Engine recommendation

[Werewolf]

Hi,

I am interested in doing some testing to evaluate the style of play of well known chess engines. To do this I would like them to play a "dummy" - an engine around 1800 elo. Please could someone recommend:

- a 1800 elo (ish) UCI engine
- must be 100% stable, capable of playing 1000 games without crashing
- able to run on a 64 bit windows OS

Many thanks.

[Henk]

An engine of 1800 elo only needs alpha beta for search and material count plus random value for evaluation.

[Werewolf]

Thanks, but do you know of such an engine?

I'd be prepared to go up to 2000 elo.

[Robert Pope]

Would a Winboard engine be acceptable, or do you insist on native UCI? microMax and TSCP are both very solid and 1700-1800ish.

[Henk]

Micromax and fairy max have a time exceeded bug. Sometimes in a total won position they don't move. But maybe that only happens when time control is about five or two minutes. I always test Skipper against fairy max and once in a while I encounter that bug.

[Ferdy]

Thanks, but do you know of such an engine?

I'd be prepared to go up to 2000 elo.

Have a look here.

http://talkchess.com/forum/viewtopic.ph ... ght=stable

[hgm]

Micromax and fairy max have a time exceeded bug. Sometimes in a total won position they don't move. But maybe that only happens when time control is about five or two minutes. I always test Skipper against fairy max and once in a while I encounter that bug.

Note that in micro-Max this was not a bug but a calculated risk. In most of the cases where this happens it was in a totally lost position (it tends to happen when it loses its last piece, and switches off null move as a result). So it hardly lost any Elo because of it, and was not worth the number of characters needed to fix it.

Note that the latest Fairy-Max commit in the hgm.nubati.net source repository fixes this problem for Fairy-Max. The latter doesn't have the maximum Elo/char goal, so it seemed a bit silly to make it inherit this problem. It now invalidates the hash table at the point where it switches off null move, so that it isn't tempted to improve on on an iteration obtained with null move in a search without null move.

[Henk]

So you can use latest version of fairy-max.

[Henk]

Micromax and fairy max have a time exceeded bug. Sometimes in a total won position they don't move. But maybe that only happens when time control is about five or two minutes. I always test Skipper against fairy max and once in a while I encounter that bug.

Note that in micro-Max this was not a bug but a calculated risk. In most of the cases where this happens it was in a totally lost position (it tends to happen when it loses its last piece, and switches off null move as a result). So it hardly lost any Elo because of it, and was not worth the number of characters needed to fix it.

Note that the latest Fairy-Max commit in the hgm.nubati.net source repository fixes this problem for Fairy-Max. The latter doesn't have the maximum Elo/char goal, so it seemed a bit silly to make it inherit this problem. It now invalidates the hash table at the point where it switches off null move, so that it isn't tempted to improve on on an iteration obtained with null move in a search without null move.

Null move is a bug in all engines that use it. Even in Skipper. Skipper does not have Zugzwang detection at all.

[hgm]

Note that the OP explicitly asked for UCI engines. WHich are rather rare in this Elo range.

I thought Roce was pretty stable.