Test Position - avoid move b3

[gordonr]

[d]1r2r1k1/4bppp/2qp4/p3n2n/Ppp1P3/5P2/BPPQNBPP/3R1R1K w - - 1 23

Avoid b3. Ra1 is better; maybe others too.

Komodo 9.42 struggles with this one - took about 1 hour on my 8 core. Latest Stockfish about 6 mins. Houdini 4 (default config; not tactical mode) and Fizbo 1.7 appear to "understand" immediately.

This is from a very quick game (1 min + 8 sec increment) where Komodo played b3, allowing c3 and getting its bishop blocked out for the rest of the game. Texel, as Black, eventually opened up the centre/kingside and won with its "extra" piece.

[peter]

Hi Gordon!

[d]1r2r1k1/4bppp/2qp4/p3n2n/Ppp1P3/5P2/BPPQNBPP/3R1R1K w - - 1 23

Avoid b3.

Have my problems with this positions cause I'd always like to have a best move or at least some candidates with distiinctable eval- relations between the ones to be chosen and the ones to be avoided, that's not so easy here, as well as for engines evals as for my own ones.

The lines I tried all tend to rather decrease White's advantage and look drawish at the first few moves, with 23.b3 maybe especially so, but with others too, so it's difficult for me to prove b3 a clear blunder.

Yet as for human positional understanding I'd say you're fully right, why bury your Bishop without any need?
And why isn't there any penalty in engines' positional eval for that obvious follow- up position after ...c3, which is the answer to be expected at once anyhow?

[peter]

The lines I tried all tend to rather decrease White's advantage and look drawish at the first few moves, with 23.b3 maybe especially so, but with others too, so it's difficult for me to prove b3 a clear blunder.

Did let komodo and SF play 4 games against each other with change of colours twice for 23.b3 and for 23.Ta1 each.

Two times I adujdicated as drawn after about 35 moves, once this was judged after several 0.00- moves by the GUI at White's 35th, but there was one single game I did let run a little longer because Black got advantage and came to a won position:

[Event "SCHWARZERBLITZ, Lang 100m+30s"]
[Site "SchwarzerBlitz"]
[Date "2016.05.03"]
[Round "3"]
[White "Komodo 9.42 64-bit"]
[Black "Stockfish 010516 64 POPCNT"]
[Result "0-1"]
[Annotator "0.67;0.00"]
[SetUp "1"]
[FEN "1r2r1k1/4bppp/2qp4/p3n2n/Ppp1P3/5P2/BPPQNBPP/3R1R1K w - - 0 23"]
[PlyCount "43"]
[TimeControl "6000+30"]

