Move 18 discussion: Everyone vs stockfish

[richs]

Here is a deeper look at the line I advocated in post #20:

18. Rab1 Qe6
19. c3 Qf5
20. Qf2 Rae8
21. Nd4 Qg5
22. g3 Ne3
23. Rfe1 f5
24. Re2 f4
25. Kh1 Kh8
26. Rbe1 Re5
27. gxf4 Qxf4
28. Qg3 Qf6

And we have good play.

[gd_barnes]

In looking at all of the analysis here, I believe that we can conclude two things:

1. Qxb6 gives us an extremely high chance to draw against a top-rated machine.

2. Rab1 is possibly a better move against an opponent that we are equal with. If I was playing correspondence chess against an equal opponent, I would likely attempt it and hope that I could out-analyze him in the resulting complex positions.

I would like to see us have that good chance for us to draw with Qxb6 because I feel we have no chance to beat SF.

If we play Rab1, the play will be far more complex. Do we want to chance that against the most powerful chess software in the world? The question to ask ourselves is this:

Can we out-think and out-analyze a top-rated machine in very complex positions? Who on this forum has the time to attempt such a task? That is what Rab1 will likely cause us to have to do.

I am back home from my trip now and will have more time to analyze moves. I will see what more I can come up with on the Rab1 line but the lines are so involved and complex, I don't think an entire day of analyzing positions would be enough.

[LaurV]

Very well said, sir!
Additionally, we started this "draw" philosophy from the beginning, when we played the safer Nd2 instead of the more aggressive Nc3, and we continued it along the game. We simplified the board fast, last time beating that horse in f6, when we could play Qd4 and keep the advantage of the bishop, so the queen exchange is just the logical following of that plan. There may be better moves, but here we see a draw. That's it.

Ignoring the facts that went (luckily) in our advantage, like SF selecting that "boring" opening which is a "known draw", and ignoring the fact that we were allowed to use the online DB, we performed very well up to now. Therefore I consider this is already a win, 1-0 for the humans!

[gd_barnes]

With fresh eyes, I just came up with a very good line that we had all missed for Qxb6:

18. Qxb6 axb6
19. Nd4! Nxb2
20. Rfb1 Nc4
21. Rb5 Re5 (a)(a1)
22. a4 Rae8
23. Kf2
(black can make no progress)

(a)
21...Ra5
22. Rab1 Rxb5
23. Rxb5 Re1+ (b)
24. Kf2 Rd1
25. Rxd5 Rd2
26. Ke1 (c)
(the material is even and black can do nothing)

(a1)
21...Rc8
22. Kf2 Ne3
23. Re1
(black cannot move Nxc2 due to 24. Rc1 and hence can't make progress here)

(b)
23...Rd7
24. Nf5 Kf8
25. Kf2
(black is completely tied down and can do nothing)

(c) black cannot take the pawn, i.e.
26...Rxg2
27. Rd8+ Kg7
28. Nf5+ Kg6
29. Nh4+ and the rook falls

My point is that we sack the b2 pawn but black becomes completely tied down to his defense and/or just can't make progress and/or our king can approach the center much faster giving us an advantageous end game.

Please take a look at the resulting positions here. What can black do? He can make no progress. This is an easy and fairly quick draw with little stress.

One final thing:
If I have missed something with this line, we can revert back to 19. Rfd1. I kind of think that 19. Nd4 may lead to a quicker draw but we can wait to see if the queens are exchanged.

[science_man_88]

Originally Posted by LaurV

Very well said, sir!
Additionally, we started this "draw" philosophy from the beginning, when we played the safer Nd2 instead of the more aggressive Nc3, and we continued it along the game. We simplified the board fast, last time beating that horse in f6, when we could play Qd4 and keep the advantage of the bishop, so the queen exchange is just the logical following of that plan. There may be better moves, but here we see a draw. That's it.

Ignoring the facts that went (luckily) in our advantage, like SF selecting that "boring" opening which is a "known draw", and ignoring the fact that we were allowed to use the online DB, we performed very well up to now. Therefore I consider this is already a win, 1-0 for the humans!

if the computer blunders enough I can see a win potentially like I said earlier ( already voted) if not for the Nxb6 possibility and a few other things we could get the opponent down to king and pawns.

[gd_barnes]

Originally Posted by science_man_88

if the computer blunders enough I can see a win potentially like I said earlier ( already voted) if not for the Nxb6 possibility and a few other things we could get the opponent down to king and pawns.

SF won't blunder...at least not that humans will notice.

[axn]

Originally Posted by gd_barnes

SF won't blunder...at least not that humans will notice.

Keep in mind that SF doesn't have a lot of end game knowledge baked in. In TCEC, there were many instances where SF would end up in an opposite-colored bishop game with a pawn or two up, but where any human could trivially see how to hold the draw. SF will play for many moves with a +2 (or something) score until the realization finally dawns.

Anyway, don't think SF will blunder into a losing position, but still...

[gd_barnes]

Originally Posted by richs

Here is a deeper look at the line I advocated in post #20:

18. Rab1 Qe6
19. c3 Qf5
20. Qf2 Rae8
21. Nd4 Qg5
22. g3 Ne3
23. Rfe1 f5
24. Re2 f4
25. Kh1 Kh8
26. Rbe1 Re5
27. gxf4 Qxf4
28. Qg3 Qf6

And we have good play.

As I stated before I thought that 19...Qf5 was not the best move for black because of 20. Nd2. Hence it inflates the value of 18. Rab1, i.e.
20. Nd2 Nxd2 21. Qxd2 Rae8 22. Rbd1 Re2 23. Rfe1 Rxe1+ 24. Rxe1 Rxe1 25. Qxe1 with an easy draw -or- 20. Nd2 b5 21. Rbe1 Rae8 22. Rxe7 Rxe7 23.Nxc4 bxc4 24. Qd2 and neither side can easily progress.

I started looking over all of the reasonable possible responses for SF against 18. Rab1. I quickly was overwhelmed. Here are a few, some of which have already been analyzed here.

Qe6, Qd6, Rae8, Re5, Re2

Re5 and Qd6 have not been looked at and I don't feel that Qe6 has been sufficiently looked at because I have not seen a very good move 19 for black. I think we're mostly OK against all of them but...

I guess this makes my previous point. Rab1 is a decent move but it involves so much complexity at each move.

[MooMoo2]

Originally Posted by LaurV

And the second reason, well, even more subjective: I am dying with curiosity to see if SF beats back with the a2 pawn or with the horse, haha.

Originally Posted by LaurV

Edit: I see 4 votes already, now the funniest would be if we take the queen and he doesn't beat with the pawn!

Stockfish's evaluation almost immediately went to 0.00 pawns (indicating a drawn game) and wouldn't change no matter how much additional time I gave it. This was the longest time spent on any position in the game. From this point on, the evaluation remained at 0.00 until the end of game 1. Taking white's queen with the horse instead of the pawn would have resulted in a +0.72 pawn advantage for white.