MatPlus.Net

 Website founded by
Milan Velimirović
in 2006

23:29 UTC
ISC 2024
 
  Forum*
 
 
 
 

Username:

Password:

Remember me

 
Forgot your
password?
Click here!
SIGN IN
to create your account if you don't already have one.
CHESS
SOLVING

Tournaments
Rating lists
1-Oct-2024

B P C F





 
 
MatPlus.Net Forum Internet and Computing Lomonosov tablebases bug?
 
You can only view this page!
(1) Posted by Peter Wong [Monday, Sep 17, 2018 13:16]

Lomonosov tablebases bug?


On the official Lomonosov TB site (tb7.chessok.com), there's a forum (accessible to registered users only) in which someone points out an apparent error in the TB assessment of a position. That thread was created back in 2016, had no responses, and the bug still seems to exist, at least in the Android version of Lomonosov.

(= 3+4 )
White to play

The TB gives the DTM as Black mates in 19, but in fact 1.Bd7! draws. The draw is confirmed when this move is played, and indeed if the WB starts on f5 or e6, Lomonosov does find 1.Bd7 as a drawing move. It's especially worrying considering that the bug also affects other positions that could lead to the diagram, e.g. retract the BP from a4 to a5, and 1...a4 is incorrectly accessed as DTM-18. Does this mean there could be other buggy positions as well?

An unrelated tablebase question:
The Wiki article on tablebases provides a table of DTM length records for various materials (https://en.wikipedia.org/wiki/Endgame_tablebase#Tables). What about the DTM length records for dual-free play, likewise for each type of material? Seems to be worthy of investigation but I haven't heard of such searches for the longest "sound" miniatures.
 
(Read Only)pid=16840
(2) Posted by Alain Villeneuve [Monday, Sep 17, 2018 21:02]

Very interesting, thank you Peter.

I noticed some oddities on the 7-tablebases, but only about pawn moves of 2 steps, like h2-h4, which they obviously does not like (the 6-men tablebases either). But this one is new for me. I just notice that 1 Bd7 is not included in the candidates moves. Maybe the problem is here : why are not considered all legal moves ? Maybe the real bug is here ?

It is very funny that the old "Fritz 5,32" (25 years ago ?) find obvious the draw after 1 Bd7. Like we say in french : "on n'arrête pas le progrès" !
 
 
(Read Only)pid=16843
(3) Posted by Peter Wong [Tuesday, Sep 18, 2018 12:25]

Thanks, Alain! Perhaps the issue does relate to the P 2-steps, since in what could be main variation here, White does need to play a tempo move with the h2-P against a similar move by the h7-P in order to draw. 1.Bd7! a3 2.Be6 Kd6 3.Kf7 h6 4.h3! h5 5.h4 and Black is in zugzwang.

This P move issue seems more likely to be the culprit than the display of the candidate moves. In most positions the number of candidates is usually much lower than the number of legal moves, though I wouldn't know how they decide which get shown.

Feel free to share other "oddities" you may have noticed on the 7-unit TB.
 
   
(Read Only)pid=16845
(4) Posted by Alain Villeneuve [Tuesday, Sep 18, 2018 15:02]

Consider this study given in the WCSC of Budapest 1988 : Kc6, Pa4 & h3 / Ke4, Pa7, g7 & h5 draw (L. Anyos 1968). The main line is 1 a5! h4 2 Kb7! but suppose you want to analyse, after the key 1 a5, the line 1...g5.

You may analyse yourself (fortunately it is not complicated), but you can't use the 7-tablebase ! The machine answers to you "invalid position" !
 
   
(Read Only)pid=16846
(5) Posted by Eric Huber [Wednesday, Sep 19, 2018 20:00]

As mentioned on this MatPlus thread http://matplus.net/start.php?px=1537379835&app=forum&act=posts&fid=it&tid=2182&pid=16764#n16764
please use this site https://syzygy-tables.info/ powered by Ronald de Man's Syzygy endgame tablebases.
The abovementioned issues do not exist with Syzygy endgame tablebases.
 
   
(Read Only)pid=16849
(6) Posted by Peter Wong [Thursday, Sep 20, 2018 10:58]; edited by Peter Wong [18-09-20]

@Alain
For some reason I can't replicate that error. After 1.a5 g5, the TB gives 2.a6 and 2.Kb7 as drawing. Maybe the issue only affects the desktop version of Lomonosov?

@Eric
The Syzygy TB seems to give only DTZ results for 6- and 7-unit positions, not DTM. Since the 50 moves rule doesn't apply to problems, I'm puzzled as to how problemists find the Syzygy TB useful? Is it because in studies there's no move limit, that means there's no practical difference between knowing the DTM and DTZ?
 
   
(Read Only)pid=16850
(7) Posted by Eric Huber [Thursday, Sep 20, 2018 20:26]

@Peter
I have absolutely no idea why problem composers should find the tablebases useful, but they are certainly welcome for chess players, solvers and studies composers.
 
 
(Read Only)pid=16851

No more posts


MatPlus.Net Forum Internet and Computing Lomonosov tablebases bug?