Engine-assisted game vs Stockfish, move 36 discussion. Drawn game?

[MooMoo2]

The most popular move on the move 35 poll was ...Rd2. Stockfish played 36. g3, and the FEN is:
5r1k/6bp/8/1B3p2/4pP2/1P4P1/3r3P/2R2R1K b - - 0 36

Game so far: 1.e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 Nc6 6. Ndb5 d6 7. Bf4 e5 8. Bg5 a6 9. Na3 b5 10. Bxf6 gxf6 11. Nd5 f5 12. c3 Bg7 13. Bd3 Be6 14. Nc2 O-O 15. O-O fxe4 16. Bxe4 Rc8 17. a4 f5 18. axb5 axb5 19. Bf3 e4 20. Be2 Kh8 21. Qd2 Be5 22. f4 Bxd5 23. Qxd5 Qb6+ 24. Kh1 Bg7 25. Rad1 b4 26. Nxb4 Nxb4 27. cxb4 Qxb4 28. b3 Rc2 29. Bc4 Qc5 30. Qe6 Bd4 31. Rc1 Rd2 32. Bd5 Qb5 33. Qxd6 Bg7 34. Bc4 Rxd6 35. Bxb5 Rd2 36. g3

Depth: 60 ply. Nodes searched: 1,566,256,298,320 (~1566 G)

Stockfish is a pawn up but willing to accept a draw. Does anyone want to take it? If not, both me and Stockfish will be happy to play on

[gd_barnes]

A draw sounds good to me.

[WMHalsdorf]

This is the draw line, so a draw is fine with me.

[richs]

Let's move on by accepting the draw.

[MooMoo2]

A draw it is. Thanks for playing, everyone!

Here's Stockfish's analysis:

Move No.	Depth (ply)	  Nodes Searched	Eval	Next Move Expected
1		46		  604,794,005,386	0.18	e6
2		46		  688,388,460,608	0.22	Nc6
3		47		  984,675,908,000	0.31	cxd4
4		48		1,336,079,061,575	0.32	Nc6
5		46		  804,034,774,000	0.38	Nc6
6		46		  583,679,655,843	0.33	Bb4
7		51		  868,975,245,726	0.31	e5
8		51		  691,518,869,565	0.34	a6
9		49		  538,734,199,621	0.36	Be6
10		51		2,954,025,368,677	0.25	gxf6
11		51		1,059,887,583,011	0.33	f5
12		49		1,218,382,516,080	0.33	Bg7
13		49		2,384,023,466,000	0.32	Be6
14		49		1,992,315,367,000	0.16	O-O
15		52		3,170,038,679,293	0.07	fxe4
16		50		  703,385,374,994	0.11	Rc8
17		51		1,951,730,222,231	0.00	f5
18		53		  707,934,602,304	0.00	axb5
19		56		2,187,604,159,550	0.00	Kh8
20		52		1,008,111,836,741	0.00	Kh8
21		51		1,033,580,272,025	0.07	Be5
22		53		1,061,117,102,247	0.00	Bxd5
23		57		  917,798,447,531	0.00	Qb6+
24		56		  575,755,189,305	0.00	Bg7
25		59		2,017,940,068,015	0.00	b4
26		60		1,604,015,294,210	0.00	Nxb4
27		61		  952,704,596,348	0.00	Qxb4
28		58		1,046,398,035,315	0.00	Qc5
29		58		  699,065,743,522	0.00	Qc5
30		60		  884,981,046,826	0.00	Bd4
31		63		2,815,167,346,378	0.00	Rd2
32		60		  815,743,220,395	0.00	Qb5
33		61		1,138,146,215,505	0.00	Bg7
34		61		2,360,445,914,151	0.00	Rxd6
35		59		1,277,115,230,510	0.00	Rd2
36		60		1,566,256,298,320	0.00	Rb8

[LaurV]

Quite interesting that it clearly went for a draw at move 17, and he saw a draw at that point. We were arguing similar in the thread.

This needs to be put on the board and looked to it deeper... maybe this (long) weekend (we are free for Labor Day)....

[science_man_88]

Rb8 ( skewering the bishop) or Bb2 attacking the rook are the only goodish moves I see.

[LaurV]

That is the expected move.. (see the table). If you want to win, you have to surprize him...

[science_man_88]

Originally Posted by LaurV

That is the expected move.. (see the table). If you want to win, you have to surprize him...

yeah but we only have 33 moves to choose from and they only have 29 if I counted correctly so in theory there's still 33*29 = 990-33 = 957 games possible with just us making a move and them making one. most are pointless though.

[MooMoo2]

Originally Posted by LaurV

Quite interesting that it clearly went for a draw at move 17, and he saw a draw at that point. We were arguing similar in the thread.

This needs to be put on the board and looked to it deeper... maybe this (long) weekend (we are free for Labor Day)....

Here's the discussion: http://www.mersenneforum.org/showthread.php?t=21810

I also let Komodo 10 run on this move...to 37 ply. As stated by DH1, it clearly likes 16...Rc8.

The best part about Rc8 is that the analysis kept showing as trending better and better for black as the plies got deeper. At 37 ply it was at -.07. It was ranging between -.10 and -.15 between plies 30 and 35 and DH1 says it is .00 at ply 40. The question is has it already found a drawish line for the game? In other words will the analysis ever be better for us than .00 as the plies get deeper?

If we are going to have a chance to win this game we will need to find moves that trend better for our side as the plies get deeper even if the moves in question are not rated as the best ones. Here the best rated move has the best trend so we have the best of both worlds. The trend is what could allow us to ultimately look deeper than is shown by the program and is one of the few things that could give humans armed with a strong engine a slight advantage over another strong engine.

I now have it looking at a move 17 response by white. It is debating amongst a4, g3, and g4 after 27 ply. It is still early in the analysis. I will post the results of its look late tonight or early Sat. This is another thing that we will have to do...begin looking at responses by white before it is his turn and considering our responses to those before it is our next turn. This will ultimately give us and our engines additional time to look at each move.

Let's see if we can rock the world and win with black!

Originally Posted by LaurV

That is the expected move.. (see the table). If you want to win, you have to surprize him...

I've attached the screenshots of Move 17 and Move 36 in case anyone's interested.

[MooMoo2]

The entire Stockfish log (for all moves) may be downloaded at: https://www.sendspace.com/file/jycr27

You'll notice that:
- The move numberings are usually correct but may be off by one or two (for example, move 13 may be in the move 12 or move 14 folder). Sorry about this, but I don't have the motivation to make sure everything's in the right folder
- There were occasions when I wasn't completely sure which moves the forum would make. In those cases, there'll be multiple moves in a single folder.