Martin Brown wrote:

Robt Hyatt once told me that Crafty would never play the Ruy since 3. Bb5
looks like a very poor move.

Ruy Lopez wasn't all that convinced by his own named opening either.

Maybe this opening is inferior, though it fooled many for
centuries. "You can fool some of the people, all of the time."

I don't doubt your word that Dr Hyatt said that. But it seems he was
wrong. Crafty 19.01 with 256MB cache sometimes gets 3. Bb5 as the
preferred line rather than 3. Bc4 in searches to plys around 16 (after
about 12 hours on a 3GHz machine). Currently at ply 17 it has 3. Bc4
0.09 and 3. Bb5 0.01 (at ply 16 they were 0.01 and 0.1 - ie roughly
swapped around).

And if you allow for say 5 centipawns of Gaussian noise dither

A clever improvement! Now Rybka can think without being
bothered by spectator noise, and the grinding of teeth which
begins anew after each crushing move. Does anyone sell
tooth protectors for this?

on the
raw evaluation function it might still sometimes play the nominaly
"inferior" line.

Engines like Shredder with better pawn structure and positional
awareness get 3. Bb5 top for a fair proportion of the time (and at
depth it prefers Bb5 for plys 19 through 22, with 23 reverting to Bc4,
and 24 still running). At ply 20 [ 23 ] the top 5 ranking moves a

3. Bb5 0.42 [ 0.24 ]
3. d4 0.31 [ 0.19 ]
3. Bc4 0.30 [ 0.31 ]
3. Nc3 0.23 [ 0.17 ]
3. Be2 0.23 [ 0.03 ]
(everything else negative)

Wow. I had no idea chess programs were already up so high
in search depth. I can remember back to when 8 or 9 plys
were considered impressive.

But I think it is a very poor idea to try and use what humans
merely *believe* is a superior move as the test for computers.
Better to come up with something purely objective, like say,
checkmate or snatching a free Queen.

The fact(?) that humans have not *yet* found a satisfactory
defense in the Ruy Lopez in no way proves that one does not
exist, nonetheless. A long time ago, after 1.P-K4 the response
....c5 was scoffed at as unorthodox (at best); now the move has
become acceptable -- even preferred.

After my next computer upgrade, I may buy Rybka and have
a look at its best line of play to see *why* it doesn't like 3.B-b5
much. Maybe there is a perfectly good reason. Just guessing,
I would say the programmers taught the program that Bishops
are better than Knights, and if the exchange is therefore scored
as a bad thing, the Bishop is a target on that square, and it is
possible that the scoring of pawn weakness which ensues
after ...a6 and ...b5 is insufficient to fully offset other factors.

In fact, it seems likely that IM Innes' complaint relates more
toward the programmers than it does toward computers
themselves. It is the programmers who have, indirectly, made
it unlikely for their machines to choose IM Innes' pet move.

-- help bot