Kornrade Test

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

Moderators: hgm, Rebel, chrisw

peter
Posts: 3185
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: Kornrade Test

Post by peter »

MikeGL wrote: Fri Sep 21, 2018 8:53 pm Would try those specific ones now, (by cheating)
I mean tweaking some engine settings in case I get lucky :mrgreen:
That's fine!
Yet I think, luck's what you'd need.
:)
One more tipp out of my tries in this direction: the engine or it's setting, that's best in Backward of the positions should have best chances with long computing started with empty hash.

I didn't try Sting very long till now, cause it can use 8G hash only, so TCs of more than an hour probably wouldn't help much with this one, even if it for sure is the best one as for cyclic Zugzwang together with lazy King- demand. E.g. it solves Constrictor(1), which I wouldn't have thought to be possible some years ago, when Mihai composed it.

As for tweaking his engine Marek could give Sting special code for each an every study probably, but he's fair enough to say so, if he did, and probably it would make the engine weaker as for other positions.

McBrain with switched off nullmove and maybe ICCF- option chosen too, Matefinder with Study Mode on are both good at Chichitza, Houdini (ev.Tactical) for some other positions too, maybe to change Houdini's 50 move- boundary could help as for the fortress Stubbornness, lowering it much, and or letting it compute rather long, question would be, if the output- line then would be a correct one, even it the eval would get somewhat near to 0.0 together with correct first move.

Here's some output after short computing with 50Moves set to 10 and Tactical1:

5: Mihai Neghina5 - Stubbornness, White to move and draw
8/b4kp1/2p1p3/5pB1/p1PP4/1p6/3K1P2/3N4 w - - 0 1

Analysis by Houdini 6.03 Pro x64-popc:


1.Kc3 Bxd4+ 2.Kxd4 b2 3.Nxb2 a3 4.Nd3 a2 5.Ne5+ Ke8 6.f4 a1Q+ 7.Kc5 Qa6 8.Kd6 Qb7 9.Kxe6 Qc7 10.Bh4 Qa5 11.Kxf5 Qa1 12.Bg5 Qh1 13.Ke6 Qd1 14.Kf5 Qb3 15.Kg6 Kf8 16.Kf5 Qh3+ 17.Ke4 Kg8 18.Kd4 Qa3 19.Bh4 Qd6+ 20.Ke3 g6 21.Bg5 Qa3+ 22.Kd4 Qb4 23.Ke3 Kg7 24.Kd3 Qb6 25.Ke2
-/+ (-1.25) Depth: 35/75 00:06:00 13167MN

Now one could say, -1.25 at least isn't a for sure losing one eval anymore, and of course, going into the line it's getting smaller more and more.
Yet 0.0 at the starting position still is something different and -1.25 isn't a real drawing eval neither.

Give me your news.
:)
Peter.
peter
Posts: 3185
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: Kornrade Test

Post by peter »

peter wrote: Fri Sep 21, 2018 8:58 pm 5: Mihai Neghina5 - Stubbornness, White to move and draw
8/b4kp1/2p1p3/5pB1/p1PP4/1p6/3K1P2/3N4 w - - 0 1

Analysis by Houdini 6.03 Pro x64-popc:


1.Kc3 Bxd4+ 2.Kxd4 b2 3.Nxb2 a3 4.Nd3 a2 5.Ne5+ Ke8 6.f4 a1Q+ 7.Kc5 Qa6 8.Kd6 Qb7 9.Kxe6 Qc7 10.Bh4 Qa5 11.Kxf5 Qa1 12.Bg5 Qh1 13.Ke6 Qd1 14.Kf5 Qb3 15.Kg6 Kf8 16.Kf5 Qh3+ 17.Ke4 Kg8 18.Kd4 Qa3 19.Bh4 Qd6+ 20.Ke3 g6 21.Bg5 Qa3+ 22.Kd4 Qb4 23.Ke3 Kg7 24.Kd3 Qb6 25.Ke2
-/+ (-1.25) Depth: 35/75 00:06:00 13167MN
Ok, 50Moves to 5 and Tactical1 brings eval down to -0.2 soon, probably it will go down furthermore and the first 6 moves, which are the important ones, are correct, so I'd say yes, Stubbornness is to be solved by tricking with this one setting. That would work with a SF and it's 50Moves boundary too, which isn't so difficult to code in for a well trained knower of SF- code. To be sure it would work, some tactical code like "Romeo23" gave to CFish sometimes, would help too probably.

