Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Top engines fail to found obvious best move! A good test position!

Author: Martin Grabriel

Date: 21:28:55 11/24/99

Go up one level in this thread


I love these posts.
Using the Rebel10 engine on epd2diag, Rebel 10 started looking at Bf5xh3 after
about 3 mins, confirmed that it is Bf5xh3 in 11 mins 4 secs, and then gave the
line below in about 19 mins 59 secs :

19:59  08.16  0.41  Bf5xh3 g2xh3 Qd7xh3 Ra1-c1 Nf6-g4 Nc3-e4 Ng4xe3


Hardware : PII 350 Mhz, 32 Mb Ram (running simultaneously many things like lotus
notes, Ms Word, etc.)
and the engine was only given 2 Mb ram for use under epd2diag.

I don't have Rebel Century, I think it will do even better.



On November 24, 1999 at 22:42:52, yibing fan wrote:

>This position was rearched after white's 15th move in a game between
>Duz_Hotimirsky F I  - Yudovich Mikhail M, 1948. Black to move. The game only
>lasted 20 moves.
>
>3r1rk1/p2q1ppp/2pb1n2/5b2/1nB5/2N1BN1P/PP2QPP1/R4RK1 b - - id=Duz_Hotimirsky F I
> - Yudovich Mikhail M ; bm=f5h3;
>
>All my top chess softwares failed to find the best move BxP(h3) within 8 hrs
>search on a PII 266.  I tried crafty 16.6, Rebel 8, CM6000. Try yours!  It is a
>good position to keep your chess engine busy for whole Thanksgiving holidays!
>
>Rebel 8 eventually came up with the best move after ~16 hrs seaching!
>It reach this conclusion after evaluated ~150 million positions. Before that, it
>prefers Rfe8 pinning white Queen.
>
>Crafty on the other hand eveluated >500 million positions and still perfered
>Nc2.
>(Sorry Bob, I terminated search after about 10 hrs, It may came up with the same
>conclusion as rebel 8 if I gave it another 10 hrs.)
>CM6000 also prefers Nc2.
>
>It is not suprising or new that computer miss best moves, what it surprised me
>is that they miss such obvious (for human at least) one.  I am a club player, I
>have no problem to see BxP and to win against any above software at this
>position. The reason to play BxP is simple, here is how I think: after bishop
>for two pawns, white K is traped. white's N, f-pawn and Q is terribly tied
>togeter; other white pieces are just too far away to save their King. On the
>contrast, black enjoys direct attack with his Q, N and B; He also has a reserved
>Rook to deliver final KO if neccessary.  You don't need a lot calculations, just
>make sure not make a terrible blunder. Go for it!
>
>Unfortunately, none of those software can think this way. In addition, I think
>this position is more of tactic than positional.
>
>Bob, Ed you may want to add this position into your testing suite.
>
>By the way, I like both your programs, they are excellent for analyse my games.
>
>Yibing Fan



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.