![[in]](/mwf/v12/flags/in.png "India")
Just Want to know which Engines play the best Rook endings ?
Has any one done any tests on endgames ?
One thing Am sure is Stockfish is worse in this case and Houdini tends to keep rook in front of passed pawns.
Has any one done any tests on endgames ?
One thing Am sure is Stockfish is worse in this case and Houdini tends to keep rook in front of passed pawns.
![[gb]](/mwf/v12/flags/gb.png "United Kingdom")
I have not found one engine that understands this as well as I do - and if we were to compare to one of the best endings players that I have had the delight of studying then the chasm between our silicon friends and Smyslov would be even wider!
My advice; get a decent book or books {perhaps study some of Smyslov's games where the master showed great use of Rook play} - you will learn far more than any engine will be able to show you.
All the best.
My advice; get a decent book or books {perhaps study some of Smyslov's games where the master showed great use of Rook play} - you will learn far more than any engine will be able to show you.
All the best.
I am interested in the Chess engines just curious to know which one is best .
Then you will need to be more precise in what you are asking for as chess engines work on a 'best fit' approach; they do not provide an exact answer for every position encountered unless perhaps, you already have a position where an endgame table base is being hit.
I am asking in general. Like HIARCS performs very well in endgames than in middle and openings.
perhaps this could be a spur for someone to create a set of endgame testsets?
![[se]](/mwf/v12/flags/se.png "Sweden")
Shredder and Fritz
Rooks behind passed pawns is more a rule of thumb for human chessplayers. Implied is that they also know, if moving a rook or queen from in front to behind a passed pawn is too much manoeuvering trouble. Rainbow Serpent has some code not in Stockfish for rooks behind the passed pawn but I am not sure it can be tuned very well. It is possible Houdini also has similar code, but that it is not triggered in your case, because the engine also has to make this assessment if it can move a rook in front to behind a passed pawn and that is much harder to assess correctly if it is in front for instance, than if the rook is already on the right side. Protection in front is still better than no protection at all, and there are endgames where the protection by the rook has to be from the side. I think rook endings are more strategic endgames than purely tactical and mostly still out of reach for tablebases, the rules for this are guidelines only, where humans have to find a way with superior pattern recognition. The rules in Smyslov's books (Jonathan Speelman comes to mind too, or John Nunn's books when it is more about the tablebase endgames) are probably not all that useful for programming, well I don't mind that
About Smyslov, there was a story in 'Computerschaak', the two monthly magazine from our CSVN many years ago where Smyslov was taking on a Fidelity machine, I think it was a Fidelity Elite so that was in the time of 8 bit processors still, before the Motorola 68000, and without any hashtables. There was a rookendgame on the board and Smylov had some trouble beating the machine. "I wrote a book about zis, but I forgot what I wrote!" It was not a very serious game and I think Smyslow still won in the end purely on routine.
Eelco
About Smyslov, there was a story in 'Computerschaak', the two monthly magazine from our CSVN many years ago where Smyslov was taking on a Fidelity machine, I think it was a Fidelity Elite so that was in the time of 8 bit processors still, before the Motorola 68000, and without any hashtables. There was a rookendgame on the board and Smylov had some trouble beating the machine. "I wrote a book about zis, but I forgot what I wrote!" It was not a very serious game and I think Smyslow still won in the end purely on routine.
Eelco
I have Symslov and Levenfish's book on rook endings :) It is maasterpiece and I still use it even when analysing with engines and TBs. I agree with all your points and those of Razor too. This is still an area where a human who is competant at rook endings can add significant value even when using engines.
But, as an attempt to address Omprakash's orriginal queston, I offer the following thoughts. IMO there is no simple answer to the question of which is the best engine for R endings. It depends how many pawns are on the board and whether you are looking at it from the stronger or weaker side.
With more pawns on I think Stockfish is strong, as is Komodo. Both are good at finding winning attempts ("plans"). IMO they are better than Houdini and Rybka in this repsect. Boith suffer with few pawns because of the lack of tablebase support.
Houdini is the best at finding drawing lines, though H 1.5 occasionally suffers from the lack of 6 man TBs.
