a Chest(UCI)-bug !?

[Spliffjiffer]

i know that i formerly asked a very similar question already (maybe 5-10 years ago) concerning Chest...sry for those who might be pissed off by seeing this again but im probably just too stupid and penetrating to avoid this (i hope for an adequate answer that im able to understand or accept and i will be able to remember about):

this time i put Chest's abilities in question (hopefully unreasonably, so plz help !) by looking at a recent position that a member presented (his name is Hobacle)...the position is this:
[d] 8/4p2p/7p/4pp1k/8/5pp1/P2Kprrb/4nbqn w - - 0 1
ONLY move 1.Kxe1...and its even winning btw

i talk about Chest in absolutely untouched mode ( this means: SearchMode: normal ), just to be sure that there is no user-interaction involved that might be guilty:

now...my 1st step is this: i ask Chest for a "defender move" and i set "SearchDepth" = 10...the result is this:

Code: Select all

FEN: 8/4p2p/7p/4pp1k/8/5pp1/P3prrb/4Kbqn b - - 0 1

ChestUCI:
FEN: 8/4p2p/7p/4pp1k/8/5pp1/P2Kprrb/4nbqn w - - (2+16)
Stellungs-Analyse: W:2/6
Suche nach Matt in -10 ... (Hash=1024MB)
Suche abgeschlossen ... (Zeit=0.10s)
Kein Matt in -10 gefunden ! (00:00)
 2/10	00:00	 114k	1.190k	-M2	1.Kd2-e3 Se1-d3 2.a2-a3 e2-e1D+
 2/10	00:00	 114k	1.190k	-M2	1.Kd2-e3 Se1-c2+ 2.Ke3-d3 e2-e1D+
 3/10	00:00	 114k	1.190k	-M3	1.Kd2-c1 Se1-d3+ 2.Kc1-c2 e2-e1D+ 3.Kc2-b3 De1-b4+
 4/10	00:00	 114k	1.190k	-M4	1.a2-a4 Se1-d3 2.Kd2-c3 e2-e1D+ 3.Kc3-c4 Tf2-c2+ 4.Kc4-d5 Sd3-f4+
 4/10	00:00	 114k	1.190k	-M4	1.a2-a4 Se1-c2 2.Kd2xc2 e2-e1D+ 3.Kc2-b3 De1-b1+ 4.Kb3-c3 Db1-b2+
 4/10	00:00	 114k	1.190k	-M4	1.a2-a3 Se1-d3 2.Kd2-c3 e2-e1D+ 3.Kc3-c4 Tf2-c2+ 4.Kc4-d5 Sd3-f4+
 4/10	00:00	 114k	1.190k	-M4	1.a2-a3 Se1-c2 2.Kd2xc2 e2-e1D+ 3.Kc2-b3 De1-b1+ 4.Kb3-c3 Db1-b2+
 5/10	00:00	 114k	1.190k	-M5	1.Kd2-c3 Se1-c2 2.Kc3-c4 e2-e1D+ 3.Kc4-c5 De1-b4+ 4.Kc5-c6 Tf2-d2 5.a2-a3 Db4-b6+
 10/10	00:00	 114k	1.190k	+M0	1.Kd2xe1

SugaR AI 2.40 avx2:

so it shows that there is only one move that is able to avoid the direct mate !...ok, np...maybe it just does not consider "anti-winning-lines" as a refutation when it searches for a "NOT-"DefenderMove""


2nd step is this (thats an important one!):
now i ask it for a "normal" mate with "SearchDepth" = 30....and it needs ages to consider moves like 1.a3 and 1 a4 (which are both -mate4 as we know already...ok, np so far...)

3rd step (the most important one!):
i play 1.a4 (or a3) Nc2 by my own and make a normal search with "SearchDepth"=126 and it says instantly: "no mate found"

...so what is this ???...any explanation or workaround ?....plz help

[peter]

3rd step (the most important one!):
i play 1.a4 (or a3) Nc2 by my own and make a normal search with "SearchDepth"=126 and it says instantly: "no mate found"

...so what is this ???...any explanation or workaround ?....plz help

I'm not sure, if I got you right, Peer, but if yes, you asked for mate for White this time (3rd step), didn't you?
And after 1.a4 (nor a3) there isn't any mate for White, is there?
So I'd say the answer is correct anyhow regards

[Spliffjiffer]

3rd step (the most important one!):
i play 1.a4 (or a3) Nc2 by my own and make a normal search with "SearchDepth"=126 and it says instantly: "no mate found"

...so what is this ???...any explanation or workaround ?....plz help

I'm not sure, if I got you right, Peer, but if yes, you asked for mate for White this time (3rd step), did you?
And after 1.a4 (nor a3) there isn't any mate for White, is there?
So I'd say the answer is correct anyhow regards

i hope that i get it right,Peter (and im getting more and more confused)....maybe it was "stummering" (thx Walter) articulated but the difference is this: anything else than 1.Kxe1 looses for white within 5 moves...and Chest needs a "million" years to avoid moves like 1.a4 or 1.a3 but they are definately loosing in 4 moves (whereas it gets the mate within 5 min with good parameters after 1.Kxe1!!)....my step 2 and step 3 comments were about "white looks for sth" and step 1 was about "black looks for sth" ...i hope im able to defend myself ...does anything work wrong in my head atm ?

