The versions below, No.1146.1;2, are published after the end of tournament, 28.10./12.11.2017, but before the award is made, accepted by the judge of PG/Retros 2015-16.
Definition: (click to show/hide)
Masand: A unit which moves and actively checks changes the colour of all the units – enemy and its own (not Ks) – which it observes after the move. A checking rear (non-moving) piece of a battery would not recolour the units it observes.
|
No.1146 Paul Rãican
Romania
original – 24.10.2016
|
Solution: (click to show/hide)
|
|
white Bf1c1 Ke1 Qd1 Ph2g2f2e2d2c2b2a2 Sg1b1 Rh1a1
black Bf8c8 Ke8 Qd8 Ph7g7f7e7d7c7b7a7 Sg8b8 Rh8a8
PG 16.5 (8+11)
Masand
|
1.a2-a4 d7-d5 2.Ra1-a3 d5-d4 3.Ra3-h3 d4-d3 4.Sb1-c3 d3*c2 5.d2-d4 Qd8*d4 {
} 6.Bc1-h6 c2*d1=Q [d4=w] [a4=b] [e2=b] 7.Sc3*d1 e2*f1=R[g1=b] [f2=b]+ {
} 8.Ke1*f1 Sg1-f3 9.Kf1-e2 f2-f1=B [g2=b]+ 10.Rh1*f1 g2-g1=S [f3=w] [h3=b]+ {
} 11.Ke2-d3 Sb8-c6 12.Qd4-d8 [c8=w] [c7=w] [e7=w] Sc6*d8 13.e7*f8=R [g8=w] [f7=w]+ {
} Ke8*f8 14.Sg8-f6 Kf8-e7 15.f7-f8=B [g7=w] Rh8*f8 16.g7-g8=S [h6=b] [f6=b] Sf6*g8 {
} 17.Rf1*g1 {dia
(C-)}
qrbs-black Schnoebelen, RBS-white Schnoebelen in echo. (Author)
|
|
|
No.1146.1 Paul Rãican
Romania
version of No.1146 – 28.10.2017
|
Solution: (click to show/hide)
|
|
white Bf1c1 Ke1 Qd1 Ph2g2f2e2d2c2b2a2 Sg1b1 Rh1a1
black Bf8c8 Ke8 Qd8 Ph7g7f7e7d7c7b7a7 Sg8b8 Rh8a8
PG 17 (10+9)
Masand
|
1.a2-a4 d7-d5 2.Ra1-a3 d5-d4 3.Ra3-h3 d4-d3 4.Sb1-c3 d3*c2 5.d2-d4 Qd8*d4 6.Bc1-e3 {
} c2*d1=Q [d4=w][a4=b][e2=b]+ 7.Sc3*d1 e2*f1=R [g1=b][f2=b]+ {
} 8.Ke1*f1 Sg1-f3 9.Kf1-e2 f2-f1=B [g2=b]+ 10.Rh1*f1 g2-g1=S [f3=w] [h3=b]+ {
} 11.Ke2-d3 Sb8-c6 12.Qd4-d8 [c8=w][c7=w][e7=w]+ {[switchback]} Sc6*d8 {
} 13.e7*f8=R [g8=w][f7=w]+ Ke8*f8 14.Sg8-h6 Kf8-e7 15.f7-f8=B [g7=w]+ Rh8*f8 {
} 16.Be3*g1 Rf8-f6 17.g7-g8=S [h6=b][f6=w]+ Sh6*g8 {dia
(C-)}
qrbs-black Schnoebelen, RBS-white Schnoebelen in echo. (Author)
|
|
|
No.1146.2 Paul Rãican
Romania
version of No.1146 – 12.11.2017
|
Solution: (click to show/hide)
|
|
white Bf1c1 Ke1 Qd1 Ph2g2f2e2d2c2b2a2 Sg1b1 Rh1a1
black Bf8c8 Ke8 Qd8 Ph7g7f7e7d7c7b7a7 Sg8b8 Rh8a8
PG 17 (13+6)
Masand
|
1.a2-a4 d7-d5 2.Ra1-a3 d5-d4 3.Ra3-h3 d4-d3 4.Sb1-c3 d3*c2 5.d2-d4 Qd8*d4 {
} 6.Bc1-e3 c2*d1=Q [d4=w][a4=b][e2=b]+ 7.Sc3*d1 e2*f1=R [f2=b][g1=b]+ {
} 8.Ke1*f1 Sg1-f3 9.Kf1-e2 f2-f1=B[g2=b]+ 10.Rh1*f1 g2-g1=S [f3=w][h3=b]+ {
} 11.Ke2-d3 Bc8-f5 [h3=w][h7=w]+ 12.Qd4-d7 [f5=w][a4=w][c7=w][e7=w]+ Sb8*d7 {
} 13.e7*f8=R [f7=w][g8=w]+ Ke8*f8 14.Sg8-h6 Kf8-e7 15.f7-f8=B[g7=w]+ Rh8*f8 {
} 16.Be3*g1 Rf8-f6 17.g7-g8=S [f6=w][h6=b] Sh6*g8 {dia
(C?)}
qrbs-black Schnoebelen, RBS-white Schnoebelen in echo. (Author)
|
Julia, the n° of this diag is 1146 !