{Intel(R) Xeon(R) CPU X5650 @ 2.67GHz W=37.2 plies; 11.596kN/s; 1.
618.897 TBAs B=34.2 plies; 10.180kN/s; 819.521 TBAs} 23. b3 c3 24. Qe3 Bd8 25.
Nd4 Qd7 {Beide letzter Buchzug} 26. g4 {[%eval 67,31] [%emt 0:02:03]} Nf6 {
[%eval 0,30] [%emt 0:03:47]} 27. Nf5 {[%eval 59,32] [%emt 0:00:00]} Bc7 {
[%eval 0,27] [%emt 0:00:34] (Te6)} 28. Bg3 {[%eval 73,27] [%emt 0:03:57] (De2)}
Rbd8 {[%eval 0,34] [%emt 0:01:22] (h6)} 29. Bf4 {[%eval 16,29] [%emt 0:05:55]
(Lh4)} d5 {[%eval -77,31] [%emt 0:04:36] (De6)} 30. Nxg7 {[%eval -58,34] [%emt
0:14:28]} Kxg7 {[%eval -158,38] [%emt 0:00:00]} 31. Bh6+ {[%eval -109,35]
[%emt 0:02:10]} Kg8 {[%eval -165,39] [%emt 0:00:00]} 32. Qg5+ {[%eval -127,35]
[%emt 0:01:42]} Ng6 {[%eval -182,38] [%emt 0:01:39]} 33. Qxf6 {[%eval -125,37]
[%emt 0:00:00]} Be5 {[%eval -189,35] [%emt 0:01:35]} 34. Qf5 {[%eval -138,37]
[%emt 0:00:27]} Bf4 {[%eval -191,37] [%emt 0:01:06]} 35. Qxd7 {[%eval -159,42]
[%emt 0:10:18]} Rxd7 {[%eval -201,41] [%emt 0:00:00]} 36. Bxf4 {[%eval -165,40]
[%emt 0:01:49]} Nxf4 {[%eval -220,34] [%emt 0:02:23]} 37. Kg1 {[%eval -174,42]
[%emt 0:00:14] (Lb1)} d4 {[%eval -241,33] [%emt 0:03:05]} 38. Bb1 {[%eval -173,
44] [%emt 0:00:00]} Ne6 {[%eval -244,31] [%emt 0:00:31]} 39. Kf2 {[%eval -212,
42] [%emt 0:12:22]} Rc8 {[%eval -265,36] [%emt 0:00:00] (Sc5)} 40. Ke2 {[%eval
-210,34] [%emt 0:01:58] (Tc1)} Nc5 {[%eval -289,31] [%emt 0:02:45]} 41. Rg1 {
[%eval -213,39] [%emt 0:00:00]} d3+ {[%eval -318,33] [%emt 0:04:17] (Kg7)} 42.
Ke3 {[%eval -221,39] [%emt 0:01:09]} Kg7 {[%eval -317,33] [%emt 0:00:01]} 43.
h3 {[%eval -245,42] [%emt 0:09:29] (Tc1)} d2 {[%eval -456,31] [%emt 0:02:54]}
44. h4 {[%eval -314,41] [%emt 0:00:00] adjud.} 0-1

Both engines had 12 threads each with ponder on and 8G hash.
Of course a sinlge game doesn't mean much and 4 games don't neither, yet I still don't see the why engines consider 23.b3 best, most of them doing so for quite a hardware-time

[gordonr]

Hi Peter,

Sorry for the delay in replying. I wanted to do some more analysis first. As well as using Komodo and Stockfish, I've been finding analysis from Fizbo 1.7 to be interesting in this position - it is more optimistic about Black winning.

The more I look at b3, the more it looks to me like White is struggling for a draw. Some example lines where I see Black with, at least, a small advantage and growing:

[pgn][Event "Computer chess game"]
[Site ""]
[Date "2016.05.03"]
[Round "?"]
[White "GR"]
[Black "Stockfish_1_May_2016"]
[Result "*"]
[WhiteElo "2400"]
[BlackElo "2400"]
[SetUp "1"]
[FEN "1r2r1k1/4bppp/2qp4/p3n2n/Ppp1P3/5P2/BPPQNBPP/3R1R1K w - - 0 23"]
[PlyCount "20"]
[EventDate "2016.??.??"]

23. b3 c3 24. Qe3 Bd8 25. Nd4 Qd7 26. g4 Nf6 27. Nf5 Bc7 28. Bg3
(28. Qg5 Ng6 29. Bd4 Qe6 30. Bxf6 Qxf6 31. Qxf6 gxf6)
28... Qc6 29. Bxe5
(29. Qe2 Rbd8 30.Qb5 Qxb5 31. axb5 d5 32. g5 Nh5)
(29. Nd4 Qa6 30. Nb5 Rxb5 31. axb5 Qxb5 32.Rde1 (32. Qf2 Bb6 33. Qg2 Bc5 34. Rfe1 Qb6 35. h3 g5 36. h4 h6 37. Re2 Ng6)
32... h6 33. Qe2 Qc6)
(29. Rd4 d5 30. g5 Nh5 31. Rxd5 Qa6 32. Qf2 Bb6)
29...dxe5 30. Qg5 g6 31. Qh6 gxf5 32. gxf5 Kh8 *[/pgn]


