Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Test Positions / WCC 2010: Anand-Topalov. Game 7
- - By Dadi Jonsson (Silver) [is] Date 2010-05-08 13:02
8/8/4q1kp/1Q4p1/2p3P1/2Pp4/5NK1/8 w - -


Most analysts seem to think that Anand missed a win in this position by not playing 42.Qa4!!

1) How does your engine evaluate this position

2) What is your own assessment? Is the position won for White after 42.Qa4?

My answers:

1) I have run Stockfish 1.7.1 and it gave the following result:

[+3.07]  d=35  42.Qa4 Qd5+ 43.Kf1 Qe6 44.Qd1 Qf6 45.Qe1 Kf7 46.Kg2 Kg7 (0:32:12) 15377463kN
[+2.34]  d=35  42.Qa4 Qc8 43.Qd1 Qe6 44.Qf3 d2 45.Nd1 Kg7 46.Kg3 Kg6 47.Kf2 Qe1+ 48.Kg2 Kg7 (0:25:35) 12207775kN
[+2.82]  d=35  42.Qa4 Qd5+ 43.Kf1 Kg7 44.Qa7+ Kh8 45.Qe3 (0:23:03) 10944246kN

I guess that most users will interpret an evaluation of +3.07 as completely won for White.

2) Although 42.Qa4 is a very nice try, I think the position is drawn, due to resources that analysts (or should I say chess engines, since so many "analysts" just run their engines and read the meter) have missed so far :)
Parent - - By turbojuice1122 (Gold) [us] Date 2010-05-08 15:14
Conceptually, this seems drawn to me looking at it now: black can shuffle the queen, and if white tries to bring the knight over to help pick off c4, black can either pick off g4 at the opportune time, march h5 to free g5, allowing either a mating attack or a perpetual.
Parent - - By turbojuice1122 (Gold) [us] Date 2010-05-08 15:15
Also, generally speaking, to put the Stockfish evaluation in terms of the other engines and something more objective, I think you generally have to divide it by 3. :-)
Parent - - By Dadi Jonsson (Silver) [is] Date 2010-05-08 17:00
OK, but many believe that Stockfish is very strong in the endgame and I didn't see a single Stockfish user saying: "Oh, it only gives a +3 so it's a clear draw" :) I'm running Shredder now. I'll post the result shortly.
Parent - - By turbojuice1122 (Gold) [us] Date 2010-05-09 02:19

> I didn't see a single Stockfish user saying: "Oh, it only gives a +3 so it's a clear draw"


Actually, I wouldn't be surprised at all to hear such a thing, and I have heard similar things before with similar evaluations in other endgame positions.
Parent - By Uly (Gold) [mx] Date 2010-05-09 04:46
I can say "Oh, it only gives a +2.5 but it's a clear draw" on a position where the only engine that knew it was a draw was Zappa Mexico.
Parent - - By Dadi Jonsson (Silver) [is] Date 2010-05-08 16:56
White's plan is reflected in the variation given by Stockfish. He wants to prevent the d3 pawn from moving forward. Then he wants to seize control of the e-file and push Black back. I don't think it is trivial to prevent this plan. There is also very little danger that Black will ever get a mating attack, but White must be careful to avoid a perpetual.

The variations given at various places on the Internet show Black ending up with serious problems. Do you have an improvement for Black?
Parent - - By turbojuice1122 (Gold) [us] Date 2010-05-09 02:20
The only possible "improvement" I have is that I don't see why black played ...Qf6 at that point, as I think that the e6-d5 shuffle can still continue, or if triangulation is a worry, go to f7 at the right time.
Parent - By Dadi Jonsson (Silver) [is] Date 2010-05-09 10:52

> I don't see why black played ...Qf6 at that point, as I think that the e6-d5 shuffle can still continue


This also allows White to implement his plan. Let's see what could happen..

