09 July 2011

Chess960 Chaos

Over on Chess960 Jungle, in a post titled Tricky Tricky Chess960, HarryO investigated a start position that I encountered in A Chess960 Catastrophe. The position, SP941 RKRBNQBN, was the start of a game at Mainz 2005, where GM Bacrot was completely lost against GM Aronian after four moves. HarryO concluded, '[Bacrot] stumbled on what I think could be amongst a few very very tricky SP's for Black to play and even worse it was against Aronian!' and 'There will be a very small number of SP's where Black's first move may have to be memorized or at least the tactical motifs will have to be memorized.'

Critics of chess960 like to say that there are some positions where the odds are stacked against Black at the beginning of the game and present this as a reason for not considering Fischer's greatest invention. Since these critics never give examples with analysis, unwitting readers might assume that they are right. Is SP941 such a position? I decided to take a closer look, subjecting it to engine analysis.

I'm not a big fan of using engines to analyze the early opening, where the best moves are often based on positional ideas like development and the center, rather than tactics. Engines also know nothing about piece harmony and coordination, a central concern in chess960 where the pieces are usually not as well coordinated as in the traditional start position, SP518 RNBQKBNR. In the case of SP941, where strong tactical motifs are present at the beginning of the game, an engine can help to cut through the tangle.

The first diagram shows the Aronian - Bacrot game after Aronian's first move. The most striking feature of the position is the RKR sitting in the corner. Castling is not going to be easy in this game and castling O-O is highly unlikely. The second feature is the weakness of the b- & d-Pawns. This was the problem that Bacrot underestimated.

To analyze the position, I used a normal engine that has no specific knowledge of chess960, i.e. without knowledge of the castling rules. Since castling does not play a role in the early moves, this is not a big disadvantage. The engine is the strongest I have, analyzing the early opening to >20-ply in only a few minutes, a feat that my chess960 engines can't do. After running the engine against the position with 1.e4, I looked at the top four moves it suggested: 1...f5, 1...f6, 1...e5, and 1...c5.

The first candidate move, 1...f5 leads to the early development of the Black Queen after 2.exf5 Qxf5. Now White can attack the Queen immediately with 3.Ng3, when 3...Qa5 calls to mind the Scandinavian Defense in SP518. Not all players are comfortable with an early deployment of the Queen and, while it looks safe on a5, its exposed position will certainly become a factor sooner or later.

Like 1...f5, the second candidate, 1...f6, opens the diagonal for the Bg8, but does little for the center. White can play an immediate 2.Bg4, hitting the weak d-Pawn, when Black responds 2...e6. This line of play looks passive and might not appeal to players who are looking for more than solidity in the opening.

The third candidate, 1...e5, was Bacrot's choice. It certainly looks natural. It uses a strategy of symmetry, which is often a good strategy for Black in the early moves of a chess960 game, and it opens diagonals for a Bishop and the Queen. The engine favors Aronian's move 2.Nd3 over all others, then suggests 2...f6 and 2...Ng6 for Black. Bacrot chose 2...Ng6, when Aronian answered 3.f4, also the heavy favorite for the engine. The two-fisted threat is 4.Nc5 & 5.Qb5, winning immediately. Instead of Bacrot's 3...Bf6, which overlooked the threat, the engine suggests 3...c5.

After the similar 2...f6 3.f4 c5, shown in the second diagram above, Black's position looks chaotic. The three Black Pawn moves aren't harmonized into any obvious plan and the e-Pawn is hanging. Although Black can recover it, the variations I looked at were all equally chaotic. It would take more analysis to determine if there is anything more to Black's game than parrying White's threats, but I stopped there.

The fourth candidate move, 1...c5 (the SP941 Sicilian?), also leads to chaotic looking positions. Its first merit is to interfere with White's Nc5, and by opening the file for a lift of the c-Rook and a diagonal for the dark squared Bishop, it heads for a fast ...O-O-O. HarryO advises to forget about it, but I think it's worth a try.

If I were faced with the position after 1.e4, I would probably play 1...e5, because I like advancing in the center. Would I see the idea with 4.Nc5 & 5.Qb5? If one of the world's top grandmasters missed it, what are the chances for us grandpatzers?


Ichabod said...

I checked it with an engine that can play FRC. Although castling didn't come up, it still had some different ideas. Up to 17 ply it preferred the 1. ... f5 line, but without the Queen recapture: 1. ... f5 2. exf4 Nd6 3. f4 Nxf5 4. Nd3 d6 5. f4 Nh4 6. Ng3 c5 7. c4 Ba5 8. Ne4 e6 9. Bc2 N8g6 10. Re1 h6. Still a very chaotic position.

But on the 18th ply it switched to 1. ... e6, which your engine didn't like: 1. ... e6 2. f4 f5 3. e5 d6 4. d4 g5 5. g3 gxf4 6. Bf3 c6 7. gxf4 Ng6 8. Nd3 Qh6 9. Be2 dxe5 10. dxe5 Bb6 11. Ng3. That gives a less chaotic position, but seems rather odd to a patzer like me with no piece moves until 6. Bf3.

HarryO said...

Mark thanks for the post. Let me know if I am shooting myself in the foot discussing these extreme cases in Chess960. Maybe I just open myself up to critics to condemning it even more!

My own gut feeling is that it is better to talk about anything to do with Chess960 rather than to suppress discussion, because after all a critic is a critic no matter whether we suppress the extremes or not. Ask yourself the question, "when have I actually fundamentally and genuinely changed my position on an issue?" It is pretty rare for any of us.

It's a bit like a democracy. In a wealthy democracy we have a small percentage of people that are homeless and on the streets. Do we want to talk about that or just suppress it? The situation is not pretty either but there is something we can do about it...

SP941 is a Picasso crazy position but it still has it's place amongst the Rembrandt's is my belief. I think you have experienced the joy of analyzing it as much as I have and so there is something in the position to give us that even if it is a bit crazy! There is something that can be done to improve the situation for black! At the same time black still has chances if white slips up just once or twice...

Chess960 is a wonderful pastime is my belief because it actually makes you think on many levels including questions about "what is actually my notion of perfection anyway?"


HarryO said...

Wow the 1...e6 line that Ichabod found does produce some interesting sequence of pawn moves. It makes me think that when the king is at the extreme edge of the board like this with the main king attack tactics temporarily suppressed, both sides can do this extensive maneuvering in the centre/opposite wing in ways you cannot do when the king is more centralised.

The problem with 1...e6 to me is not 1...e6 actually but ...f5?! because the position looks totally lost for black simply because white's light bishop is 200% better than black's light bishop which as far as I can tell has absolutely no escape path after f5 (please correct me if I'm wrong!). This is because f5 hemms black's dark bishop along the vital escape diagonal h7-b1 which is actually an attacking diagonal against the king!

On the other hand if black breaks a few opening principals and plays:
1. e4 e6
2. f4 Ng6
3. Ng3 c5!
Now that actually seems to work on many levels. Black's Ng6 can jump to e5 or it can even harass on h4 attacking f3.

Wow....this opening combines Ichabod's find with Mark's c5 proposition. At all times black does not play ...f5?! instead black now continues along the pawn framework d6 and the h6 method to release the light bishop against white's king...

So many rules are being broken but it is really interesting to know how black is actually surviving so well in this line!

Holly molly if black develops Bb6/Ne7/Qe7 they can even castle g-side! However I think that in reality black's dark bishop cannot be developed well enough to make such a reality actually happen (Bb6 is dubious) what about Bh4 after the queen is taken off e7!?


HarryO said...

SP941 is so interesting, in case anyone is interested I've done another blog on it over here:


Mark Weeks said...

Ichabod: It's not that my engine didn't like 1...e6, it's that I arbitrarily cut off the candidate moves at four. If I had stopped the analysis a ply earlier or a ply later, 1...e6 might have been in the top four. There's a lot of chance in these computer analyses, another reason why I don't rely on them too much for the early opening. I agree that the move 1...e6 is interesting.

HarryO: Any start position (including White's first move) that leads to a quick win has to be examined objectively. Ditto for any start position (including White's first move and Black's response) that looks too drawish. Is there a defect in the start position or is it poor play that skews the results?


Of course, in real competition we don't have computers to guide us, so the question becomes how do we find the way through the thicket of possible variations. The more chaotic the position looks, the more we depend on calculation and the less we depend on positional factors. Can we recognize chaos before it actually appears on the board? - Mark