Enjoy!>

click for larger view

Analysis by Houdini 4 Pro w32: d 39 done

Again Berlin Defence

[Event "2nd Sinquefield Cup 2014"]
[Site "Saint Louis"]
[Date "2014.09.06"]
[Round "10"]
[White "Topalov, Veselin"]
[Black "Carlsen, Magnus"]
[Result "1/2-1/2"]
[ECO "C67"]
[WhiteElo "2772"]
[BlackElo "2877"]
[Annotator "Robot 3"]
[PlyCount "38"]
[EventDate "2014.??.??"]
[EventCountry "USA"]
[TimeControl "40/5400+30:1800+30"]

1. e4 e5 2. Nf3 Nc6 3. Bb5 Nf6 4. O-O Nxe4 5. d4 Nd6 6. Bxc6 dxc6 7. dxe5 Nf5 8. Qxd8+ Kxd8 9. h3 Ke8 10. Nc3 b6 11. Bf4 Bb4 12. Ne4 Ba6 13. Rfc1 Be2 14. Nfd2 Bxd2 15. Nxd2 Nd4 16. Be3 Nf5 17. Bf4 Nd4 18. Be3 Nf5 19. Bf4 Nd4 1/2-1/2>

Surprise, surprise, 4.0-0 led to a draw in less than 20 moves.

Any questions?

I have had little time to spend on this game but I have been posting Stockfish analysis at the AT and trying to read most of the analysis. I have not plumbed the depths of the Hocus Pocus line but I do have questions.

Assuming 17...Bxd4 18. Bg5 Nf5 19. Nxg6 fxg6, we end up here:

http://www.jepflast.com/chesstree/n...

The hocus pocus line depends on us choosing 20. Rxe4. However, both your Houdini analysis and my Stockfish analysis shows 20. g4 to be better. Stockfish even likes 20. e6 better than 20. Rxe4.

I understand that the hocus pocus proponents claim that following the 20. Rxe4 path will / may lead us to a position where we obtain a game winning advantage. Here are my questions:

1) How many <other> reasonable choices does GMARK have in this line along the way that would avoid our hoped for advantageous position? For example, if there are three points along the way where he has two equally plausible moves, then the chances of our reaching the desired position are no better than 1 in 8.

2) How exhaustively have the 20. g4 lines been researched? Are we convinced that we have adequately explored these lines and that all reasonable continuations lead to a dead draw?

That is after 18...Nc6, not 18...Nf5.

I think this is only because 14.e5 was played instead of 14.exd5. After 12... Bg6 we had better evals, which carried into the 14.exd5 lines. 14.e5 lowered our evals and made 12... Bg6 more viable for Black.

Of course, all this is very debatable and should be left to the post-mortem.

You or anyone else is justified to be annoyed with flippant move suggestions. But don't you think that it is a little bit harsh to suggest that those who vote for these absurd moves be sent to the border of Syria and Iraq? Whatever happened to "let the punishment fit the crime?

After all, this is just a game, and it is supposed to be fun. Some people find these flippant move suggestions to be their idea of "fun". I personally don't, but as long as it is not overdone and doesn't interfere with the game I am more than willing to tolerate them. And I hope that others will tolerate my weird attempts to have "fun".

Now, I'm sure that you know by now that I can't resist such an "interesting" question. So, first, let me show the final position from Carlsen vs Aronian, 2014 after 48.Kxc3.

click for larger view

Three pawns up! How many of us would like to be in that situation against Aronian! But the FinalGen tablebase generator indicates that this is a draw after 48...Kg8, 48...Kh8, or 48...Rf5. White wins after all other Black moves.

So here is how 6 engines evaluated the position, with and without 5-piece tablebases, with the tablebase used in parenthesis:

Engine Depth With_TB Without_TB

Critter 1.6a (Gaviota) 39 [+3.83] [+3.82]

Hiarcs 14 (Nalimov) 40 [+4.72] [+4.95]

Houdini (Gaviota) 35 [+3.60] [+5.04]

Komodo 7a (Syzygy) 31 [+3.17] [+3.12]

Rybka 4.1 (Nalimov) 37 [+4.73] [+4.61]

Stockfish 5 (Syzygy) 60 [+5.19] [+5.04]

Except for Houdini, all the engines' evaluations were similar, but nowhere near the correct evaluation, [0.00]. So it doesn't seem like using tablebases improves the accuracy of the engines' evaluation, except perhaps when the position reached has the same number of pieces as the tablebase used. This came as a surprise to me although I had seen similar behavior when comparing analyses done by Stockfish with and without tablebases. I foolishly thought that it was just an aberration; I should know better.

So then the next question is how much, if at all, does tablebase probing slow down the engine? Here are the comparable times for each engine to reach the indicated search depth:

Engine Depth With_TB Without_TB

Critter 1.6a (Gaviota) 39 00:16:46 00:30:41

Hiarcs 14 (Nalimov) 40 00:08:00 00:07:43

Houdini (Gaviota) 35 00:05:03 00:07:27

Komodo 7a (Syzygy) 31 00:22:10 00:54:53

Rybka 4.1 (Nalimov) 37 01:33:13 01:43:49

Stockfish 5 (Syzygy) 60 00:06:14 00:09:31

Five of the 6 engines reached the indicated depth <faster> with the tablebase enabled than with the tablebase disabled, and Hiarcs' elapsed times with and without tablebases were very close. Since tablebase probing is supposed to be slower than node evaluation, this was surprising to me. All I can think of is that having tablebase information available allows for more aggressive search tree pruning, so that the reduced search tree searching more than compensates for the additional time required to probe the tablebases. If anyone has any other ideas, please let me know.

click for larger view

Analysis by Houdini 4 Pro w32:

<The hocus pocus line depends on us choosing 20. Rxe4. However, both your Houdini analysis and my Stockfish analysis shows 20. g4 to be better>