[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: SV: [computer-go] Score estimating
In message <LMENKGJEFIOCOJLOMOFOEEMPCAAA.ulla_kjeld@xxxxxxxxxxxxxxxxx>,
Ulla & Kjeld <ulla_kjeld@xxxxxxxxxxxxxxxxx> writes
Well score estimation is one of the key issue about making a computer
program that will play Go at a good level.
I'm working on some ideas on how to calculate the score.
My idea is that you divide the score by the numbers of stones. This give you
a rate of how much every stone was worth. This can be any real number, but
mostly it will come around 0.5 to 1 (best guess, could be found by
calculating some games)
What do you mean by "score"? My guess is that you mean the territory
surrounded by one player, plus the prisoners captured by that player.
I would think that the value/stone in the game goes up and down as the game
gones on. In the beginning a stone will be about 5-15 point worth, but later
it will only be 0 to 2 point worth. (could also be negative if you make bad
moves)
At the start of the game, the value of a move is twice the correct komi.
I.e., around 15 points.
But if is possible to calculate the rate between score and the number of
stone at the end of the game, I think it should be possible to do it
halfways in the game too.
It should be possible. Why do you think it will be useful?
The rate of a stone is put together from the number of stones around the
stone and number of free points also.
But so far, I have not made a complete algorithem.
Nick
--
Nick Wedd nick@xxxxxxxxxxxxxxxxx
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/