[Uri Blass]

i know that i formerly asked a very similar question already (maybe 5-10 years ago) concerning Chest...sry for those who might be pissed off by seeing this again but im probably just too stupid and penetrating to avoid this (i hope for an adequate answer that im able to understand or accept and i will be able to remember about):

this time i put Chest's abilities in question (hopefully unreasonably, so plz help !) by looking at a recent position that a member presented (his name is Hobacle)...the position is this:
[d] 8/4p2p/7p/4pp1k/8/5pp1/P2Kprrb/4nbqn w - - 0 1
ONLY move 1.Kxe1...and its even winning btw

i talk about Chest in absolutely untouched mode ( this means: SearchMode: normal ), just to be sure that there is no user-interaction involved that might be guilty:

now...my 1st step is this: i ask Chest for a "defender move" and i set "SearchDepth" = 10...the result is this:

Code: Select all

FEN: 8/4p2p/7p/4pp1k/8/5pp1/P3prrb/4Kbqn b - - 0 1

ChestUCI:
FEN: 8/4p2p/7p/4pp1k/8/5pp1/P2Kprrb/4nbqn w - - (2+16)
Stellungs-Analyse: W:2/6
Suche nach Matt in -10 ... (Hash=1024MB)
Suche abgeschlossen ... (Zeit=0.10s)
Kein Matt in -10 gefunden ! (00:00)
 2/10	00:00	 114k	1.190k	-M2	1.Kd2-e3 Se1-d3 2.a2-a3 e2-e1D+
 2/10	00:00	 114k	1.190k	-M2	1.Kd2-e3 Se1-c2+ 2.Ke3-d3 e2-e1D+
 3/10	00:00	 114k	1.190k	-M3	1.Kd2-c1 Se1-d3+ 2.Kc1-c2 e2-e1D+ 3.Kc2-b3 De1-b4+
 4/10	00:00	 114k	1.190k	-M4	1.a2-a4 Se1-d3 2.Kd2-c3 e2-e1D+ 3.Kc3-c4 Tf2-c2+ 4.Kc4-d5 Sd3-f4+
 4/10	00:00	 114k	1.190k	-M4	1.a2-a4 Se1-c2 2.Kd2xc2 e2-e1D+ 3.Kc2-b3 De1-b1+ 4.Kb3-c3 Db1-b2+
 4/10	00:00	 114k	1.190k	-M4	1.a2-a3 Se1-d3 2.Kd2-c3 e2-e1D+ 3.Kc3-c4 Tf2-c2+ 4.Kc4-d5 Sd3-f4+
 4/10	00:00	 114k	1.190k	-M4	1.a2-a3 Se1-c2 2.Kd2xc2 e2-e1D+ 3.Kc2-b3 De1-b1+ 4.Kb3-c3 Db1-b2+
 5/10	00:00	 114k	1.190k	-M5	1.Kd2-c3 Se1-c2 2.Kc3-c4 e2-e1D+ 3.Kc4-c5 De1-b4+ 4.Kc5-c6 Tf2-d2 5.a2-a3 Db4-b6+
 10/10	00:00	 114k	1.190k	+M0	1.Kd2xe1

SugaR AI 2.40 avx2:

so it shows that there is only one move that is able to avoid the direct mate !...ok, np...maybe it just does not consider "anti-winning-lines" as a refutation when it searches for a "NOT-"DefenderMove""


2nd step is this (thats an important one!):
now i ask it for a "normal" mate with "SearchDepth" = 30....and it needs ages to consider moves like 1.a3 and 1 a4 (which are both -mate4 as we know already...ok, np so far...)

3rd step (the most important one!):
i play 1.a4 (or a3) Nc2 by my own and make a normal search with "SearchDepth"=126 and it says instantly: "no mate found"

...so what is this ???...any explanation or workaround ?....plz help

I guess that chest simply starts the search with a line like 1.a4 e4 and not with 1.a4 Nc2 so chest needs ages to prove that white has no mate after 1.a4 e4
Maybe the mistake is later in the search and it starts with 1.a4 Nc2 but does not continue correctly for black.

The not correct line for black does not allow white to mate but also does not allow black to mate so it need to use a lot of time to prove that white cannot mate.

[peter]

anything else than 1.Kxe1 looses for white within 5 moves...and Chest needs a "million" years to avoid moves like 1.a4 or 1.a3 but they are definately loosing in 4 moves (whereas it gets the mate within 5 min with good parameters after 1.Kxe1!!)....my step 2 and step 3 comments were about "white looks for sth" and step 1 was about "black looks for sth"

Guess I got you now, Uri has posted an explanation, that sounds good to me regards

[Spliffjiffer]

I guess that chest simply starts the search with a line like 1.a4 e4 and not with 1.a4 Nc2 so chest needs ages to prove that white has no mate after 1.a4 e4
Maybe the mistake is later in the search and it starts with 1.a4 Nc2 but does not continue correctly for black.

The not correct line for black does not allow white to mate but also does not allow black to mate so it need to use a lot of time to prove that white cannot mate.

this is not satisfining to me....it makes no sense to me when i nead 3 min to solve this in ply 40 and 1 million years in ply 40 (if every move beyond that is simply mate in 5)..its just stupid to not consider this mate in 5...while every move instead of the best move is just loosing in 5... the engine sees it but is not able to carry it over to the next ply !?...i think thats the problem and thats the bug !??...im not programming but its at least easy to me to understand sth (maybe not to fix)....i think this is buggy and "easily" to fix (no matter how hard it is coding-wise), just a brain-thing imo, and i never made a program but i made mathematics, yeah maybe wrong but its a bug for me !?...Chst is an amazing engine and the Franz Hubert is the same !!...Chest works so to say flawlessly, and i rely on it very very much, thx !!