"Poor Fish": very hard test suite

[peter]

So the winners are...

Code: Select all

DBI 2 (super hard)
1b6/2p5/1kB5/1PN5/8/3K4/8/q7 w - - bm Nb3; DBI 2
1r3rk1/5p1p/p2pb1p1/3Np1P1/2pRPR2/qP2Q1N1/P1PK3P/8 w - - bm Nf5; c1 "diff=5"; DBI 2
2r1r1k1/4bppp/1q4n1/3pP3/p1pP4/PnR2NBP/1P4P1/1B1QR2K w - - bm e6; c1 "diff=8"; DBI 2
...

The third of those seems doubteable to me as for the single best move:

[pgn]
[Event "Costalba DBT2"]
[Site "?"]
[Date "2016.09.28"]
[Round "?"]
[White "Lejeune, Vincent"]
[Black "Hard, 2016Nr29"]
[Result "*"]
[Annotator "Martan,Peter"]
[FEN "2r1r1k1/4bppp/1q4n1/3pP3/p1pP4/PnR2NBP/1P4P1/1B1QR2K w - - 0 1"]
[SetUp "1"]
1.e6 ( 1.Bf5 $5 Nf8 2.Bxc8 Rxc8 3.Rf1 Ne6 4.Bf2 Rf8 5.Be3 Bd8 6.Rc2 Qb7 7.Qe1 Qa6 8.Qf2 Bb6 9.Rd1 Qb5 10.Qg3 $16 ) 1...f6 2.Bf2 ( 2.Bxg6 $5 hxg6 3.Nh4 Qxd4 4.Qc2 f5 5.Nxg6 Qf6 6.Nf4 Rcd8 7.Qd1 d4 8.Rxc4 d3 9.Nxd3 Nc5 10.Rxc5 $16 ) ( 2.Nh4 $5 Nxh4 3.Bxh4 ( 3.Bxh7+ Kxh7 4.Qh5+ Kg8 5.Bxh4 Nxd4 6.Qf7+ Kh8 7.Qh5+ Kg8 8.Qf7+ Kh8 9.Qh5+ $11 ) 3...f5 ( 3...g6 $2 4.Bxg6 $18 ) 4.Bxf5 Qxd4 5.Bxe7 Qxd1 6.Rxd1 d4 7.Bb4 dxc3 8.e7 c2 9.Bxc2 g6 10.Re1 $16 ) 2...Bf8 3.Nh4 ( 3.Bf5 $6 Ne7 4.Bxh7+ Kxh7 5.Nh4 Kg8 6.Qh5 Nd2 ( 6...Nxd4 $2 7.Qf7+ Kh7 ( 7...Kh8 $2 8.g4 $18 ) 8.Rg3 Qxb2 9.Bxd4 Qxd4 10.Ng6 $18 ) 7.Qf7+ Kh8 8.Qh5+ Kg8 9.Rg3 Qxb2 10.Qf7+ Kh8 11.Qh5+ $11 ) 3...Nxh4 4.Bxh4 Rxe6 5.Qh5 Rxe1+ 6.Bxe1 g6 7.Qxd5+ Kh8 8.Rxc4 Rxc4 9.Qxc4 Qxd4 10.Qxd4 Nxd4 $16 *
[/pgn]

Have I missed a winning line for 1.e6 and does anybody know the source of the position?
Best

[Vinvin]

So the winners are...

Code: Select all

DBI 2 (super hard)
1b6/2p5/1kB5/1PN5/8/3K4/8/q7 w - - bm Nb3; DBI 2
1r3rk1/5p1p/p2pb1p1/3Np1P1/2pRPR2/qP2Q1N1/P1PK3P/8 w - - bm Nf5; c1 "diff=5"; DBI 2
2r1r1k1/4bppp/1q4n1/3pP3/p1pP4/PnR2NBP/1P4P1/1B1QR2K w - - bm e6; c1 "diff=8"; DBI 2
...

The third of those seems doubteable to me as for the single best move:

...

Have I missed a winning line for 1.e6 and does anybody know the source of the position?
Best

Probably this line (nice advantage but I'm not sure it's winning) :

[Event "?"]
[Site "?"]
[White "W"]
[Black "B"]
[Result "*"]
[FEN "2r1r1k1/4bppp/1q4n1/3pP3/p1pP4/PnR2NBP/1P4P1/1B1QR2K w - - 0 1"]
[Comment "bm e6; c1 "diff=8";"]

1. e6 f6 2. Bc2 Bd6 3. Nh4 Bxg3 4. Rxg3 Qxd4 5. Nxg6 Qxd1 6. Rxd1 hxg6 7.
Rxd5 Rxe6 8. Bxg6 Kf8 9. Rg4 Rc7 *

[Vinvin]

Vincent, it would be great if you could double check these positions (at least the DBI 2 group) with your multiple-run method to verify these are really hard and DBT works.

I've started an analyses for the 160 positions, 3 runs.
Will end probably in 12 hours.