Dmitri’s Py2Web is priceless for such problems.
It makes me aware that I don’t actually see why not 11.Rxg1/17.Kd3.
Because after 11. Rxg1 Sc6, 12. Qd8 results in a recolouring of Sd1.
Thanks Joost!
Nice line-closing and of course unexpected for those (it’s me) who unduly but stubbornly look for the answer only on e,f,g files.
Quite a misconception 🙁
Quite an achievement, do I see it right as Schnoebelen AUW by both sides? Great choice of fairy condition right for the idea…
No white queen promotion.
Personally, I think the chance of this being correct is less than 1%, but it takes too much time to find the cook.
Cooked in 15.5 moves:
1. d4 e5 2. Bg5 Qxg5 3. dxe5 Qxe5 4. a4 Qc3 [c7,g7=w][b2,c2=b]+ 5. Sxc3 Ke7 6. Ra3 cxd1=Q [d7=w][e2,a4=b]+ 7. Sxd1 Sc6 8. Kd2 exf1=R 9. Rh3 Sd8 10. gxh8=B Rxf2 [f7=w][g2=b]+ 11. Kd3 Bh6 12. Bf6 [b2=w]+ Rxf6 13. dxc8=B Rf1 14. f8=R Rf3 [h3,f8=b]+ 15. Sxf3 g1=B 16. Rxg1
This solution was found by an experimental fairy PG solver I’m working on. The program uses heuristics to search the most likely moves first, and this problem seemed like a good test given Joost’s comment. In 25 min it found cooks in both 16.5 and 16.0 moves. Letting it run for 13h, it found the cook shown above in 15.5 moves.
My program looks really promising so I’ll keep working on it and I plan to make it available. Timescale: a few months.
Astounding.
Hats off!
Even if the problem has been broken, I am glad that we finally have such a versed solver in Canada. Congratulations, Francois! A little question: If the last white move is dropped, is the problem still cooked?
Yes, of course: 15….g1=S!
Version 1146.1 is cooked by Jacobi in “PG demolition mode”:
1.Sf3 e6 2.a4 Qf6 3.d3 Qc3 [c7,g7=w][b2,c2,d3=b]+ 4.Sxc3 dxe2 5.Be3 f6 6.Bxe2 Ke7 7.gxf8=R cxd1=Q [d7=w][a1,e2,a4=b]+ 8.Sxd1 Rc1 9.Kxe2 Rc2 [b2=w][c7=b]+ 10.Kd3 Rxf2 11.Rf1 Rxg2 12.Bg1 Sc6 13.Rxf6 Sh6 14.d8=Q [c7,c8,h8=w]+ Sxd8 15.Bxe6 Rg3 16.Bc8 Rh3 17.Rg8 Sxg8
Finding this cook was harder for my program than for 1146. It took 15h to find this cook on my computer. Trying in 16.5 moves, it hasn’t found any cook after 43 hours.
A very hidden alternative game, virtually impossible to discover with the human brain. Chapeau!
1146.2 isn’t C+, contrary to what the caption says. At publication time the problem had merely survived 4 days of “demolition mode” on my computer. In an unfortunate game of “Chinese whispers” between me, Paul and Julia, this information turned into “checked for 4 days” and then into the label “C+”.
Today the problem has survived 14 days of demolition mode, but all this means is that Jacobi v0.1 finds no obvious cook so far. The problem is still C-.
