On Apr 12, 10:48*pm, David Richerby
wrote:
Tom F. wrote:
I know programing a chess engine is very difficult. I took a stab at
it in college and gave up.

I've done it and I don't think it's *so* difficult to write a chess
engine. *It's reasonable to be proud of having written one and there's
no shame in having tried and not got there -- it's not easy.

It is certainly not easy to write a chess engine that is competitive
with the top rank engines these days. It was easier in the past when
top engines were not very good. But it is possible given enough
patience. I have an old toy engine I play with sometimes derived from
a VAX Pascal code with various modern improvements.

BUT, I know for sure I wouldn't post results like that.

Amen. *I've not looked at any of Sanny's engine's games for a long
time but, when I did, they were either of a very low standard or had
taken an unacceptably long time to play.

Or both.

As it happens, my engine was about as strong as me in the middlegame
but played a very weak endgame. *I wouldn't have been ashamed to post
its games here but there wouldn't have been much point because the
individual moves give little information about what's going on. *This

Sometimes it does though. I don't think Sanny is strong enough to
analyse his own games or understand the analysis of a stronger engine
so his best bet is actually to post games here and ask for humans to
comment.

whole approach of `here's a game -- tell me what I need to improve' is
fundamentally flawed.

Having a local tame strong chess player to ask for advice and guidance
on the evaluation function would be a lot better. Almost all the
serious teams do this and/or are themselves already strong
chessplayers.

Your engine is not even beta. It's very rudimentary alpha. I guess it
doesn't make any illegal moves. That's a great accomplishment,
actually.

Wellllll, a chess engine works by generating a list of legal moves and
then choosing one of them. *Generating the list is by far the easiest
part of that and I'd hope that any competent programmer would be able
to enumerate the legal moves in any position.

You'd be surprised how often the en passant rule seems to cause grief
in the move generator (even the Deep Blue chip designer admits to
having had trouble with phantom pieces appearing after en passant move
unmove special cases go wrong).

If I were you, I'd keep quiet, at least, until your engine can hold
its own against Crafty or some similar engine.

That's a big ask. *There aren't many who've written an engine that
strong.

Curiously these weaker engines with reasonable tactical nouse play a
lot better against humans than they do against machines. Humans can
miss something and let the machine take advantage. Sannys program
doesn't have sufficient speed or lookahead to stand any chance against
Crafty (still less Rybka) at present.

Unless something is done by way of endgame heuristics (or tablebases)
they are completely blind in the endgame though. I have previously
suggested to Sanny a few small tweaks that would allow his engine to
at leats win the more obvious KRk situations cleanly.

Regards,
Martin Brown