Yet I wouldn't trust H6 with 50Moves set to 5 only as for other positions very far.
:)
Last edited by peter on Fri Sep 21, 2018 9:52 pm, edited 1 time in total.
Peter.
MikeGL
Posts: 1010
Joined: Thu Sep 01, 2011 2:49 pm

Re: Kornrade Test

Post by MikeGL »

Thanks for these additional tips on specific engines.
Would try, and post success stories.
peter wrote: Fri Sep 21, 2018 8:58 pm
MikeGL wrote: Fri Sep 21, 2018 8:53 pm Would try those specific ones now, (by cheating)
I mean tweaking some engine settings in case I get lucky :mrgreen:
That's fine!
Yet I think, luck's what you'd need.
:)
One more tipp out of my tries in this direction: the engine or it's setting, that's best in Backward of the positions should have best chances with long computing started with empty hash.

I didn't try Sting very long till now, cause it can use 8G hash only, so TCs of more than an hour probably wouldn't help much with this one, even if it for sure is the best one as for cyclic Zugzwang together with lazy King- demand. E.g. it solves Constrictor(1), which I wouldn't have thought to be possible some years ago, when Mihai composed it.

As for tweaking his engine Marek could give Sting special code for each an every study probably, but he's fair enough to say so, if he did, and probably it would make the engine weaker as for other positions.

McBrain with switched off nullmove and maybe ICCF- option chosen too, Matefinder with Study Mode on are both good at Chichitza, Houdini (ev.Tactical) for some other positions too, maybe to change Houdini's 50 move- boundary could help as for the fortress Stubbornness, lowering it much, and or letting it compute rather long, question would be, if the output- line then would be a correct one, even it the eval would get somewhat near to 0.0 together with correct first move.

Here's some output after short computing with 50Moves set to 10 and Tactical1:

5: Mihai Neghina5 - Stubbornness, White to move and draw
8/b4kp1/2p1p3/5pB1/p1PP4/1p6/3K1P2/3N4 w - - 0 1

Analysis by Houdini 6.03 Pro x64-popc:


1.Kc3 Bxd4+ 2.Kxd4 b2 3.Nxb2 a3 4.Nd3 a2 5.Ne5+ Ke8 6.f4 a1Q+ 7.Kc5 Qa6 8.Kd6 Qb7 9.Kxe6 Qc7 10.Bh4 Qa5 11.Kxf5 Qa1 12.Bg5 Qh1 13.Ke6 Qd1 14.Kf5 Qb3 15.Kg6 Kf8 16.Kf5 Qh3+ 17.Ke4 Kg8 18.Kd4 Qa3 19.Bh4 Qd6+ 20.Ke3 g6 21.Bg5 Qa3+ 22.Kd4 Qb4 23.Ke3 Kg7 24.Kd3 Qb6 25.Ke2
-/+ (-1.25) Depth: 35/75 00:06:00 13167MN

Now one could say, -1.25 at least isn't a for sure losing one eval anymore, and of course, going into the line it's getting smaller more and more.
Yet 0.0 at the starting position still is something different and -1.25 isn't a real drawing eval neither.

Give me your news.
:)
I told my wife that a husband is like a fine wine; he gets better with age. The next day, she locked me in the cellar.
peter
Posts: 3185
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: Kornrade Test

Post by peter »

peter wrote: Fri Sep 21, 2018 8:58 pm 5: Mihai Neghina5 - Stubbornness, White to move and draw
[d]8/b4kp1/2p1p3/5pB1/p1PP4/1p6/3K1P2/3N4 w - - 0 1
Fritz 15 is to be mentioned too, when used by the GUI in a way called Monte Carlo- analysis, even if this isn't, what it's meant to be today.
It's just a GUI- feature, using the engine simply to play out lots of very fast games and judge the moves tried statistically.

F15 Monte Carlo is starting with 1.Kc3, even with Tree size set to narrow and with Search depht set to 5 plies, it reaches several thousand games in a few minutes for this one move only having a percentage of about 55%.

