Sorry, I just saw Landon's latest post. Actually, the free Rybka is quite strong in its own right. According to CEGT, the free Rybka is actually a few points stronger than the newest Toga. But although the free Rybka is strong, it has some irritating flaws (it's a beta, after all.) It can't use tablebases, and it's endgame technique is horrendous. Also, it can only display the principle variation (no additional variations.)
So if you go the free route, using Rybka is only slightly better than Toga, but with more bugs and irritations. Or you can give up just a few ratings points and go with the more stable Toga.
I disagree Xanthos. Which engine you use makes a huge difference. Rybka without tablebases will thrash many other lesser engines, even with tablebases. Tablebases make far less of a difference than one might think. In the majority of games, the tablebases are never even accessed. Yes, every now and then having tablebases will salvage a draw from a loss or a win from a draw, but not that often. Better to have tablebases, but it's not a huge difference in rating points.
Originally posted by Mad RookPlay a game against an engine with tablebases with your tablebases disabled and you'll see.
I disagree Xanthos. Which engine you use makes a huge difference. Rybka without tablebases will thrash many other lesser engines, even with tablebases. Tablebases make far less of a difference than one might think. In the majority of games, the tablebases are never even accessed. Yes, every now and then having tablebases will salvage a draw from a loss or a ...[text shortened]... but not that often. Better to have tablebases, but it's not a huge difference in rating points.
Originally posted by Mad RookAnd I stand by mine. I've won more than a handful of games over 'superior' engines during the final stages (ie. when tablebases become useful) from even (and occasionally lost) positions.
Been there, done that. I play engine-engine matches on my PC all the time. I stand by my statement.
And how are you not reaching the point where tablebases enter the search tree in every game?
Originally posted by XanthosNZEvery game? Tablebases only become a factor in the endgame, when there are 5 pieces or less. Thus, if you get a strong engine up a piece in the middle game, with or without tablebases, it will usually win. In fact, the other engine will probably resig before the endgame. (depends on settings)
And I stand by mine. I've won more than a handful of games over 'superior' engines during the final stages (ie. when tablebases become useful) from even (and occasionally lost) positions.
And how are you not reaching the point where tablebases enter the search tree in every game?
Originally posted by exigentskyThe game will only be decided in the middlegame if one engine is significantly stronger than the other (as mistakes aren't an issue). In my experience ( > 5000 games in the Playchess engine room) most games are decided either in the opening (one side plays a line that an engine shouldn't play) or in the transfer from the mid to the endgame (this is a specific point in engine chess). Whichever side can make the transition first gains an excellent advantage.
Every game? Tablebases only become a factor in the endgame, when there are 5 pieces or less. Thus, if you get a strong engine up a piece in the middle game, with or without tablebases, it will usually win. In fact, the other engine will probably resig before the endgame. (depends on settings)
Also I've seen tablebase activity in Fritz with 8 or 9 pieces remaining on the board and I could easily construct examples of occasions with many more pieces. If a position in a tablebase occurs during a search tree then you are gaining the advantage of the tablebases.
...tablebases greatly imporve the search horizon of a Computer Engine.
If it gets 13 ply deep and spots a tablebase postion it might be able to quickly jump another 8 ply ~ where it transposes to a won tablebase position with perfect play.
- Thus it will improve perfermance and rating, becuase the other engine is only looking 13 deep.
Now all it has to do is get to an evenish middlegame and wait until it can "out-think" (i.e. search deeper) than the other engine.
as for other things to improve engines.
better (faster) hardware.
high Quality database (which should complement the Engines opening choices...corr. games are the best to use)
Good opening Book
Better engine (of course!!)
TABLEBASES
etc, etc
Originally posted by LanndonKaneI'm basing my emphasis on tablebases on my own experiences. They added close to 100 points to the rating of my engine in engine-engine matches. There is an awful lot of bad advice out there about engine play.
http://www.aarontay.per.sg/Winboard/egtb.html#[A.1]
I says only 25 elo points on avaerage.
Is this the huge difference you've been ranting about, Xanthos?