I agree that the position is unclear enough to prevent it from being a definite good test position.

cheers

Gordon

[F. Bluemers]

Hi,
Interesting position.
I agree that b3 does not look very good....
I don't like the Ra1 suggestion as a rook locked in on a1 is not much better.
Maybe 1.f4 might work.
Best
Fonzy

[rabbits23]

What's wrong with Qd5?
Allan

[peter]

The more I look at b3, the more it looks to me like White is struggling for a draw. Some example lines where I see Black with, at least, a small advantage and growing:

[pgn][Event "Computer chess game"]
[Site ""]
[Date "2016.05.03"]
[Round "?"]
[White "GR"]
[Black "Stockfish_1_May_2016"]
[Result "*"]
[WhiteElo "2400"]
[BlackElo "2400"]
[SetUp "1"]
[FEN "1r2r1k1/4bppp/2qp4/p3n2n/Ppp1P3/5P2/BPPQNBPP/3R1R1K w - - 0 23"]
[PlyCount "20"]
[EventDate "2016.??.??"]

23. b3 c3 24. Qe3 Bd8 25. Nd4 Qd7 26. g4 Nf6 27. Nf5 Bc7 28. Bg3
(28. Qg5 Ng6 29. Bd4 Qe6 30. Bxf6 Qxf6 31. Qxf6 gxf6)
28... Qc6 29. Bxe5
(29. Qe2 Rbd8 30.Qb5 Qxb5 31. axb5 d5 32. g5 Nh5)
(29. Nd4 Qa6 30. Nb5 Rxb5 31. axb5 Qxb5 32.Rde1 (32. Qf2 Bb6 33. Qg2 Bc5 34. Rfe1 Qb6 35. h3 g5 36. h4 h6 37. Re2 Ng6)
32... h6 33. Qe2 Qc6)
(29. Rd4 d5 30. g5 Nh5 31. Rxd5 Qa6 32. Qf2 Bb6)
29...dxe5 30. Qg5 g6 31. Qh6 gxf5 32. gxf5 Kh8 *[/pgn]


I agree that the position is unclear enough to prevent it from being a definite good test position.

At the one already shown game komodo- SF,

http://www.talkchess.com/forum/viewtopi ... 20&t=60022

Black came to clear advantage after 28...Rbd8 instead of your ...Qc6, but komodo's reply 29.Bf4?! seemed to be supoptimal at least, with Bh4 instead Black still has the better edge, I guess, but isn't for sure winning.
If it is with best White play after 28...Qc6, I doubt too, but as your lines show clearly, engine- evals start climbing for Black's side and the ones right after 23.b3 are too good for White from the startup and get relativated by Backward.

In my lines White deviates from your main line with 26.Nb5 instead of g4 and in one game SF- komodo that went on like this:

1r2r1k1/4bppp/2qp4/p3n2n/Ppp1P3/5P2/BPPQNBPP/3R1R1K w - - 0 23

23. b3 c3 24. Qe3 Bd8 25. Nd4 Qd7 26. Nb5 Re6 27. Qa7 Qxa7 28. Bxa7 Bb6 29. Bxb6 Rxb6 30. Nc7 Rh6 31. Kg1 g5 32. Nd5 Ra6 33. Nc7 Rc6 34. Nd5 Ra6 35. h3

Here I adjudicated as equal.

Overall I'd say the position is about eval, but I do see more problems for White after 23.b3 yet, and these are self- created problems only, it cannot be good to continue game with one light figure less without any compensation.

The best move yet isn't easy to name cause Black can hold after 23.Ra1 as well as after e.g. 23.Nd4, 23.f4, 23.Ba7 as well, I guess, and I don't see a forced White win neither at any try of mine.