[Dann Corbit]

2r1r1k1/4bppp/1q4n1/3pP3/p1pP4/PnR2NBP/1P4P1/1B1QR2K w - - acd 31; acs 3967; bm e6; c1 "diff=8"; c2 "Gaprindashvili-Bliss, CorrGame 1994"; c3 "e6"; ce 130; id "W.Eigenmann: Brillant 057"; pm e6; pv e6 f6 Nh4 Nxh4 Bxh4 f5 Bxf5 Qxd4 Bxe7 Qxd1 Rxd1 d4 Bb4 dxc3 e7 c2 Bxc2 g6 Re1 Kg7 Kh2 Rc7 Re4 Kf7 Rh4 Kg7 Bc3+ Kg8 Be4 Rc5 Bb4 c3 Bxc3 Rxe7 Bb4 Ree5 Bxc5 Nxc5 Bf3 Kg7 Kg3 Re8 Rc4 Re5 Kh4 Kf6 g4 Ke6 Kg3;

[Dann Corbit]

See also:
https://glarean-magazin.ch/2010/01/22/d ... n-magazin/

[peter]

[
Probably this line (nice advantage but I'm not sure it's winning) :

[Event "?"]
[Site "?"]
[White "W"]
[Black "B"]
[Result "*"]
[FEN "2r1r1k1/4bppp/1q4n1/3pP3/p1pP4/PnR2NBP/1P4P1/1B1QR2K w - - 0 1"]
[Comment "bm e6; c1 "diff=8";"]

1. e6 f6 2. Bc2 Bd6 3. Nh4 Bxg3 4. Rxg3 Qxd4 5. Nxg6 Qxd1 6. Rxd1 hxg6 7.
Rxd5 Rxe6 8. Bxg6 Kf8 9. Rg4 Rc7 *

That's the point, Vincent, I'm not sure neither.
At the end of your line SF's eval is even a little lower than at the end of my pv and it's not any better than at the end of the 1.Bf5- line.
That we are not sure of the winning of 1.e6 wouldn't matter, a bm doesn't have to be a clear game- winner to me, but if there indeed is a second move winning as well or as well not...

[Vinvin]

See also:
https://glarean-magazin.ch/2010/01/22/d ... n-magazin/

Thanks, Dann !

2 winning lines :
1. e6 f6 2. Nh4 Nxh4 3. Bxh4 g6 4. Bxg6 hxg6 5. Rg3 Kh7 6. Rxg6 Kxg6 7. Qg4+ +-
1. e6 Bf6 2. exf7+ Kxf7 3. Ne5+ +-

Reference : Gaprindashvili-Bliss, CorrGame 1994

[Vinvin]

I have just finished testing with Double Blind Test (DBT) the full suite of 160 hard positions provided by Vincent. I have used current Stockfish modified by me to run automatically the DBT test on a epd file without need of human supervisor, so it ran overnight.
...

Why this position is not selected ?
1r6/4k3/r2p2p1/2pR1p1p/2P1pP1P/pPK1P1P1/P7/1B6 b - - bm Rxb3;

[peter]

Thanks for the source, Dann.
The line you give is up to move 10 within my pgn already in braces, if you want to have a look at it again.

At the end of your line, I get this SF- output with full Syzygys:

[D]8/7p/4k1p1/2n1r3/p1R3P1/P4BKP/1P6/8 b - - 0 1

Analysis by pedantFishW_2016-09-25_popcnt:

25...h6 26.h4 g5 27.h5 Kd6 28.Rd4+ Ke7 29.Bd1 Re1 30.Bxa4 Rb1 31.Rb4 Nxa4 32.Rxa4 Rxb2 33.Ra6 Rb3+ 34.Kg2 Kf7 35.Rxh6 Rxa3 36.Rd6 Ra5 37.Rd3 Kg7 38.Kf3 Kf6 39.Ke4 Ke6 40.Rc3 Re5+ 41.Kd3 Ra5 42.Ke2 Kf6 43.Rf3+ Ke6 44.Kf2 Ra4 45.Kg3 Ra5 46.Kh3 Ra1 47.Rf2 Ra3+ 48.Kg2 Ra4 49.Kf3 Ke7 50.Kg3 Ra3+ 51.Kg2 Ra1 52.Rc2
+/- (1.54) Depth: 40/65 00:11:07 20976MN, tb=50375321

Bliss played 1...Bf6 instead of ...f6 in the original corr.- game, as it seems maybe this was suboptimal.

But my main doubt isn't the winning of 1...e6, which I like more than 1.Bf5(!?) too, but it's not quite easy to discriminate it as single best move as for my pov.
I'd have no problem with giving 1.Bf5 as a less beautiful and maybe even not quite as good seond move, meaning, if an engine having it in output maybe wouldn't be fully wrong neither.

[peter]

2 winning lines :
1. e6 f6 2. Nh4 Nxh4 3. Bxh4 g6 4. Bxg6 hxg6 5. Rg3 Kh7 6. Rxg6 Kxg6 7. Qg4+ +-
1. e6 Bf6 2. exf7+ Kxf7 3. Ne5+ +-

No, Vincent, sorry, but no.
I have 3...g6(?) after 2...f6 2.Nh4 Nxh4 3.Bh4 already as a black blunder in my .pgn posted here:
http://www.talkchess.com/forum/viewtopi ... 46&t=61526
, if you want to have a look again, 3...f5 instead is clearly better.
And in the original game 1...Bf6 is at least ?!, here ...f6 is clearly better