42.Qa4 Qd5+ 43.Kf1 Qe6 44.Qd1 Qd5 (instead of 44...Qf6) 45.Qe1

8/8/6kp/3q2p1/2p3P1/2Pp4/5N2/4QK2 b - -


and White threatens 46.Qe4+ winning. Black's position looks hopeless.

Black needs to find something better if he wants to draw :)
Parent - By Dadi Jonsson (Silver) [is] Date 2010-05-08 18:58
Here are the results for Deep Shredder 12 using 2GB hash and tablebases (3-5 + some 6). Looks promising for White with an evaluation of +1.43:

[+1.43]  d=22  42.Qa4 h5 43.gxh5+ Kxh5 44.Qa7 Qd5+ 45.Kf1 Kg6 46.Qe3 Kf5 47.Qh3+ Kg6 48.Qg4 Kf7 49.Qe4 (0:24:00) 13569062kN
[+1.41]  d=21  42.Qa4 h5 43.gxh5+ Kxh5 44.Qa7 Qd5+ 45.Kf1 Kg6 46.Qe3 Kf5 47.Qh3+ Kg6 48.Qg4 Kf7 49.Qe4 (0:11:18) 6244942kN
[+1.37]  d=20  42.Qa4 Qd5+ 43.Kf1 Qe6 44.Qa2 h5 45.gxh5+ Kxh5 46.Qa8 Kg6 47.Qe4+ Kf7 48.Qxe6+ Kxe6 49.Ne4 Kd5 50.Nxg5 Kd6 51.Ke1 Kd5 52.Kf2 Kd6 53.Ne4+ Kc6 54.Ke3 Kd5 55.Kf4 Ke6 (0:02:51) 1261551kN
[+1.35]  d=19  42.Qa4 Qd5+ 43.Kf1 Qe6 44.Qa2 h5 45.gxh5+ Kxh5 46.Qa8 Kg6 47.Qe4+ Kf7 48.Qxe6+ Kxe6 49.Ne4 Kd5 50.Nxg5 Kd6 51.Ke1 Kd5 52.Kf2 Kd6 53.Ne4+ Kc6 54.Ke3 Kd5 55.Kf4 Ke6 (0:01:29) 542629kN
[+1.24]  d=18  42.Qa4 Qd5+ 43.Kf1 Qe6 44.Qa2 h5 45.gxh5+ Kxh5 46.Qa8 d2 47.Qf3+ g4 48.Qe2 Qxe2+ 49.Kxe2 g3 50.Nh3 Kh4 51.Ng1 Kg4 52.Kxd2 (0:01:04) 393343kN
[+0.54]  d=18  42.Qc5 d2 43.Kf3 Qf6+ 44.Ke2 Qxc3 45.Qf5+ Kg7 46.Qd7+ Kg6 47.Qe8+ Kg7 48.Qe7+ Kg6 49.Qe4+ Kg7 50.Qb7+ Kg6 51.Qc6+ (0:00:23) 168205kN
Parent - By Dadi Jonsson (Silver) [is] Date 2010-05-08 20:25
Deep Sjeng 3 using 2GB hash and tablebases (3-5 + some 6) likes 42.Qb1 (which is also an interesting move) and ends with with evaluation of +0.87 at depth 25:

[+0.87]  d=25  42.Qb1 Qe3 43.Qb8 d2 44.Qg8+ Kf6 45.Qh7 Qxf2+ 46.Kxf2 d1Q 47.Qxh6+ Kf7 48.Qh7+ Ke6 49.Qg6+ Kd7 50.Qf7+ Kd6 51.Qxc4 Qd2+ 52.Kf3 Qd1+ 53.Qe2 Qd5+ 54.Kf2 Kc7 55.Qe3 Qf7+ 56.Kg3 Qc4 57.Qa7+ Kc8 58.Qd4 Qc7+ 59.Kf3 Qb7+ 60.Qe4 Qxe4+ 61.Kxe4 (0:38:04) 22972525kN
[+0.87]  d=24  42.Qb1 Qe3 43.Qb8 d2 44.Qg8+ Kf6 45.Qh7 Qxf2+ 46.Kxf2 d1Q 47.Qxh6+ Kf7 48.Qh7+ Ke6 49.Qg6+ Kd7 50.Qf7+ Kd6 51.Qxc4 Qd2+ 52.Kf3 Qd1+ 53.Qe2 Qd5+ 54.Kg3 Kc7 55.Qe3 Kc6 56.c4 Qh1 57.Qd4 Qe1+ 58.Kg2 Qe2+ 59.Kg3 Qe1+ (0:32:26) 19570848kN
[+0.86]  d=24  42.Qb1 (0:31:32) 18979309kN
[+0.46]  d=23  42.Qb1 Qe3 43.Qb8 d2 44.Qg8+ Kf6 45.Qh7 Qxf2+ 46.Kxf2 d1Q 47.Qxh6+ Kf7 48.Qh7+ Ke6 49.Qg6+ Kd7 50.Qf7+ Kd6 51.Qxc4 Qd2+ 52.Kf3 Qd1+ 53.Qe2 Qd5+ 54.Kg3 Kc7 55.Qe3 Kc6 56.c4 Qh1 57.Qd4 Qe1+ 58.Kg2 Qe2+ 59.Kg3 Qe1+ (0:24:04) 14518502kN
[+0.48]  d=22  42.Qb1 Qf6 43.Qe1 Kg7 44.Kg1 Kf7 45.Ne4 Qe6 46.Qf2+ Kg7 47.Qf3 Qb6+ 48.Kg2 Qb2+ 49.Kh3 Qe2 50.Qf6+ Kh7 51.Qf7+ Kh8 52.Qxc4 Qf1+ (0:14:20) 8455744kN
[+0.49]  d=21  42.Qb1 Qf6 43.Qe1 Kg7 44.Kg1 Kf7 45.Ne4 Qe6 46.Qf2+ Kg7 47.Qf3 Qb6+ 48.Kg2 Qb2+ 49.Kh3 Qe2 50.Qf6+ Kh7 51.Qf7+ Kh8 52.Qxc4 Qf1+ (0:08:50) 5175353kN
[+0.41]  d=21  42.Qb1 (0:00:49) 464001kN
[+0.01]  d=21  42.Qc5 d2 43.Kf3 Qe1 44.Qb6+ Kg7 45.Qd6 Kh7 46.Qd7+ Kh8 47.Qd8+ Kg7 48.Qd6 (0:00:42) 389687kN

After forcing 42.Qa4, the evaluation increases to +1.04:

[+1.04]  d=24  42...Qc8 43.Qd1 Qe6 44.Kf1 Qf6 45.Qe1 Kf7 46.Kg1 Kf8 47.Ne4 Qe6 (0:26:48) 16439487kN
[+1.04]  d=23  42...Qc8 43.Qd1 Qe6 44.Kf1 Qf6 45.Qe1 Kf7 46.Kg1 Kf8 47.Ne4 Qe6 (0:11:44) 7296236kN
[+1.04]  d=22  42...Qc8 43.Qd1 Qe6 44.Kf1 Qf6 45.Qe1 Kf7 46.Kg1 Kf8 47.Ne4 Qe6 (0:08:09) 5051345kN
[+1.06]  d=22  42...h5 43.gxh5+ Kxh5 44.Qa7 Qd5+ 45.Kf1 Kg6 46.Qe3 Kf5 47.Qh3+ Kg6 48.Qg2 Qf7 49.Qe4+ Kg7 50.Kg2 Qc7 51.Qd4+ Kg6 52.Qd5 Qf4 53.Qe4+ Kh6 54.Qxf4 gxf4 55.Ne4 Kg7 56.Kf3 Kf7 57.Kxf4 Ke6 58.Kf3 Kd5 (0:05:40) 3504313kN
[+0.71]  d=21  42...h5 43.gxh5+ Kxh5 44.Qa7 Qd5+ 45.Kf1 Kg6 46.Qe7 Kf5 47.Qf8+ Ke6 48.Qg8+ Kd6 49.Qg6+ Kc7 50.Qe4 Kc6 51.Ke1
Parent - By Dadi Jonsson (Silver) [is] Date 2010-05-08 21:16
Naum 4.1 using 2GB hash and tablebases (3-5 + some 6) switches to 42.Qa4 at higher depths, with an evaluation over +1:

[+1.09]  d=28  42.Qa4 Qd5+ 43.Kf1 Qe6 44.Qd1 Qf6 45.Qe1 Kf7 46.Kg2 Kg7 47.Nd1 Qc6+ 48.Kg1 Qf3 49.Qe5+ Kf7 50.Qc7+ Kg6 51.Nf2 Qd5 52.Qb6+ Kg7 53.Kf1 Qe5 54.Qb7+ Kg6 55.Qe4+ Qxe4 (0:35:08) 9646226kN
[+0.64]  d=28  42.Qa4 (0:14:35) 3856348kN
[+0.63]  d=28  42.Qb1 Qe3 43.Qb8 d2 44.Qg8+ Kf6 45.Qh7 Qxf2+ 46.Kxf2 d1Q 47.Qxh6+ Kf7 48.Qh7+ Ke6 49.Qg8+ Kd7 50.Qxc4 Qd2+ 51.Kf3 Qd1+ 52.Ke3 Qc1+ 53.Kd4 Qf4+ 54.Kc5 Qd6+ 55.Kb5 Qb8+ (0:07:45) 1980511kN
[+0.63]  d=27  42.Qb1 Qe3 43.Qb8 d2 44.Qg8+ Kf6 45.Qh7 Qxf2+ 46.Kxf2 d1Q 47.Qxh6+ Kf7 48.Qh7+ Ke6 49.Qg8+ Kd7 50.Qxc4 Qd2+ 51.Kf3 Qd1+ 52.Ke3 Qc1+ 53.Kd4 Qf4+ 54.Kc5 Qd6+ 55.Kb5 (0:05:23) 1341055kN
[+0.55]  d=26  42.Qb1 Qe3 43.Qb8 d2 44.Qg8+ Kf6 45.Qh7 Qxf2+ 46.Kxf2 d1Q 47.Qxh6+ Kf7 48.Qh7+ Ke6 49.Qg8+ Kd7 50.Qxc4 Qd2+ 51.Kf3 Qd1+ 52.Ke3 Qg1+ 53.Kd2 Qh2+ 54.Qe2 Qf4+ (0:03:55) 985906kN
[+0.36]  d=26  42.Qb1 (0:02:29) 600368kN
[+0.35]  d=26  42.Qb7 d2 43.Qb1+ Kg7 44.Kf3 Qe1 45.Qb7+ Kf6 46.Qb6+ Kf7 47.Qa7+ Kg6 48.Qd7 Qc1 49.Qd6+ Kg7 50.Qd4+ Kg6 51.Qb6+ Kh7 52.Qb7+ Kg6 53.Qe4+ Kf6 54.Qd5 Kg7 (0:01:27) 315570kN
Parent - - By Mark (****) [us] Date 2010-05-09 09:34
I let IDeA run at d=15 for about 15 hours now.  After Qa4 it's showing +.48.  (With basically no interactivity yet.)
Parent - - By Dadi Jonsson (Silver) [is] Date 2010-05-09 10:54
Thanks for checking. I look forward to hearing more about your results. In any case, +0.48  matches my own evaluation of the position better than +3 ;)
Parent - - By Mark (****) [us] Date 2010-05-09 11:58
I'll let it run.  Hopefully, I'll get some time today to input some promising lines.  Should I keep it at 15 ply or increase it?

