Re: computer-go: Good Play (was FPGA)

[Amol Dharmadhikari <dharam2@xxxxxxxxx>]

> 
>    Sam Sloan wrote:
> 
>    > The average chess position has 27 possible legal moves of which no more
>    > than 6 or 7 are reasonable.
>    > 
>    > Almost every move is reasonable in the early stages of a game of go, except
>    > for moves on the edge.
> 
>    I disagree with that. The number is certainly higher than that of chess,
>    but 'almost every move' is much overdoing it. I think the 6 or 7 moves
>    in chess would correspond with 20 or 30 in Go.
> 
>    > I have heard estimates that God, if he exists, could give the top go
>    > players in the world 13 stones and beat them easily.
> 
>    The estimate I have heared most is 3 to 4 stones, rather than 13.
> 
> 
> 
> All  of this is   of course extremely subjective.   I  believe that we
> shouldn't  be  talking about  what   is "reasonable" without  defining
> reasonable.
> 
> I would define reasonable as any move which doesn't change the result.
> If the position  is   a win, only a    move that maintains  a  win  is
> reasonable and  if the position  is a draw,  no losing moves should be
> considered reasonable.
> 
> In chess, I don't believe that 6 or 7 moves are reasonable on average,
> at least by my definition.   I would say more  like 2 or 3.  There are
> way too  many positions where  only 1 move is   playable but of course
> this varies considerably from game to game.
> 
> Even though these are all just guesses, I am quite sure that 6 or 7 is
> too many.   This is of course difficult to prove!
> 
> 
> Don
> 

Well, u are assuming that there is just a single winning plan in any
position in which case there might be 6 to 7 moves(i dont agree on the
numbers but i dont have any clue about what they might be).

But in any given position, there might be multiple winning plans
depending on how one plays. A positional player like Petrosian will try
to deprieve his opponent of squares.....and try to strangle him. A more
"direct" approach would be taken by someone like Bobby Fischer who will
try to find a "break" or open up the position. Tal might try some more
speculative sacrifice. And i am sure there are much more than just 6 to 7
moves corresponding to each plan. 

Here we are not talking of a fixed sequence of moves(a mini forcing
combination) in between a game but we are taking a "high-level" view of
the entire game. 

In my opinion, it is entirely senseless to make these generalizations
about the number of moves without actual figures(even with actual figures
it is entirely subjective).

One more important thing: Even if the number of sensible moves is less,
that does not make Chess "easy" in comparision to "Go" or vice versa. 
This is like a computer saying that he has 5 branches to search instead
of 6 and so the "5 branches" game is easier. But what we are forgetting
is that a good heuristic function about the position(if we are able to
express it in quantitive terms!) will solve this problem as it will take
the shortest path to the solution in A* search...........


Regards,

Amol