Rybka gets better as the pawns come off IMO and as the TBs kick in.
Zappa is a good all rounder at Rook endings, if you are prepared to give it time. I regard it as slow but sure, just as in other parts of the game.
I only have old copies of Shredder (version 4 and 9). It was good in its day at rook endings but these versions have been surpassed. Maybe the later versions are stronger.
I don't have Fritz so can't coomment on that and don't bother with Critter or Haircs 13 for endings - I have never found that they add anything to the other engines.
As an aside I have been thinking about an efficient way of generating 7 man TBs, as I am especially interested in R+2 vs R+1. In principle it is an easy problem, in practice it becomes hard because of the shear volume of data. I can now see why Nalimov's code is as complex as it is!
But, as an attempt to address Omprakash's orriginal queston, I offer the following thoughts. IMO there is no simple answer to the question of which is the best engine for R endings. It depends how many pawns are on the board and whether you are looking at it from the stronger or weaker side.
With more pawns on I think Stockfish is strong, as is Komodo. Both are good at finding winning attempts ("plans"). IMO they are better than Houdini and Rybka in this repsect. Boith suffer with few pawns because of the lack of tablebase support.
Houdini is the best at finding drawing lines, though H 1.5 occasionally suffers from the lack of 6 man TBs.
Rybka gets better as the pawns come off IMO and as the TBs kick in.
Zappa is a good all rounder at Rook endings, if you are prepared to give it time. I regard it as slow but sure, just as in other parts of the game.
I only have old copies of Shredder (version 4 and 9). It was good in its day at rook endings but these versions have been surpassed. Maybe the later versions are stronger.
I don't have Fritz so can't coomment on that and don't bother with Critter or Haircs 13 for endings - I have never found that they add anything to the other engines.
As an aside I have been thinking about an efficient way of generating 7 man TBs, as I am especially interested in R+2 vs R+1. In principle it is an easy problem, in practice it becomes hard because of the shear volume of data. I can now see why Nalimov's code is as complex as it is!
![[us]](/mwf/v12/flags/us.png "United States")
I have been meaning to ask you how you have been doing on FICGS? and any possible chance I can get a comment on this thread?
http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=24418
http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=24418
Firstly thanks again for introducing me to FICGS. I like it! I have also given LSS a go but only 4 games so far.
I am still working my way up the FICGS ladder. My next rating will be over 2200. It may be tough to get to 2300 or 2400 but I shall give it a try. As my goals is to learn about the game I accept that my methods (just loads of interactive analysis, openings from my OTB repetoire and not a systematic analysis of DBs, no Aquarium/IDEA, rarely long seaches) will probably not give the best results, added to which I have a limited amount of talent!
I will probably take a full or partial break to work on the tablebases which will also slow my progress down.
Re your other thread, I really can't comment much. I mainly use Komodo 3/4, Stockfish 2.2.1 and Houdini 1.5, and rarely anything else until the endgame. I like using these three as they are different from each other and complementary. My main add ons are R3 and Zappa in the endgame and ooccasionally the middlegame, and very occasionally Critter (any fairly recent version) and Hiarcs 13.2 in the middlegame.
As an aside, whilst I rarely use R3 it still delights me with its all round quality. It was really special in its day!
I am still working my way up the FICGS ladder. My next rating will be over 2200. It may be tough to get to 2300 or 2400 but I shall give it a try. As my goals is to learn about the game I accept that my methods (just loads of interactive analysis, openings from my OTB repetoire and not a systematic analysis of DBs, no Aquarium/IDEA, rarely long seaches) will probably not give the best results, added to which I have a limited amount of talent!
I will probably take a full or partial break to work on the tablebases which will also slow my progress down.
Re your other thread, I really can't comment much. I mainly use Komodo 3/4, Stockfish 2.2.1 and Houdini 1.5, and rarely anything else until the endgame. I like using these three as they are different from each other and complementary. My main add ons are R3 and Zappa in the endgame and ooccasionally the middlegame, and very occasionally Critter (any fairly recent version) and Hiarcs 13.2 in the middlegame.
As an aside, whilst I rarely use R3 it still delights me with its all round quality. It was really special in its day!
Thanks.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill