Question about Chest UCI

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

Spliffjiffer
Posts: 416
Joined: Thu Aug 02, 2012 7:48 pm
Location: Germany

Question about Chest UCI

Post by Spliffjiffer »

Outgoing from this position:
[D]8/8/8/8/5p1p/K2B1Bpp/1p1P2rq/2R3bk w - - 0 1

I changed engine parameters like this:

Type: Mate
SearchDepth: 20
SearchMode: Special
ChecksOnly: 16
KingSquares: 1
PieceLimit: 2
MaxMoves: 10

1. Now i let run ChestUCI searching for a mate and it takes hours...i abbort

2. Now i play Rd1 (just for example) and turn on DefenderMoves in the options, let it run and it takes hours....i abbort

3. Now i play b1Q, turn off DefenderMoves, let it run and it instanty shows that there is no mate in 20 moves

So why the hell ChestUCI on step 2. took hours when it sees that after b1Q there will be no mate ??? (on step 2 i tried instead of Rd1 all other possible moves with an appropriate response with the same result ! )

Im going crazy :-(...is it a bug or can i fix this somehow or am i doing something stupid?....please help :-)
Wahrheiten sind Illusionen von denen wir aber vergessen haben dass sie welche sind.
User avatar
F.Huber
Posts: 853
Joined: Thu Mar 09, 2006 4:50 pm
Location: Austria

Re: Question about Chest UCI

Post by F.Huber »

Spliffjiffer wrote: So why the hell ChestUCI on step 2. took hours when it sees that after b1Q there will be no mate ???
Because there are other promotions for black (b1R, b1B, b1N), for which it is not so obviously clear that there is no following #20.
With DefenderMoves ChestUCI checks all moves for a possible mate in the given depth.

So nothing is wrong with the behaviour you see, neither a bug on ChestUCI's nor any error on your side. ;)

BTW, your parameters are very restrictive, and with such a special setting it's quite hard to understand what ChestUCI is in fact doing internally.

Regards,
Franz
Spliffjiffer
Posts: 416
Joined: Thu Aug 02, 2012 7:48 pm
Location: Germany

Re: Question about Chest UCI

Post by Spliffjiffer »

Ah...now i understand
Thank you very much for your reply and your explanation Franz :-)
ChestUCI is GREAT but wouldnt it make sense to "tell" Chest that it has to consider only one promotion when this promotion is sufficent for a "nonmate"?

Grüsse aus Deutschland :-)
Wahrheiten sind Illusionen von denen wir aber vergessen haben dass sie welche sind.
User avatar
F.Huber
Posts: 853
Joined: Thu Mar 09, 2006 4:50 pm
Location: Austria

Re: Question about Chest UCI

Post by F.Huber »

Hallo 'Nachbar'! ;)
Spliffjiffer wrote: ChestUCI is GREAT but wouldnt it make sense to "tell" Chest that it has to consider only one promotion when this promotion is sufficent for a "nonmate"?
Well, ChestUCI is a matesolver, its purpose is to find a mate but not to defend against a mate.
The name 'DefenderMoves' may not be very descriptive, but it does just the following:
it makes all moves for the current side, and then checks if there is any mate (under the given conditions) for the other side.

Franz
Paloma
Posts: 1167
Joined: Thu Dec 25, 2008 9:07 pm
Full name: Herbert L

Re: Question about Chest UCI

Post by Paloma »

Is the position correct?
Can you give the source ... and solution? :)
Candidate moves seems to be: Rc3, c4, c6, c7, c8 - further Rc5, c2 - and Bxg2+ or Rxg1+, but none of them leads to mate in 20 :?
Spliffjiffer
Posts: 416
Joined: Thu Aug 02, 2012 7:48 pm
Location: Germany

Re: Question about Chest UCI (original Testposition)

Post by Spliffjiffer »

Hello Herbert :-)
Im sorry for the irritation....the initial position i posted is a dead draw ...there is no mate ;-)
The position arose after some personaly investigation in a study by Jan Timman which im still not able to solve and it might be that there is no solution...
[D]6r1/p1p1p1pP/P1PpP1P1/8/2P1B3/2K3pp/3P2rq/R5bk w - - 0 1
Timman proclaims a white win here...
the solution should start with 1.hxg8N d5 2.bf3 d4+{...dxc4 leads to mate in 16 according to Chest (with the given parameters but MaxMoves=5 you will quickly see)} if there is one i suppose!?
now, where to move the king?...after that white should underpromote g,e,c pawns to knights again i suppose but im not sure about the a-pawn....there are some tricks with underpromoting it to a bishop or a knight again but i cant make it work with the right defensive moves until now.....
btw: if you remove the c4-pawn to a4 in the initial position the solution is quite easy by underpromoting the a-pawn into a bishop and mating on g2 in the end but with the pawn on c4 this problem is driving me mad :-)
Chapeau to a solver !
Wahrheiten sind Illusionen von denen wir aber vergessen haben dass sie welche sind.