Author: Robert Sherman
Date: 19:01:54 11/10/97
Go up one level in this thread
On November 10, 1997 at 08:37:11, Robert Hyatt wrote:
>On November 09, 1997 at 23:56:48, Robert Sherman wrote:
>
>>On November 09, 1997 at 17:00:03, Robert Hyatt wrote:
>>
>>>On November 09, 1997 at 16:14:30, Robert Sherman wrote:
>>>
>>>>On November 09, 1997 at 13:50:58, Jack Leverette wrote:
>>>>
>>>>>Dear Rob Sherman:
>>>>>This isn't right. 4. ...Nc6. was move 2.
>>>>>I believe the move is 4. .. Nf6.
>>>>>
>>>>>Yours,
>>>>>The AceyMan
>>>>
>>>>Sorry I typed it in fast from memory. Your right it should be 4....Nf6.
>>>>
>>>>Robert Sherman
>>>
>>>I ran this thru Crafty at one minute per move on my P5 notebook. First,
>>>here is the annotated game score:
>>>
>>> 1. e4 e5
>>> 2. Nf3 Nc6
>>> 3. Bc4 Bc5
>>> 4. Nc3 Nf6
>>> 5. Nd5 Nxe4
>>> {9:+0.70 Nxe4 Qe2 Nxf2 Rf1 Ng4 h3 Nf6 Nxe5 Nxd5 Bxd5}
>>> 6. d3 Nxf2
>>> {9:+5.24 Nxf2 Qe2 Nxh1 Nxe5 Bf2+ Kd1 Nd4 Qf1 f6 Nxd7
>>>Bxd7 Qxh1}
>>> 7. Qe2 Nxh1
>>> {8:+4.94 Nxh1 Nxe5 Nd4 Nxc7+ Qxc7 Bxf7+ Ke7 Ng6+ Kxf7
>>>Nxh8+ Kf8 Qf1+ Kg8 Qxh1 Kxh8}
>>> 8. Ng5 O-O
>>> {8:-2.14 O-O Qe4 Bf2+ Kd1 Qxg5 Bxg5 b5 Nf6+ gxf6 Bxf7+
>>>Rxf7}
>>> {8:+5.74 f5 Nf7 Kxf7 Nxc7+ d5 Bxd5+ Kg6 Nxa8 Bf2+ Kd1
>>>Qxd5}
>>
>>Wow! I am surprised at the difference in score here. I have tested this
>>game with almost all the top commercial programs (Genius, Mchess, Fritz,
>>Rebel etc. and they all pick O-O over f5. While a couple of IM's prefer
>>f5. I think this could put an end to DB not playing Qb6 conspiracy
>>theory. Very human-like approach of not walking into an attack although
>>I don't think your scores are accurate.
>
>depends on your definition of "accurate score". *my* scores don't
>reflect
>material only, they reflect lots of other things. Particularly if it is
>a
>O-O type move. I compare safety after O-O with the potential safety
>after
>O-O-O, and if O-O-O would be much better, then O-O gets a *big* penalty
>to
>make it want to defer castling. The penalty is proportional to the
>difference val(O-O-O) - val(O-O). Without some such trick, a program
>will
>castle right into an attack because: (a) most programs receive a bonus
>for
>castling, to make it castle *now* (before it gets into trouble by being
>trapped in the center); (b) most programs want to castle *period*, and
>if
>castling long is difficult, or needs time to prepare, *now* looks
>better.
>
>And it can get you crushed...
>
>self-crushed might be a better term...
>
>
>
I am impressed that Crafty steers itself out of trouble. But a
seven-point swing for a computer program based on positional safety of
the king is incredible. It certainly makes that Deep Blue Be4 move
totally normal because computer programs were only a point or two
different. Your score is 7-8 points different from all the commercial
programs that I have seen.
>>> 9. Nxc7 Qxc7
>>> {9:+2.57 Qxc7 Qh5 h6 Bxf7+ Kh8 Qg6 Bb4+ Kf1 Ng3+ hxg3
>>Odd doesn't walk into an attack but walks into a mate in 6. Again
>>Crafty show that it is possible for a computer to be scared of an attack
>>but still miss a deep mate.
>
>note this was run for a short time on a P75 notebook. I wasn't really
>hot on the trail of deep tactics, rather I was interested in making sure
>it wouldn't impale itself on white's attack... I can see what a deeper
>search will see if you like...
>
>
I would very much like you to try a deeper search. I would like to see
if there is any tactical merit or was it only Crafty's intuition :-) I
would like to see you prove the analysis of all the commercial programs
wrong. I mean it is not every day that you see such a divergence is
computer evaluation. Komputer Korner likes his scores exact you know
:-) He must be freaking about all this asymytry :-)
>
>>
>>>Rxf7+ Nxf7+ Kg8 Bxh6}
>>> 10. Qh5 h6
>>> {9:-2.44 h6 Bxf7+ Kh8 Qg6 Qa5+ Bd2 Bf2+ Ke2 Ng3+ hxg3
>>>Qxd2+ Kxd2
>>> Be3+ Kxe3 hxg5}
>>> 11. Bxf7+ Kh8
>>> {12:-Mat05 Kh8 Qg6 Qa5+ c3 Bf2+ Kf1 Ng3+ hxg3 hxg5 Qh5#}
>>> 12. Qg6 Qa5+
>>> {11:-Mat04 Qa5+ c3 Bf2+ Kf1 Ng3+ hxg3 hxg5 Qh5#}
>>> 13. c3 Qxc3+
>>> {14:-Mat03 Qxc3+ bxc3 Bf2+ Kd1 hxg5 Qh5#}
>>> 14. bxc3 Bf2+
>>> {14:-Mat02 Bf2+ Kd1 hxg5 Qh5#}
>>> 15. Kd1 Ne7
>>> {14:-Mat01 Ne7 Qh7#}
>>> {14:-Mat01 hxg5 Qh5#}
>>> 16. Qh7#
>>>
>>>
>>>I didn't study it carefully, but black's move 8 seems horrible, castling
>>>right into the attack. there is an 8 pawn difference between O-O and
>>>f5,
>>>although I can't say what would happen at a deeper depth.
>>>
>>>Where were you wanting Rebel to play a different move? IE everything
>>>looks
>>>bearable up until it castles into white's pieces...
>>
>>f5 is better than o-o but castling is bearable. The ?? move is Qxc7
>>
>>Robert Sherman
>
>"bearable" might be too lax. Black looks to be in serious difficulty
>as opposed to his position after deferring and castling long or sticking
>in the center and attacking on the kingside...
What I mean by "bearable" is that I believe that black can get away with
that move and still win. If you can prove that white has a win or draw
after that move I will be very surprised. In my opinion, with astute
defensive play black survives and wins.
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.