Wednesday, June 17, 2015

Chess960: Playing Komodo-9 64bit

I decided to play Komodo-9 in a game of Chess960 because the consensus is that it plays natural human looking moves, which might be a nice way to practice Chess960 openings.

So I played Komodo-9 on SP456 - a nice easy number to remember!

Here is the situation at move 2:

SP456 White (Komodo-9) to play move 2

My first move as black was a very natural move that takes good control of the center. Komodo's first move f4 surprised me a little. I'm not sure if it is a good idea to commit to weakening the light squares on the kingside so early when there are better moves. f4 does threatens f5 at some point which would kick black's knight if I planted it on e6. However that is not a threat. I would instead play Nc6/Bd7 which opens up some nice lines on the light squares for my bishop against white's kingside. 

But now I really got a shock. Komodo-9 decides to play Qf2?! This cannot be a good move. There is no point playing out the queen so early while also blocking the dark square bishop in the process. Trying to defend f4/f5 so soon seems to be silly.

So I analysed what my other engine Stockfish thought of Komodo's move. Here is the result:

Stockfish at depth 22 thinks that after 2.Qf2?! Black should play ...Nd6 with a slightly better score -0.16. On the other hand, Komodo-9 at depth 18 thinks that black should play ...Nd6 but that white is slightly better at 0.14.

So after 2.Qf2?! we can see that Komodo-9 has issues playing Chess960 openings. Firstly, it is a slower engine than Stockfish when it comes to achieving the same depth. Stockfish got to depth 22 really quickly. Perhaps the depth of the calculation is really important in the Chess960 opening which favours Stockfish.

But Komodo's issue is not just depth, but also evaluation. There is no way in hell you can tell me that after 2.Qf2?! white is a bit better? Stockfish's evaluation of the position seems to me to be about right if not even a bit too conservative in it's favouring of black after 2.Qf2?! Nd6.

So I thought I would take a look at the CCRL 24-hour-a-day computer engine competition. In standard Chess Komodo appears to have the current lead. But in Chess960 the story is different. Stockfish is still the leader with a 50 point ELO rating higher than Komodo.

Well there you go. A little bit of engine trivia for you.

Enjoy 960

Tuesday, June 2, 2015

Chess960: Puzzle no.16

Dear followers of this blog, enjoy this unusual puzzle. It requires an understanding of what a pawn island in Chess is. For more information see a definition Pawn islands (wiki). Often in Chess960 openings we try to avoid creating long lasting pawn islands because the isolated pawn or pawns are a long term target of attack for our opponent to exploit. There are exceptions of course!

In this position, both white and black have three pawn islands each.

White has three pawn islands (a-e),(f),(g-h).
Black has three pawn islands, (a-f), (g), (h).

Here is the puzzle,  the answer is surprising!

SP455 move 2 white to play:
If white captures the g-pawn, who will have the
greatest number of pawn islands before move 7?
White, black or equal number?
Solution given bottom left hand corner

Answer: White!
2. fxg5      white has three pawn islands, black two.
2. ...h6     no change, white=3, black=2.
3. gxh6 Qxh6 white=2, black=1
4. h3   Be5! no change
5. g4   Bh2  no change
6. Rg2  Qxh3 white=2, black=1

White has more pawn islands than black. If instead white tries 5.Nbc3 or any other move, white still has one extra pawn island than black.

Monday, May 4, 2015

Chess960: Puzzle no.15

SP535 White to play: find the winning move

Solution given bottom left corner

5.Nh6! winning the rook

Thursday, January 22, 2015

Add this Chess960 start to the list of difficult SP's black must face?

Since Chess960 was invented, SP408 has never been played in real world competition. It is a serious intelligence test for black after white plays the move 1.d4.

SP408 Black to play: How to respond to 1.d4?

The problem is that the d5 square cannot be comfortably supported in such a way that assist black's development. For example, ...d5/e6 blocks a development square for a knight. If black plays ...d5/c6, yes that frees the dark bishop, but white has natural developing moves that undermine the d5 square:

1. d4...d5
2. Nfe3 {attacking d5} c6
3. c4! {threatening to draw a pawn onto the weak d5 square}

The big question is, what does black do about supporting the weak d5 square in a way that assists development?

