Zugzwang detection

[Uri]

I think programs (especially Toga) need something called need zugzwang detection in endgames.

For example in this position Toga CMLX 1.4 beta5d thinks it's draw but the next move Kc6 wins for white because black is in zugzwang.

[d]8/2K1k3/7p/4p1p1/4P1P1/5P1P/8/8 w - - 0 57

[ernest]

I think programs (especially Toga) need something called need zugzwang detection in endgames.

Strange: all other Togas I tried (including beta5c), have no problem with this position... Maybe it's linked to Teemu's beta5d...

[Uri]

I think programs (especially Toga) need something called need zugzwang detection in endgames.

Strange: all other Togas I tried (including beta5c), have no problem with this position... Maybe it's linked to Teemu's beta5d...

Without endgame tablebases or with endgame tablebases?

I think programmers don't invest much time in endgames because endgames are for the most part solved for 3-4-5-6-7 men.

[Cubeman]

My Toga versions that run on my PPC know that White is clearly and immediately winning.And my PPC does not have any form of endgame tablebases.

[Uri]

My Toga versions that run on my PPC know that White is clearly and immediately winning.And my PPC does not have any form of endgame tablebases.

Which versions of Toga are you using?

[Cubeman]

I use Togall 1.2.1a and now the latest Togallv3.1.2se running under the PocketGrandMaster GUI.

[Uri Blass]

I think programs (especially Toga) need something called need zugzwang detection in endgames.

Strange: all other Togas I tried (including beta5c), have no problem with this position... Maybe it's linked to Teemu's beta5d...

Without endgame tablebases or with endgame tablebases?

I think programmers don't invest much time in endgames because endgames are for the most part solved for 3-4-5-6-7 men.

You are clearly wrong.
1)There is no need to invest time on endgame to solve this position
Every program that I know except special buggy untested toga can solve it.

2)Tablebases are not a reason not to invest time on endgame knowledge when tablebase access is slow.
Endgames usually are not solved by tablebases and tablebase give very little to playing strength of chess programs.

If programmers do not invest much time on endgame knowledge is because other things like middle game are more important but it has nothing to do with tablebases.

Uri

[Uri]

You are clearly wrong.
1)There is no need to invest time on endgame to solve this position
Every program that I know except special buggy untested toga can solve it.

All other Toga versions i have (Toga 3.1.2SE, Toga 1.4.1SE and Toga II 1.4 beta5d) couldn't see it's a win for white in this position.

[d]8/2K5/4k2p/4p1p1/4P1P1/5P1P/8/8 b - - 0 56

[Uri Blass]

You are clearly wrong.
1)There is no need to invest time on endgame to solve this position
Every program that I know except special buggy untested toga can solve it.

All other Toga versions i have (Toga 3.1.2SE, Toga 1.4.1SE and Toga II 1.4 beta5d) couldn't see it's a win for white in this position.

[d]8/2K5/4k2p/4p1p1/4P1P1/5P1P/8/8 b - - 0 56

Now you give a different position and I guess that the draw evaluation is only with a game because toga1.41SE does not see a draw from your position without a game.

I guess that it has nothing to do with zugzwang detection but with the simple solution of evaluating repetition as a draw(it is possible to do something else but it is not very important for games and programs are usually designed for games and not foe finding blunders in human-human games).

Probably Ke7 caused repetition of previous position and toga evaluates it as a draw.

Programs are designed to play games and toga does not blunder here because it already has a lost position and everything that toga does with black lose.

There may be rare cases when evaluating repetition different than 0.00 can help but these cases can happen mainly when the opponent blunders from winning position and allow toga to choose between repetition and another drawing move and this usually does not happen in comp-comp games.

Uri

[Cubeman]

You are clearly wrong.
1)There is no need to invest time on endgame to solve this position
Every program that I know except special buggy untested toga can solve it.

All other Toga versions i have (Toga 3.1.2SE, Toga 1.4.1SE and Toga II 1.4 beta5d) couldn't see it's a win for white in this position.

[d]8/2K5/4k2p/4p1p1/4P1P1/5P1P/8/8 b - - 0 56

Even with this 2nd position the same versions of Toga I have still see that White is clearly winning first evals are +4.98 then +5.23 at D=22 then +11.04 at D=23 and time 90 seconds.