Author: Albert Silver
Date: 16:33:24 01/25/02
Go up one level in this thread
On January 25, 2002 at 19:15:26, Dann Corbit wrote:
>On January 25, 2002 at 19:03:21, Albert Silver wrote:
>
>>On January 25, 2002 at 18:31:10, Dann Corbit wrote:
>>
>>>>There are two things I think you aren't
>>>>appreciating: first is Kasparov's non-propensity to make fatal mistakes. The key
>>>>word there is fatal. You are presuming that not only will the perfect player
>>>>will have forced winning sequences at hand at every move, but also that Kasparov
>>>>will forcibly make a fatal mistake. I think you are very strongly mistaken about
>>>>the number of non-losing moves. In many quiet positions the chances of him or
>>>>another player of his knowledge to make a fatal blunder is _extremely_ low IMHO.
>>>
>>>In this case, a blunder is making an imprecise move that costs you a pawn 30
>>>plies later and the game 100 full moves later. I don't think Kasparov (or any
>>>of the others) can see it.
>>
>>Sometimes even being up a bishop and a pawn isn't enough, or a queen versus a
>>rook+pawn, etc. The ways to not lose are many times many.
>>
>>Ok, let me put it differently. Let's say that in a given position there are 35
>>possible moves, of which 32 are forcibly losing (and I'm being ridiculously
>>harsh IMO).
>>
>>20 lose material right away. The chance of a top player playing them is somewhat
>>slim.
>>
>>7 lose at a depth of 10-15 moves (by this I mean that the poor human will
>>recognize they are lost 10 moves later). The 10-15 move ones are positional
>>blunders (or very deep tactics of course). Most likely the top player will avoid
>>these too, though their chances rise somewhat. These are moves like an illogical
>>h4 that may take time to exploit, but can definitely be forced to a win.
>>
>>And 5 at 30 moves. The 30 move ones are what we would call positional errors
>>(VERY deep tactical mistakes) that can also be exploited but the road to a
>>forced win is very narrow.
>>
>>And finally you have a miserable 3 moves that don't lose forcibly.
>
>Maybe 2 of them *do* lose forcibly. Or maybe 1 wins, one loses, and one draws.
>Or maybe all 3 can still win.
>
>>I don't think
>>that it is so hard for a world-class player to hit on one of these non-losing
>>moves.
>
>If they are non-losing. You really mean "don't lose immediately." The limit of
>depth that we can see determines what these moves are. If you can see one
>single ply ahead, then every move is one that does not lose immediately. If
>you can see 7 full moves ahead, then maybe 2 or 3. But if you see 20 full moves
>ahead, then maybe all 3 of the "good moves" are losing and one of the ones that
>looks horrible at 7 full moves is actually the winning move. And when you look
>ahead 100 full plies, it changes again.
No, the figures I gave were my belief on perfect play. The three moves are what
I _pessimistically_ believe would be the number of non-losing moves with perfect
play of course. Hence my comment on what a perfect player's assessment of the
position might look like:
"Usually they are quite logical to boot, though the reasons the imperfect
human will give (occupies the column, development, control over e5 square, etc.)
are very different from that of the perfect player (Perfect DB says that Re1,
Bd2 and Qe1 are 0.00)."
Ok, here's a question that would be interesting to answer: Take all the
tablebases and see what the numbers (quantity and percentile stats) of
non-losing moves in positions where a non-losing move exists. And if possible, a
breakdown of those numbers to compare between numbers of pieces on the board.
Just to see whether a tendency in changes of stats according to the number of
pieces (3-4-5-6) is detectable.
Albert
>
>>Usually they are quite logical to boot, though the reasons the imperfect
>>human will give (occupies the column, development, control over e5 square, etc.)
>>are very different from that of the perfect player (Perfect DB says that Re1,
>>Bd2 and Qe1 are 0.00). You present this as though the non-losing moves are not
>>only extremely few, but impossibly hard to find (i.e. must look illogical
>>according to our understanding of the game). I think otherwise. I don't think
>>playing perfectly will change this either and believe this about the game of
>>chess as an absolute. I think perfect play from both sides would lead to a draw,
>>and also believe that the number of paths to that perfect draw is enormous.
>
>Along with the number of paths to losses and wins, which are also astronomical.
>I just don't believe that an imperfect player will be able to find them against
>a perfect player.
>
>>Perhaps most moves (60+%) are losing because most moves drop a piece or pawn in
>>one move, but I don't think that makes them likely candidates for a WC player to
>>play.
>
>Then there's the Evergreen game.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.