As for good or bad test positions: it's not any of the "classical" best move or avoid move problems but it is as good as each and any other chess position is, to test lines and eval-relations and eval- courses over the lines of engines and human players, I'm not at all fixed to a certain kind of that to call it test position. As well as we have to forget about Celo from engine- engine- ganes only more and more, is the more to forget about classisal position-tests too.

The very good point of this position still is to me, why are nearly all the engines so fixed to 23.b3 (?!) which at least is a clear positional blunder to me.
Houdini and Texel in fact are the only ones I found till now not to chose b3 as for default settings and not too long hardware- time

[peter]

Houdini and Texel in fact are the only ones I found till now not to chose b3 as for default settings and not too long hardware- time

And not to forget Fizbo, I just downloaded it, here's its output:

1r2r1k1/4bppp/2qp4/p3n2n/Ppp1P3/5P2/BPPQNBPP/3R1R1K w - - 0 1

Analysis by Fizbo 1.7:

23.Dd5 Dxd5 24.Txd5 b3 25.cxb3 cxb3 26.Lb1
=/+ (-0.34) Tiefe: 7/15 00:00:00 11kN
23.Dd5 Dxa4 24.Lxc4 Dxc2 25.Tc1 Dxb2 26.La6 b3
=/+ (-0.63) Tiefe: 8/16 00:00:00 14kN
23.Dd5 Dxa4 24.Lxc4 Dxc2 25.Tc1 Dxb2 26.La6 b3
=/+ (-0.52) Tiefe: 9/16 00:00:00 18kN
23.Dd5 Dxa4 24.Lxc4 Dxc2 25.La6 Dxb2 26.f4 Sf6
-/+ (-0.72) Tiefe: 10/16 00:00:00 26kN
23.b3 c3 24.De3 Da6 25.Sd4 d5 26.exd5 Lc5
=/+ (-0.65) Tiefe: 10/20 00:00:00 199kN
23.b3 c3 24.De3 Da6 25.Sd4 Tbc8 26.Sb5 h6
=/+ (-0.44) Tiefe: 11/20 00:00:00 206kN
23.b3 c3 24.De3 Da6 25.Lg1 Lh4 26.Sd4 d5
=/+ (-0.43) Tiefe: 12/20 00:00:00 232kN
23.b3 c3 24.De3 Da6 25.h3 Tbc8 26.f4 Sg6
=/+ (-0.46) Tiefe: 13/23 00:00:00 382kN
23.Dd5 Dxd5 24.Txd5 Sf6 25.Txa5 b3 26.Lb1 bxc2
= (-0.29) Tiefe: 13/25 00:00:00 750kN
23.Dd5 Dxa4 24.Lxc4 Dxc2 25.La6 Dxb2 26.Ld4 Dc2
=/+ (-0.52) Tiefe: 14/26 00:00:01 2029kN
23.Dd5 Da6 24.Dd2 h6 25.h3 Sf6 26.Ld4 Sfd7
= (-0.23) Tiefe: 15/28 00:00:01 3482kN
23.Dd5 Da6 24.Dd2 h6 25.h3 Sf6 26.Lb1 d5
=/+ (-0.40) Tiefe: 16/28 00:00:01 4172kN
23.Dd5 Da6 24.Dd2 h6 25.h3 Sf6 26.Lb1 Dc6
=/+ (-0.44) Tiefe: 17/28 00:00:02 5539kN
23.Dd5 Da6 24.Dd2 h6 25.Lb1 Dc6 26.c3 b3
=/+ (-0.45) Tiefe: 18/32 00:00:09 32799kN
23.b3 c3 24.De3 Da6 25.Sd4 Ld8 26.Sb5 Txb5
=/+ (-0.32) Tiefe: 18/32 00:00:11 38196kN
23.b3 c3 24.De3 Da6 25.Sd4 h6 26.Sb5 Tb7
=/+ (-0.52) Tiefe: 19/32 00:00:14 48747kN
23.Sd4 Dxa4 24.Ta1 Dd7 25.f4 b3 26.cxb3 Sd3
= (-0.04) Tiefe: 19/32 00:00:20 71296kN
23.Sd4 Dxa4 24.Ta1 Dd7 25.f4 b3 26.cxb3 Sd3
=/+ (-0.38) Tiefe: 20/35 00:00:27 98198kN
23.Sd4 Dxa4 24.Ta1 Dd7 25.f4 b3 26.cxb3 Sd3
=/+ (-0.52) Tiefe: 21/35 00:00:28 102MN
23.Ta1 Tbd8 24.Tfd1 d5 25.exd5 Dc7 26.Dd4 Lf6
=/+ (-0.32) Tiefe: 21/44 00:01:06 207MN
23.Ta1 Tbd8 24.Tfd1 d5 25.exd5 Dc7 26.Dd4 Lf8
=/+ (-0.40) Tiefe: 22/44 00:01:13 229MN