In a way one could call that solved too, even if I don't know, if and when other fist moves would come into trial too, neither the result then.
Peter.
lech
Posts: 1136
Joined: Sun Feb 14, 2010 10:02 pm

Re: Kornrade Test

Post by lech »

peter wrote: Fri Sep 21, 2018 8:58 pm ---
As for tweaking his engine Marek could give Sting special code for each an every study probably, but he's fair enough to say so, if he did, and probably it would make the engine weaker as for other positions.
---
The special code added to Sting is not a perfect work and need some modifications.
When I do some changes, the main goal is always to solve the "IQ" test.
Now I found why Sting was weak in 12. "Shield".
Thanks to the difficult positions I can make my ideas more perfect.
I believe that it is possible to make all the position solved without tricks, cheating or machine methods.
Maybe, I can't be friendly, but let me be useful.
ernest
Posts: 2041
Joined: Wed Mar 08, 2006 8:30 pm

Re: Kornrade Test

Post by ernest »

Kornrade wrote: Wed Sep 12, 2018 12:27 pm Hello all,

I have finally succeeded in making an up-to-date list of my studies on the personal site Chichitza
Does anybody know why this site, which also contained the Neghina studies, no longer works ?

TIA !
peter
Posts: 3185
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: Kornrade Test

Post by peter »

ernest wrote: Tue Nov 06, 2018 3:24 am Does anybody know why this site, which also contained the Neghina studies, no longer works ?
Looks like being under renovation again.
Peter.
Kornrade
Posts: 11
Joined: Fri Jun 15, 2012 11:18 am

Re: Kornrade Test

Post by Kornrade »

ernest wrote: Tue Nov 06, 2018 3:24 am
Does anybody know why this site, which also contained the Neghina studies, no longer works ?

TIA !
Yeah, sorry about that... The main page (where the list was) is not available, but the studies can still be reached individually:


01 Constrictor +- http://chichitza.ro/Chess.html?lang=en&aux=01 [FEN "3k4/2pq3p/pp5R/4P3/P6P/2PN4/1PN3K1/8 w - - 0 1"]
02 Mirror +- http://chichitza.ro/Chess.html?lang=en&aux=02 [FEN "6Nk/pp2Np1p/2p2Pp1/2R2bP1/7K/P7/1q1n1Q2/5n2 w - - 0 1"]
03 Walk +- http://chichitza.ro/Chess.html?lang=en&aux=03 [FEN "5B1q/3N3p/p3p2p/7k/B7/5PP1/K6P/8 w - - 0 1"]
04 Detour +- http://chichitza.ro/Chess.html?lang=en&aux=04 [FEN "7q/b1p5/1p1Npkb1/pPP2ppP/P1P5/3B2P1/5P1R/K3R3 w - - 0 1"]
05 Stubbornness == http://chichitza.ro/Chess.html?lang=en&aux=05 [FEN "8/b4kp1/2p1p3/5pB1/p1PP4/1p6/3K1P2/3N4 w - - 0 1"]
06 Guard Change +- http://chichitza.ro/Chess.html?lang=en&aux=06 [FEN "2k4N/Q1np4/2p2Bpp/1p1P4/pPP1p2P/P7/7q/1K6 w - - 0 1"]
Study 07 is missing on purpose. It was shown to be cooked and will be available after the publication of the repair (probably in spring).
08 Slowness +- http://chichitza.ro/Chess.html?lang=en&aux=08 [FEN "q7/b1p5/kp1p4/p2PN3/PPP5/1K6/8/5B2 w - - 0 1"]
09 Standstill == http://chichitza.ro/Chess.html?lang=en&aux=09 [FEN "1N4K1/1pp1p3/7k/2r3pP/p2pp1P1/P6N/8/2B5 w - - 0 1"]
10 Shelter == http://chichitza.ro/Chess.html?lang=en&aux=10 [FEN "4B3/k1K1pp2/p3bq2/2p1p3/1p2P1P1/4P2Q/nR6/8 w - - 0 1"]
11 Chase +- http://chichitza.ro/Chess.html?lang=en&aux=11 [FEN "4N2k/4PB1p/7P/B1p5/2P5/3q2Pp/1P3P1K/2r5 w - - 0 1"]
12 Desperation == http://chichitza.ro/Chess.html?lang=en&aux=12 [FEN "2b5/1pr4p/3bp1pk/1p6/1PpN2PP/K1P1n3/P3N1R1/3R4 w - - 0 1"]
13 Shield +- http://chichitza.ro/Chess.html?lang=en&aux=13 [FEN "8/4Pr1N/8/6PN/8/3K1ppp/4prkb/4Bbnn w - - 0 1"]
14 Fork-field +- http://chichitza.ro/Chess.html?lang=en&aux=14 [FEN "3B2r1/n3Pp2/P7/5k1p/q3NP2/7K/5P2/4NB2 w - - 0 1"]
15 Harassment +- http://chichitza.ro/Chess.html?lang=en&aux=15 [FEN "8/q1p1N1p1/3p1p1p/pk5P/1p2BB2/8/PK3P2/8 w - - 0 1"]
16 Smothering +- http://chichitza.ro/Chess.html?lang=en&aux=16 [FEN "6kq/2p4p/6P1/5nN1/K7/2P5/7B/3B4 w - - 0 1"]
* Smothering II +- http://chichitza.ro/Chess.html?lang=en&aux=16 [FEN "5k2/2p4p/6P1/1p3n2/3q4/PB4BN/KP5P/8 w - - 0 1"]
17 Fishing Rod == http://chichitza.ro/Chess.html?lang=en&aux=17 [FEN "8/1B3b1p/7P/p2p4/P2pp1kp/Q7/1N1Pq2P/7K w - - 0 1"]
18 Stronghold == http://chichitza.ro/Chess.html?lang=en&aux=18 [FEN "6k1/1qr1p2p/ppN5/3p1N2/6p1/1P1P2Pp/PP2PP1P/2RK4 w - - 0 1"]
19 Evil Eye == http://chichitza.ro/Chess.html?lang=en&aux=19 [FEN "8/pr1Pq3/br1kP3/1R2RQ2/5p2/1nPK4/5P2/8 w - - 0 1"]
20 Roundabout +- http://chichitza.ro/Chess.html?lang=en&aux=20 [FEN "4b3/5p2/2p4p/P1p1PN2/2pn2P1/P2k4/Pr1npKN1/B7 w - - 0 1"]
ernest
Posts: 2041
Joined: Wed Mar 08, 2006 8:30 pm

