I don't play against engines much any more. When I did I found there was something to be said for playing weaker engines at full strength than playing a strong engine trying to emulate weak play. In the latter case I often felt as though I was playing against a team consisting of a somewhat weaker player, a GM who could take over at any time, and a 6 year old who at rare intervals would be allowed to make the move. I'd often feel like I wasted my time after either playing well to gain an advantage then suddenly lose to impossibly brilliant play, or get steadily beaten back only to be handed the game with the mother of all dufus moves.
There's a lot of interest in seeing just how strong engines can become. For analysis these are great, but in any practical sense does it matter to a shmoe like me if my games are picked apart by a program that is 1400 Elo stronger than I am as opposed to 1200 Elo? These engines are much stronger than they need to be to provide a challenging opponent unless your name is Kasparov or Kramnik (and even then!) I think it would be interesting to see someone develop a series of engines designed to be challenging to humans at specific playing strenghts rather than taking one stong engine and dumbing it down to various degrees.
I've been training openings against gnuchess recently. I'll just try to play book moves until I or gnuchess drops out of book, check the position for how the play could go (or play a little further), and then reset and start anew.
the rare occasion I actually play the game to the bitter end, I get totally killed. but somehow it still feels more enjoyable to play against gnuchess than fritz, as it doesn't play so mindnumblingly accurate chess (at G/5) as fritz does.
it's also good training against losing material, on full strength. humans miss (or decide against) taking material, but an engine almost never does. drop the ball, and you're dead.
the chessmaster personalities are nice, but it has the same problem as fritz. on full strength they're too strong to get meaningful games, and weakened they throw in random moves with otherwise GM level play.
Originally posted by krimIt is technically possible to see a tactical line 6-8 moves deep which could be beyond the depth of the machine to see at blitz time speeds .. this can happen frequently in the opening if the machine has a limited opening book (you can defeat chess jester over and over using this technique for instance). I would think it happens relatively rarely at blitz speeds in the middle game and never in the endgame against a decent engine.
Heard this from where?
I doubt it would ever happen at competition level at normal time speeds where the machine can follow viable lines 20 moves deep .. the only hope human beings have in these situations is seeking long term positional advantages.
At this point the line between tactics and what we would think of as strategy is already starting to blur for the top machines .. they just don't make tactical errors within our search depth.
That said, a classic example (maybe "the" classic example) of humans being able to outplay machines tactically is the h2/h7 bishop sac on a castled position. These can often have forced resolutions beyond the machines horizon but even then I imagine this would only work under blitz speeds.