Pity, here it caused an Fritz- GUI- exception, but could be restarted at once without restart of GUI, tried 4G hash instead of 16, then 2G and 1024Mb, every time in about the same depth as above freezing without GUI- crash.

Edit: so far no more crashes with number of threads reduced to 12 instead of 24, Fizbo seems to dislike hyperthreading

[peter]

Edit: so far no more crashes with number of threads reduced to 12 instead of 24, Fizbo seems to dislike hyperthreading

Or a certain degree of cpu's utilisation- capazity, a game against komodo with 12 threads each and ponder on at 200'+30", I had to restart several times to get that line:

1r2r1k1/4bppp/2qp4/p3n2n/Ppp1P3/5P2/BPPQNBPP/3R1R1K w - - 0 23

23. f4 Ng4 24. Nd4 Qc5 25. Qe2 Nxf2+ 26. Qxf2 Bf6 27. Nb5 Qxf2 28. Rxf2 Rxe4 29. Nxd6 Rxf4 30. Rxf4 Nxf4

Fizbo shows his ability to keep White's chances alive against komodo, its eval changes from the Black to the White side, komodo's ones stay rather constant favoring White a little bit with slight decrease.


Here are engines' outputs at the position after move 30 with 8G hash full each:

[d]1r4k1/5ppp/3N1b2/p7/Ppp2n2/8/BPP3PP/3R3K w - - 0 1

Analysis by Fizbo 1.7:

31.Sxc4 Td8 32.Txd8+ Lxd8 33.Kg1 Lc7 34.Kf2 g5
+/= (0.51) Tiefe: 29/39 00:00:10 41123kN


1r4k1/5ppp/3N1b2/p7/Ppp2n2/8/BPP3PP/3R3K w - - 0 1

Analysis by Komodo 9.42 64-bit:

31.Nxc4 Ne6 32.Rd5 Bd8 33.Bb3 Ra8 34.Nd6 Bc7 35.Nb5 Bb6 36.Rd6 Rd8 37.Bxe6 fxe6 38.g3 Rxd6 39.Nxd6 Bc7 40.Nc4 Kf7 41.Kg2 Kf6 42.Kf3 Kf5 43.h3 h5 44.c3 bxc3 45.bxc3 Kg5 46.Ke4 Kf6 47.g4 hxg4 48.hxg4 Ke7 49.Kd4 Kf6 50.Kc5 Kg5 51.Kc6 Bd8 52.Kd7 Bf6
= (0.27) Depth: 38 00:02:35 2164MN

[gordonr]

@Peter, I agree that your "26.Nb5 instead of g4 and in one game SF- komodo" looks like an improvement for White in that main line.

Since Ra1, as a first move, looks to be more likely a draw than a win, I guess lines like f4 (Fonzy), Ng4 or Qd5 (Allan), Qxa4 aren't worse in terms of not losing either. b3 is playing for a draw at best and a difficult draw at that. This is still an interesting position even if not ideal as a typical test position.

Thanks everyone for their thoughts.

Gordon