Well since no human has played this SP in competition, I turned to the CCRL database of computer engines instead. The stats are not promising for black with a 62% - 38% win rate for white which is pretty high. A lot of engines, in particular Houdini 4, will try to play this line:

1. d4...c4
2. d5!? {preventing black's knights from developing toward the center} c4?!

However, I found one game played by Stockfish 5 that I think is a very memorable win for black, spoiled only by Critter's bad opening idea to push the a-pawn aggessively. Stockfish bypasses the d5 problem altogether by appearing to play ultra conservatively minimising any weaknesses, then only later beginning to apply pressure.

This is a stunning opening by Stockfish, as if it were intelligent. Critter is made to look foolish I think!

SP408 Critter plays a kind of Queen's Gambit Declined Slav?
Ok here is one possible answer to the problem of 1.d4 in SP408!
3.c4...Nde6! {So black attacks the very same weak square that white attacks}

Nice idea, brilliant! What is this in theory? A counter gambit on a weak square? Ideas like this are never seen in standard chess...

Wednesday, December 17, 2014

Frankfurt City Chess960 Cup 2014: Let's follow the winner (part 1)

The Frankfurt City Chess960 Cup is possibly the biggest ongoing Chess960 event in the world since Mainz ended. It has been running since 2010 and plans continue to run the event until at least 2020. So lets support it! 

On my blog you can download all the games of the entire history of competitive Chess960 ever played since it was invented by Bobby Fischer, including this tournament.

I thought I would take a close look at the games of the winner of this years event FM Zuyev, a very strong 2400 ELO Chess960 player who has played in the cup every year since it began. He won every single game he played this time, seven out of seven! 

I apologize to all the players in advance for not covering their games in this blog, but time and energy do not permit it.

Zuyev's first game was white against a 2071 player (H.Schmitt) and began as a Queen's Gambit Declined Slav Defense (exchange variation). However this was not chess, but SP428 and there is no bishop on f1. Yes, this is a very funny version of the queen's gambit!

SP428 Queen's gambit: Zuyev plays 2.c4

Zuyev offers a pawn but there is no way to capture the pawn back if black accepts the gambit. Nc1-b3 or Nd3 are then not possible and white will be down a pawn for a very long time. There are some advantages with 2.c4 however. White can get a strong center if the gambit is accepted (except that e4 blocks the bishop on b1), and black will have to sacrifice tempo to support the pawn on c4 with b5. Perhaps Zuyev wanted a really active game against a player that is almost 400 ELO lower than him.

But black did not accept the gambit (was he showing nerves?), and the game entered the Slav defense proper. The next interesting point in the game was 4.....h6!?

Black plays 4....h6!?

Black wants to develop his light bishop to h7 and oppose white's bishop on b1. This is not a bad idea, considering that the g8-a2 diagonal is blocked by d5. Perhaps a better alternative was 4....f6, preparing e4 and liberating the g8 bishop towards the center. Note that my Stockfish 5 kibitzer wanted to play 4....e6?! for quite some time until it changed it's mind to 4....f6 at depth 22.

The next critical moment is when Zuyev plays 5.f4. This move seems dubious to me since it ruins white's pawn structure, king safety and the f4 pawn is exposed to the b8 bishop. Also, the move leaves a big hole on e4 for one of black's knights to sit. Perhaps I am wrong, so I ask the question!

Why 5.f4?

The reason white played f4, is because he wants to play Nde5 where the knight will be supported twice against Bxe5(N) which would preserve two supported pawns in the center. White figured that black is unlikely to kick Ne5 with f6 because that would block a development square for black's e8 knight. This is in fact what happened in the game.

Why 8.Nb3?

I personally do not like white's idea. I think white wants to attack g7 at some point, but black can simply insert ...b7 which completely stops the b3 knight in it's tracks. Also, 8.Nb3 blocks any progress of the queenside pawns. I think better was Ncd3, where white can always withdraw his e5 knight back to f3 if necessary. This also is compatible with the old golden rule of chess, develop the knights towards the center!

Now black takes the initiative because of the hole on e4. White is uncertain about how to develop his g1 bishop and despite being 400 ELO stronger than black, is on the back foot. He plays 9.Be3:

White's decides to develop with 9.Be3

But as the diagram shows, this gives black a very simple plan to develop his knights to f5, attacking Be3, and if the bishop withdraws to f2, black has Ne4 again attacking the bishop. Although this is not fatal, black has a clear plan.

To cool things down a bit however, if black does trade a knight for white's dark square bishop, this is not necessarily bad, since the pawn construction d4-e3-f4 closes down the play for the dark squared bishops and black is better to retain a knight in a closed position (perhaps).

Here is the next key moment where black decides to trade off his undeveloped dark squared bishop for white's actively placed knight on e5:

Black plays 14....Bb8xe5(N)

I think this was a very good decision by black. No time has been wasted moving the dark squared bishop yet, while white has expended quite a lot of effort to play the knight to e5 in a forward position. Perhaps a better way to achieve this was firstly to play 14....Nxc5(N) and then Bxe5(N). The improvement is that white will be left with three isolated pawns! There will be a pawn on c5 that cannot be supported because of a5, and isolated pawns on e5 and e2.

Instead, black (may be due to time pressure) decided to simplify down the position with ...Bxe5(N), ...Nxf2(B), ...b6 (kicking the knight), ...Rc8. The problem is that all of black's initiative is gone and the game is completely even:

Move 18 white to play

White now has connected pawns in the center, is contesting the open c-file and has pressure on f7. My Kibitzer Houdini4 and Stockfish5 evaluate the game at 0.00. Black now plays a beautiful attacking game, putting pressure on black's king safety and arrives at the next crucial milestone:

Move 23 Black to play: don't blunder!

Black has pushed the h-pawn and will try to open up white's king safety. However the attack looks blunted since Qf2/Rf1/Nf4 are strong defenders. With the attack coming to a halt, black plays the natural move 23. g5? to kick the knight and continue the pressure, but this is a key blunder. Can you see why? (Solution bottom left hand corner of this blog). White sees the blunder, and black resigns in a lost position:

Move 27 Black resigns

Black's queen will be trapped after Rf3, and in any case, she is on the wrong side of the pawn with black's queenside and d5 under-defended.

Solution to move 23 blunder:

24. h3...Qg3 {deflecting the queen from protecting the knight on f5}
25. Nxd5! {capturing a crucial pawn, undermining the protection of the f5 knight, and revealing a discovered attack on the f5 knight}
White wins a pawn and black's queen is trapped.

Saturday, November 22, 2014

Chess960: Compilation Database Update for the Frankfurt City Championships

Right at the climax of the Chess World Championship at Sochi, there is also a Chess960 tournament unfolding!

The Frankfurt City Chess960 Championships have become an annual event which I only just realised has been going since 2010. Here is the link. The good news is that it looks like it is scheduled to continue until 2020! Perhaps it is the spiritual successor for the Mainz Chess960 event that ended in 2009?

The 2014 Frankfurt City competition is still unfolding with a few rounds to go.......

In the mean time, I have updated the compilation database for Chess960 that I host on this blog for another 216 games! The additional games are from the Frankfurt Championships 2010-2013 for games where at least one of the players is 2000+ rating.

Once the dust has settled on 2014 championships, I'll add them to the database as well.

Thank you to the organizers for uploading the PGNs for all the Chess960 games!

Chess960: Two steps forward one step back at Saint Louis

Disappointing news today that the two Chess960 greats Hikaru Nakamura and Levon Aronian are back for another Chess showdown at Saint Louis Chess Club but there will be no Chess960.

The questions I ask are:
  • Why is this exhibition being hosted right at the climax of the 2014 Chess world championship at Sochi?
  • Perhaps it is because the promoters hope that the internet exposure will be larger?
  • Perhaps this exhibition is trying to display an alternative more innovative format for a future chess world championship?
  • Why no Chess960 between the two great 960 players? 

A few months ago the Saint Louis Chess hosted a Chess960 match between Nakamura and Aronian at the Sinquefield cup 2014 called "Ultimate Moves" but it was botched. There was no live stream, no commentary of the match and it was tacked onto the end of the tournament when most commentators and journalists had left the venue.

So this new Chess showdown could have fixed all that damage and promoted Chess960 with a properly organised Chess960 showdown.

But no! Really disappointing. Such an opportunity wasted.