Re: Kornrade Test

Post by ernest »

Thanks, Peter and Kornrade !
Dann Corbit
Posts: 12538
Joined: Wed Mar 08, 2006 8:57 pm
Location: Redmond, WA USA

Re: Kornrade Test

Post by Dann Corbit »

For these two positions, I do not have a mechanical solution:
[d]7q/b1p5/1p1Npkb1/pPP2ppP/P1P5/3B2P1/5P1R/K3R3 w - -
[d]2k4N/Q1np4/2p2Bpp/1p1P4/pPP1p2P/P7/7q/1K6 w - -
For this one, I have the key move, but the score is zero:

2k4N/Q1np4/2p2Bpp/1p1P4/pPP1p2P/P7/7q/1K6 w - - acd 29; ce 0; acs 394; acn 905837; pv Nf7 Qg3 Qc5 Ne8 dxc6 Qd3+ Kc1 Qf1+ Kd2 Qg2+ Ke1 Qh1+ Ke2 Qg2+ Ke1 ; bm Nf7;

It was produced by an automated back and forth analysis over this move sequence:
[pgn][Event "?"] [Site "?"] [Date "????.??.??"] [Round "?"] [White "?"] [Black "?"] [Result "*"] [SetUp "1"] [FEN "2k4N/Q1np4/2p2Bpp/1p1P4/pPP1p2P/P7/7q/1K6 w - - 0 1"] 1. Nf7 Qg3 2. Qxc7+ Qxc7 3. d6 Qb8 4. c5 Kb7 5. Nd8+ Ka6 6. Kc2 Qc8 7. Kd2 Qb8 8. Ke3 Qc8 9. Kxe4 Qb8 10. Be7 Qc8 11. Ke5 Qa8 12. Kf6 g5 13. hxg5 hxg5 14. Kxg5 Qc8 15. Kf6 Ka7 16. Kf7 Ka8 17. Ke8 Ka7 18. Bf8 Ka8 19. Ke7 Ka7 20. Ne6 Ka8 21. Nd4 Ka7 22. Nf3 Ka8 23. Ne5 Ka7 24. Nxd7 Qb7 25. Bg7 Ka6 26. Bf6 Ka7 27. Ke8 Qc8+ 28. Bd8 Kb7 29. Nf6 * [/pgn]
A 62 ply search on the root position proved fruitless.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.