program commercially available.
Ingo Althoefer.
PS: More infos may be found at
http://www.neci.nj.com/homepages/mic/event.html
: Question for the Othello experts: Was this (the 6 - 0 score) totally
: unexpected?
: Mark
Not really. othello programs have been quite good for some time. And
there's been a good bit of publication by the authors detailing what they
did. There's a paper we use in our AI course here (where the final
project is an othello program, and we end with a class tournament) called
something like "how to write a world championship othello program" or
something similar. It gives a *lot* of details...
>Ingo Althoefer.
This is an incredible win in my opinion. I also hope that Logistello
becomes commercially available.
Question for the Othello experts: Was this (the 6 - 0 score) totally
unexpected?
Mark
also the above url should be:
http://www.neci.nj.nec.com/homepages/mic/event.html
Not in the least. It would have been surprising if the World Champion
had beaten Logistello in even a single game.
The top programs in othello are roughly as far ahead of the best humans
as Kasparov is ahead of a 2000 player. Here are some of the reasons:
- The programs can solve endgames exactly, by search, surprisingly
early. If there are 22 moves left, then Logistello has already solved
the game; it usually does it earlier than that. For a human, this is
the late middlegame; for the program, the game is already over and
the only question is whether the opponent will blunder.
- The programs use sophisticated evaluation functions which
are tuned by machine learning. They're accurate.
- Othello is somewhat hard for humans. Because each move can flip
a lot of disks and cause big changes on the board, it's hard
to visualize far ahead. Programs don't have any difficulty.
Jay Scott <j...@forum.swarthmore.edu>
Machine Learning in Games:
http://forum.swarthmore.edu/~jay/learn-game/index.html
I have a paper on my hard drive: "Keyano Unplugged: the Construction of an Othello Program". (I have a few others though they deal with only a specific portion of an Othello Program, e.g. "Experiments with Multi-ProbCut and a new High Quality Evaluation Function for Othello".)
Can you tell me what paper your class refers to, and where it is available via ftp? :-)
It should be
http://www.neci.nj.nec.com/homepages/mic/event.html
/ Gunnar Andersson
Jay Scott wrote:
>Mark Rawlings wrote:
>>Question for the Othello experts: Was this (the 6 - 0 score) totally
>>unexpected?
>Not in the least. It would have been surprising if the World Champion
>had beaten Logistello in even a single game.
I fully agree. However, the match of Logistello versus the human World
Champion had its justification. It was the first such match. And until
August 3, 1997, there were several ( strong ) human Othello players who
claimed that man were stronger than machine, if the playing conditions were
fair and if money was at stake. Both conditions were met in this match.
But, perhaps we will see a development like in checkers ( where Jonathan
Schaeffers program Chinook is top and better than all humans): there may
always remain some humans who claim to be stronger than the machine(s).
>The top programs in othello are roughly as far ahead of the best humans
>as Kasparov is ahead of a 2000 player. Here are some of the reasons:
>- The programs can solve endgames exactly, by search, surprisingly
>early. If there are 22 moves left, then Logistello has already solved
>the game; it usually does it earlier than that. For a human, this is
>the late middlegame; for the program, the game is already over and
>the only question is whether the opponent will blunder.
>- The programs use sophisticated evaluation functions which
>are tuned by machine learning. They're accurate.
>- Othello is somewhat hard for humans. Because each move can flip
>a lot of disks and cause big changes on the board, it's hard
>to visualize far ahead. Programs don't have any difficulty.
Ingo Althoefer.
The correct url is
http://www.neci.nj.nec.com/homepages/mic/event.html
Sorry for the typo !
Ingo Althoefer.