Author: Robert Hyatt
Date: 21:16:22 04/25/01
Go up one level in this thread
On April 25, 2001 at 19:03:14, Vincent Diepeveen wrote:
>
>According to bob 8(6) means 8 ply in software and 6 in hardware hence 14 ply.
>I believe, and i'm pretty sure here that it's 8 plies it's searching here.
>
Not "according to Bob." "According to Hsu, Campbell, etc..."
>If i turn on SE + recaptures in diep then i get nearly everywhere the same
>mainline as deep blue did in the majority of moves.
>
>I'll produce if i have a bit more time some log files of all games
>to compare.
>
>Bf5 is a tactical shot here and all programs agree here. don't tell me
>that deep blue was a 'special program' in here which needs 4 ply more
>as even programs without recapture extensions!
OK. Uri and I have both been "gentle" here. It isn't working. So to be
more blunt.. Your "tactical shot that all programs agree on" is a complete
load of _crap_.
First, here is output from crafty where I asked it to show the best 7 moves
and their scores:
({11:-0.99} 1. Bf5 exf5 2. Rxe7 Nxe7 3. Re1 a5 4. Nh4 g6 5. Bd6
Rh7 6. Qc3 $17)
({11:-1.19} 1. Re2 Qf6 2. Rae1 Nb4 3. Qc3 e5 4. Be4 exd4 5. Nxd
4 Kb7 6. Bxc6+ Nxc6 7. Nxb5 Qxc3 8. Nxc3 $17)
({11:-1.19} 1. Be4 a5 2. Bf5 Nb4 3. Qc3 exf5 4. Rxe7 Bxe7 5. Re
1 Re8 6. d5 $17)
({11:-1.37} 1. c4 Nb4 2. Qc3 bxc4 3. Qxc4 Nb6 4. Qxe6+ Qxe6 5.
Rxe6 Bxf3 6. gxf3 Kb7 7. Be4+ N6d5 8. Rc1 $17)
({11:-1.38} 1. c3 Qf6 2. c4 Nb4 3. Qc3 bxc4 4. Qxc4 Qxg6 5. Rxe
6 Nb6 6. Qxc6+ Nxc6 7. Rxg6 $17)
({11:-1.44} 1. Qe2 Qf6 2. Qxe6 Qxe6 3. Rxe6 Kb7 4. Bf5 b4 5. Ra
e1 Rd8 6. h4 Bb5 $17)
({11:-1.42} 1. Rac1 Kb7 2. c4 bxc4 3. Qxc4 Rc8 4. Ra1 Nb8 5. Qa
2 a6 6. Bf5 Bd7 7. Bxb8 Kxb8 8. Qxa6 $17)
If you look, Bf5 at 11 plies is -.99... Re2 is -1.19 as is Be4. In fact
the worst score is under 1/2 pawn away from the best. I could _easily_ see a
different evaluation favoring something besides Bf5 at most any depth. The
other moves are very close. Bf5 is not a tactically killing move as you
suggest... it is just a positional move that takes advantage of a pin along the
e-file. It doesn't win the world... lots of moves produce reasonable scores
that still say black stands better.
So get off of this "tactically worst in the world if it takes N plies to
find Bf5." There is nothing to find there. As the above shows. (the above
was done on my notebook giving 60 seconds per search using the annotate
command).
>
>When talking about evaluation it's probably very identical to Zarkov.
How can _anyone_ that aspires to be a scientist make a statement like that?
Science is based on _fact_ and actual _observations_ not on wild speculation
and guess-work and "I think"...
>
>Zarkov not only produces usually the same move like deep blue, it produced
>here at home most of the same moves at the same depth including majority
>of all bad moves played by deep blue. Zarkov's mobility is probably what
>Seirawan describes and what Deep Blue also is doing according to Seirawan:
>just count the number of squares a queen is attacking (when talking
>queen mobility).
This move is a normal move. There is nothing to suggest that finding Bf4 at
the Nth iteration marks DB as stupid. Bf4 is _just_ another move to consider.
Not a tactical winner...
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.