Actually I prefer three-state logic:
C+ – fully tested and considered sound by at least one program
C- – unsound
C? – anything else = no defect found yet, but not yet fully tested by any program yet
In this logic, 1146.2 is still C?.
I followed the JF notation which seems to be
C+: sound
cooked: unsound
C-: unknown
The letter C comes from the word “computer”, so maybe you could write “C-” for a cook found by a computer, but I don’t see why you’d write “computer minus” for a cook found by a human. You’d need four-state logic.
When the problem is fully tested and considered sound by two or more programs/computers, then should we use C++ or C#? 🙂
In general, I test all problems which can be tested by Popeye/WinChloe, and mark them as C+ if sound, or in some cases with additional comment about C+/C- by different versions of the programs.
For the problems with fairy elements not implemented in Popeye/WinChloe I’ve used C-, but maybe it would be more correct to use C?
Have changed 1146.2 to C? for now. But this case was really very much confusing for me, as this version was sent by Paul with the comment “Must say that F. Labelle already checked this last version (for 4 or 5 days) with Jacobi and don’t found an alternative game“; some days later there’s another comment “F. Labelle informs that after 9 days of “demolition mode” his computer is still running. It hasn’t found a cook and it hasn’t found the intended solution either“; and later Paul writes again “This problem is sound“?!
All of it together feels to me very much as C??? 🙂
But seriously, to mark those problems I can’t test myself with C+, probably I have to ask authors for the reports from their solving programs.
About the choice of notation for problems with unknown status (C- vs C?), with Google I found two definitions in favor of C-:
C- : problem was not tested by computer program or was tested only partially (Award of Marián Križovenský 55 JT)
C-. Computer-tested only partially (WCCI 2013-2015. Fairies)
and one in favor of C?:
C- means unsound problem. C? means not computer tested problem. (c2_text_en.doc on http://www.jurajlorinc.com)
Maybe someone subscribed to every chess magazine can tell us which convention each magazine is using?
In order to simplify the situation, there is another option followed e.g. by WinChloe:
C- = cooked
Nothing = not computer tested.
My favorite is clearly in favor of C?, so that each problem has some “C status”:
C+ = correct
C- = cooked
C? = not computer tested
Obviously C+ and C- are definitive (relatively to a given program and version), while C? may be changed to C+ or C- some day.
The big question is whether C- should denote a cooked or a not computer tested problem. François said this is 2-1 in favor of not computer tested, but if we add WinChloe’s opinion, this is 2-2!
Wouldn’t it be absurd using C- to indicate an ORIGINAL which a computer has PROVED to be cooked?
Nikola, I believe C- might happen after replacing C? with it. I won’t publish a problem proved as cooked, but could publish not tested one…
Yes Julia, that’s what I say.
C- or C? could mean “not tested” or perhaps that a computer didn’t find the solution which the humans are able to see.
A trivial example, Popeye can’t solve Loyd’s #2: Ke6,Qa6/Ke8,Ra8,Pa7c7, so it’s not C+ 🙂
Of course, as this is a retro problem… An amount of retro problems are 100/100 correct because the reasoning is enough easy to avoid any human mistake, but are nevertheless C? because computers can’t handle such retro problems, even trivial for human brains.
More votes (notation of problems with unknown status, retros section):
feenschach: nothing
PDB: nothing
Phenix: nothing
Problem Paradise: nothing
Julia Fairies: C- (prior to 1146.2)
Strategems: C-
Problemist: C?
Version 1146.2 is cooked by Jacobi after about 30 days in “PG demolition mode”:
1.a4 d5 2.Ra3 Bf5 3.e4 dxe4 4.Qh5 Sh6 5.Qxf7 [f5,e7,g7,f8=w]+ Kxf7 6.Ke2 Rxf8 7.g8=Q [h7,f8=w][g2=b]+ Kxe7 8.Rh3 Sxg8 9.f3 Qd3 [e4=w][c2,d2,f3=b]+ 10.Kxd3 d1=S 11.Sxf3 gxf1=S 12.Be3 Sg3 13.Rf1 c1=R 14.Bg1 Rc3 [c7=w]+ 15.Sxc3 Sxe4 16.Sxd1 Sef6 17.Rxf6 Sd7
This is a spectacular alternative solution. Thank you, François, for your patience!