![[gb]](/mwf/v12/flags/gb.png "United Kingdom")
It seems a bit silly for an engine such as Rybka to have an elo of over 3000 but still can't mate with B N & k
against a lone K, and yet an engine like crafy can. If you for example place the lone K on g7 or h8 as black and include the white squared B with N and K for white at furthest point away then Crafty or Fritz ,and perhaps others, seem to know what to do in driving the K over to the white square. I know that in any important game Rybka will use tablebases to finish off so i don't suppose its important and it obviously wasn't one of Vasik's priorities. However,it still upsets me. Rybka winfinder also has no idea what to do. Time used 1 or 2 minutes.
against a lone K, and yet an engine like crafy can. If you for example place the lone K on g7 or h8 as black and include the white squared B with N and K for white at furthest point away then Crafty or Fritz ,and perhaps others, seem to know what to do in driving the K over to the white square. I know that in any important game Rybka will use tablebases to finish off so i don't suppose its important and it obviously wasn't one of Vasik's priorities. However,it still upsets me. Rybka winfinder also has no idea what to do. Time used 1 or 2 minutes.
TB are part of computer chess. if they were included in the executable file, nobody would complain about this.
By Michael Waesch ![[de]](/mwf/v12/flags/de.png "Germany")
Date 2007-03-06 16:37
I would complain for getting a 1.2 TB big executable if that is possible.
Mike
Mike
![[us]](/mwf/v12/flags/us.png "United States")
more than half the 6men are 100% useless by fact (no where close to 1.2T is needed).
plus kbnk is just 4men, in which this set along with 3men is astronomically smaller than 6men and can be easily put in the exe.
maybe that is a good idea.
Vas what do you think of putting 4men tablebases (not 6men, or 5men , or even 3men) in the exe. for faster access and much better endgame play and analysis (due to bettering the pruning of lines leading into the 4men endgame)? Or just maybe an even smaller bitbase format instead. Bitbases have been found to work effectively.
plus kbnk is just 4men, in which this set along with 3men is astronomically smaller than 6men and can be easily put in the exe.
maybe that is a good idea.
Vas what do you think of putting 4men tablebases (not 6men, or 5men , or even 3men) in the exe. for faster access and much better endgame play and analysis (due to bettering the pruning of lines leading into the 4men endgame)? Or just maybe an even smaller bitbase format instead. Bitbases have been found to work effectively.
![[ph]](/mwf/v12/flags/ph.png "Philippines")
why not simply use the bitbases that Toga II 1.3x4 uses, the ones developed by Scorpio engine developer Daniel Shawul.
all3-4 shredderbases is just 1.4Mb. In any case, size is the main reason why TB are not included in the executable file. Of course, in this way everything looks "cleaner", but the issue I wanted to stress here is that TB are a part of computer chess. 3-4menTB are quite basic in endgames. If you do not use them, you can expect that the engine will play some bad moves, and you cannot complain about that.
![[hr]](/mwf/v12/flags/hr.png "Croatia")
The way I understand it, shredderbases are not so "revolutionary", namely, when consulting them, engine gets only information if given position is lost/drawn/won, and not which moves are best and path to optimal solution. It still needs to consult Nalimov TB to get the best performance ( = finish the game with best possible result).
I guess it all comes down to market; programmers decide if they should bother implementing for example KNBvK into their engines, based on how much trouble is this to them as opposed to what is the probability that an average user has 4-men TBs on his PC. Not so long ago (a year ot 2), it was not very common to have all 5-men, nowadays, if you do not have "some" (or even all) 6-men is weird.
Dragon Mist
I guess it all comes down to market; programmers decide if they should bother implementing for example KNBvK into their engines, based on how much trouble is this to them as opposed to what is the probability that an average user has 4-men TBs on his PC. Not so long ago (a year ot 2), it was not very common to have all 5-men, nowadays, if you do not have "some" (or even all) 6-men is weird.
Dragon Mist
shredderbases indeed are not so revolutionary. It is clear that during the search, you do not need to know what the best move is, but only if the position is w/l/d. You need the full info (best move, distance to mate) only when you actually have to move (ok, the distance to mate is most of the time irrelevant also in this case). shredderbases guarantee therefore a faster access during the search, and since they are much smaller, can be loaded in the RAM to have much much faster access. BTW, TB are not so relevant for strenght, since many games end with much more than 6 men, and even in games with 3-4-5-6 men endings, an engine can do very well with the "slow" Nalimov. So, shredderbases are a nice (but not new) idea, but definitely not revolutionary. I think that rybka does not support this kind of endgame bases, simply because V.R. (wisely) decided he can spend his time in more effective improvements of rybka. If rybka were developed by ten people, you probably would have similar endgame bases.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill