15 Mar 06
Originally posted by RingtailhunterIt's ok to talk about this one in progress right?
This is the game phlabi and I have. It is an unrated game. Shredder 9 is black Fritz 8 is not.
Game 1609406
Fritz thinks its a draw. Shredder thinks it's winning. We'll see.
16 Mar 06
Originally posted by Ringtailhuntermove 35??? What am I missing?
This is the game phlabi and I have. It is an unrated game. Shredder 9 is black Fritz 8 is not.
Game 1609406
Fritz thinks its a draw. Shredder thinks it's winning. We'll see.
16 Mar 06
1 edit
Originally posted by Freddie2006Good memory Freedie...
I think I remember Phlab mentioning that he made a mouseslip so they used a move to go back to how the game was before.
This game is my Fritz 9 vs RTH's Shredder 9 GUI.
Originally posted by Shinidoki
move 35??? What am I missing?
A better question is what was white doing on move 34. Here is the line that I was supposed to play (cut from my move 33 message)
1. +- (1.71): 34.Rf4 Qc7 35.Qf3 Re8 36.Rd4 h6 37.Qe4 Nb3 38.Rd6 Nc5
34. Rf3 was me doing the wrong move, and we decided it would have been a waste of time to not 'set' the pieces back to how the board looked at the end of moves 33.
So move 36 is really what the engines would have been doing for move 34. I messed a move up 34, move 35 was a 'take back' and game continues. I didn't realize there was a thread on this, I'll look around.
16 Mar 06
Originally posted by exigentskyWhich crown?
Shredder holds the crown now
Rybka is currently beating everything in engine matches (this is prior to Shredder 10 release - who knows what that will bring).
In the playchess.com engine room, the participants are dominated by Rybka versions, and although Rybka has not yet reached the latest SSDF rating list for engines, in other testing it is showing incredible results (see the Computer Chess Club site).
By-the-way, as well as Rybka, I do like Shredder for analysis. I really hope that Shredder 10 challenges or overtakes Rybka in strength as that would put the ball back in Rybka's court.
16 Mar 06
Originally posted by RingtailhunterThere are two things that improve strength of engines; search depth and complexity of position analysis.
To be honest I have never used fritz. I have read quite a bit about it though. Shredder will look a little further ahead.
From what I have read Shredder uses quite aggressive pruning (will discard unlikely looking lines so as not to waste time analysing them) and uses a slightly simpler position algorithm (the formula it uses to see if a position is good or bad) to make gains in search depth.
Hiarcs is the opposite, it doesn't prune as much and uses a more complex algorithm which slows it down but means it will determine the goodness of a position more accurately.
Fritz is in between the two.
Now obviously with competing ideas like this there is likely to be an optimum point. Of course the problem is finding it.
Attempting to search deeper at the expense of algorithm complexity has the problem that as you move through the ply (half-moves) the number of positions to check grows at an alarming rate (even with pruning) and the position scores could be a little shakey (see the game known as the Alterman Wall for an example of a position algorithm failing [the game also contains a out of horizon plan]).
Too heavy pruning means that sometimes the engine will completely disgard a line that could be winning but not for say 13 ply (well within search range but anything involving say a queen sac on ply 2 will likely have been pruned) until the secondary search (a more complete search that follows behind the primary search) picks it up. Shredder seems to suffer from this.
Attempting to check every position with no pruning (Deep Blue was the last engine to do this) or use an extremely convoluted algorithm means that the seeing horizon of the engine is reduced and any plan exceeding that horizon will remain unseen perhaps until it's too late.
16 Mar 06
Originally posted by XanthosNZThanks Xanthos, that was both interesting and informative.
There are two things that improve strength of engines; search depth and complexity of position analysis.
From what I have read Shredder uses quite aggressive pruning (will discard unlikely looking lines so as not to waste time analysing them) and uses a slightly simpler position algorithm (the formula it uses to see if a position is good or bad) to mak ...[text shortened]... is reduced and any plan exceeding that horizon will remain unseen perhaps until it's too late.
16 Mar 06
Originally posted by XanthosNZIt's true that Deep Blue was heavily brute force, especially at the hardware level. But the version that beat Kasparov also made good use of selective search extensions (pruning) which helped it avoid the horizon effect. And, of course, it made use of alpha-beta pruning.
Attempting to check every position with no pruning (Deep Blue was the last engine to do this) or use an extremely convoluted algorithm means that the seeing horizon of the engine is reduced and any plan exceeding that horizon will remain unseen perhaps until it's too late.