The last minimax has the eval down to +.44 (about 37,000 positions so far).
Parent - - By Dadi Jonsson (Silver) [is] Date 2010-05-09 22:59
Just keep it at 15 plies. It will be interesting to see how it turns out.
Parent - - By Mark (****) [us] Date 2010-05-10 10:01
Will do. It's at .56 this morning.
Parent - By Mark (****) [us] Date 2010-05-12 16:13 Edited 2010-05-12 16:15
With two more days running,  it hit as high as .64, but this morning was back down to .56.  I've put in all of the suggested lines in this thread (most of which were already in the automatic IDeA tree).  Doesn't look like there's anything overly interesting here...
- By Dadi Jonsson (Silver) [is] Date 2010-05-09 22:57
Here is an English translation of Shipov's comments after he saw Anand's 42.Qb7 (copied from Dana's blog). Do you see any improvements to this variation?

I would have played this way too, sheltering the king from checks. And, apparently, I was wrong! The computer demonstrates its fantastically subtle grasp of queen play, not allowing Black to advance his pawn to d2, with 42. Qa4!! Qd5+ 43. Kf1 Qe6 44. Qa2!!

8/8/4q1kp/6p1/2p3P1/2Pp4/Q4N2/5K2 b - -

(threatening to capture with the knight on d3) 44. … Qd5 (44. … Qc6 45. Qa1! Qd5 46. Qe1!) 45. Qa6+ Kg7 46. Qa7+ Kg6 47. Qe3!

8/8/6kp/3q2p1/2p3P1/2PpQ3/5N2/5K2 b - -

and Black is in a bitter situation. White is ready to activate his knight, and … d2 is refuted by Ke2!
- - By Dadi Jonsson (Silver) [is] Date 2010-05-09 23:04
Macolm Pein wrote the following about the position after 42.Qa4!!:

"Before we look at the lines let's consider what White would like to achieve. It's often a good idea in an endgame to imagine where you want your pieces and then try and find a way of getting them there. White would like to control the e file with his queen from e1,e3 or e4 because then Black would have to retreat. White also wants to prevent d3-d2 and that is the first point"

Then he gives the Shipov variation, and also the following variation:

"Finally the most obscure line

C) 42.Qa4 Qc8 43.Qd1 Qe6 44.Kf1! idea Qe1 44...Qf6 45.Qe1 Kf7 46.Kg2 Kg7 47.Nd1 Qc6+ 48.Kg1 Qb6+ 49.Nf2 Qd6 50.Qe4! and there is progress, Qb7+ and Ne4 may come."

I think that Black can play better here :)
Parent - By Dadi Jonsson (Silver) [is] Date 2010-05-11 08:46
Here is my improvement to the 42.Qa4 Qc8 variation, which leads to a draw.

Malcolm Pein gives 42.Qa4 Qc8 43.Qd1 Qe6 44.Kf1! with the idea of forcing the black queen off the e-file.

8/8/4q1kp/6p1/2p3P1/2Pp4/5N2/3Q1K2 b - -


I think that perhaps Malcolm was misguided by his general ideas about this position (see the quote in my post above). Black can just keep the queen on the e-file:

44...Qe5! (instead of Malcolm's 44...Qf6) 45.Qe1 (the idea behind 44.Kf1) 45...Kf6 46.Ne4+ (Repositions the knight before exchanging queens. 46.Qxe5 Kxe5 and the active position of the black king ensures the draw) 46...Ke6! 47.Nd2 Kd5 48.Qxe5+ Kxe5.

8/8/7p/4k1p1/2p3P1/2Pp4/3N4/5K2 w - -


49.Nxc4+ Kf4

8/8/7p/6p1/2N2kP1/2Pp4/8/5K2 w - -


I guess that by this time most chess engines will show an evaluation close to zero. Black picks up the g4-pawn and holds the draw. For example, 50.Kf2 Kxg4 51.Ne5+ Kf4 52.Nxd3+ is a tablebase draw.  Black can also draw in the Shipov variation (42.Qa4 Qd5).
Up Topic The Rybka Lounge / Test Positions / WCC 2010: Anand-Topalov. Game 7

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill