For the past week I've been running the resulting position through several engines in Deep Position Analysis mode, as I didn't find Komodo 6's lines too attractive. Now I had the chance to try this in the latest Komodo (v9) and while the lines are somewhat different from those produced by v6, I remain unconvinced that 15...Qe6 is that much of an improvement. As I went through the reams of analysis I came to appreciate the power of 10.Be3 much better. Sure, Black need not lose and can get playable (from an engine's perspective) positions, but most of them look thoroughly unpleasant to me. It's clearly White's game, as he gets to call all the shots.

While the engines were chugging away I had a peek into what I consider the definitive monograph on the English Opening - V. Bagirov's "Angliyskoye Nachalo" (Fizkultura i Sport, 1989). The variation <1.c4 c5 2.Nc3 Nf6 3.Nf3 d5 4.cxd5 Nxd5> (the author calls it "Modern Variation") is given 33 pages in the book's 464-page budget (for comparison, The Reverse Dragon takes 30, the Hedgehog 28 and the Double Fianchetto 12). Note that it's almost pure variations with commentary throughout, as the monograph dates from a time before it became fashionable to copy-paste a dozen complete games from a database, add a few annotations and call it an opening book.

The 5.d4 line occupies two pages, with this game providing the main variation. Toward the end of this section Bagirov gives the moves until 22.Bb6 with the comment "and White secured a material advantage". The final paragraph reads:

<The reader may be justified in asking: what is Black supposed to do in the 5.d4 system, then? We recommend having a look at the notes to Black's seventh or, earlier still, transposing into the Gruenfeld (see the notes to Black's fifth). Or perhaps Black may devise an improvement at some other point of this variation.>

One gets a distinct impression that he's not particularly enthusiastic about Black's prospects in "this variation" as a whole, at least not after 10.Be3:

<10.Be3!

The latest word in theory. Previously, it had been thought that 10.0-0 a6 11.Bxc6+ Rxc6 12.Qe3 Bxf3 13.Qxf3 g6 14.Rd1 Qc8 15.Bg5 h6 16.Bh4 Bg7 17.Rd2 Bxb2 18.Rad1 Bf6 (but not 18...Rc1?? 19.Rxb2 Qc3 20.Rb1 and White wins) leads to a complex position where White at best has compensation for the pawn. And then Portisch took up the case and introduced a substantial correction.>

Gotta love the pre-DB days! At the time this book was published "the latest word" had been ten years old (true, there were no more GM games with 10.Be3 in the meantime). The same year saw the first significant improvement in this line introduced into master practice at a minor tournament (a Swiss Open) in Czechoslovakia: in Freisler-Vodicka the unrated Black played 10...a6 and drew his 2300+ opponent in 50-something moves http://www.365chess.com/game.php?gi...

After 14.Qd4 the book gives a diagram with the following summary:

<The last few moves were relatively forced, and there appeared a position with a perceptible advantage to White. Black hasn't started on kingside development yet, which is by the way is not a simple task due to the pressure along the long diagonal a1-h8. Moreover, White is in complete control of the d-file. True, his pawn structure is somewhat spoiled but in the current position that doesn't have substantial significance. Such a factor could tell in an endgame, but things may not get that far.>

There follows an excerpt from the Portisch-Huebner until move 22 with just two notes - Bagirov mentions the trap Black sets up with 19...Qd7 (incl. a short line) and at 15...e5 <according to S. Gligoric, 15...Qe6 offered better defensive chances> (no variations given).

He also notes that 5...e6 leads to pure Queen's Gambit after both 6.e3 and 6.e4. That, incidentally, is the move he also recommends as an alternative to Black's seventh:

┊ 7...e6 is probably stronger here, and after 7...e4 not <8...Bd7 9.Be2 Na6? 10.0-0 Rc8 11.Qe5!> [etc. as in Palatnik-Gofstein, Daugavpils 1978] but <8...Nc6 9.Bb5 Qb6!? 10.a4 Qc5!> and Black has good chances to equalize - <11.Bd2 Qxc3+ 12.Bxc3 Bd7 13.0-0 a6 14.Bc4 Rg8!> [Mikhalchishin-Georgadze, Tbilisi 1978]. However, White missed the continuation <10.Bxc6+ Qxc6 11.Qxc6+ bxc6 12.0-0 Be7 13.Be3 0-0 14.Rfc1 Bb7 15.Bc5>, which secured him a nice engame [Muresan-Savereide, Tbilisi 1982]. All this deserved being the main line, however it would have been difficult to show the progress of this system's development then. ┊

Now onto the analysis of 15...Qe6!?.

Engines used:
- Komodo 9 search depth 30 plies (K9)
- Stockfish 6 d=35 (SF6)
- Deep Shredder 12 d=23 (DS12)
- Deep Rybka 4.1 d=22 (DR41)

The last two entries have been included as they prune candidate moves much less aggressively than the current top engines and have to chew through a *lot* more positions to get a given search depth (and thus are much slower but a little more thorough). So I thought their results might be useful just in case K9 and SF6 missed something early on. DS12 is particularly noteworthy because it's the only engine of the four that can find the forced draw in this position - Stockfish (Computer) (kibitz #61)

After 15...Qe6!? White's two main continuations are 16.Rc1 and 17.Qa7. The engines also like the prophylactic 16.Kg2, but it doesn't seem as critical as the other two to me, their evals notwithstanding. Rybka holds to an opinion different from the rest and prefers 15...g5 to 15...Qe6.

click for larger view

White intends to exchange off one of Black's few queenside defenders before invading. The engines unanimously agree that allowing that is not a good idea.

16...Rd8 17.Qb4 Rd7

DS12 prefers <17...b5 18.Qa5 Rd6 19.Rfd1 Rxd1+ 20.Rxd1 Kf7 21.Rd4 g6 22.a4 Bg7 23.axb5 f5 24.Ra4 axb5 25.Qxb5 > [+0.75/23]

click for larger view

18.Rfd1 Rxd1+ 19.Rxd1 b5 20.a4 Qc4 21.Qd2 bxa4 22.Qd7+ Kf7 23.Rd6 h6 [+1.32/30]

Here SF6 prefers <24.Kg2 Kg8 25.Rc6 Qb5 > [+1.30/35]

click for larger view

while the main line (K9) goes on with

24.Rb6 Kg8 25.Kg2 Kh7 26.Qf5+ [+1.46/30]

click for larger view

In either final position (K9 and SF6) Black is temporarily a pawn up, but by the time he manages to bring his kingside pieces out the situation will be reversed. In the DS12 position that already happened.

• 15...Qe6 16.Qa7

click for larger view

The most direct approach. The engines consider it to be less dangerous than 16.Rc1, although they part company early on.

16...b5 17.Rd2

Here there is a major fork:

DS12: <17.a4 Kf7 18.axb5 axb5 19.Rd5 Qh3 20.Qd7 Qxd7 21.Rxd7 g6 22.Ra1 Bg7 23.Raa7 Rhe8 24.Rab7 f5 25.b3 > [+0.87/23]

click for larger view

SF6: <17.Rd5 f5 18.Rxf5 g6 19.Rc5 Rxc5 20.Bxc5 Bg7 21.Rd1 Bxb2 22.Qb7 Rg8 23.a4 Bf6 24.axb5 axb5 25.Qxb5+ > [+0.98/35]

click for larger view

The main line (K9) goes on with

17.Rd2 g5 18.Rfd1 Rg8 19.Rd7 Rc6 20.Bc5 g4 21.Bxe7 Bxe7 22.Rxe7+ Qxe7 23.Qa8+ Kf7 24.Qxc6 Rd8 25.Rxd8 Qxd8 26.Qc3 gxf3 > [+0.97/30]

click for larger view

Three different final positions, quite similar evals. It's remarkable how the supposedly positional Komodo is the only one to opt for a counter-attack against the white King <...g5, ...g4>. Whatever the eval, I like this last position best. There are always opportunities to harass the better-standing side in Q+P endgames.

click for larger view

A slow move advocated by Stockfish and Komodo. Its only advantages seem to be avoiding any back-rank checks (that don't turn up anyway in any of the lines examined so far) and preventing ...Qh3 (a move that Komodo is inclined to in some 16.Rc1 lines). As it doesn't threaten anything directly, Black can attend to his development.

Here the engines again diverge early and again Komodo goes for the sharper line.

◦ 16...g5 17.Rc1 Bg7 18.Rxc8+ Qxc8 19.Rd1 Qe6 20.Bxg5 Rg8 21.Bh4 Kf7 and here there's a minor fork:

22.Qd3 Bh6+ 23.Bg3 b5 24.a4 bxa4 25.Ra1 [+1.08/30]

click for larger view

or

22.Qb4 Bh6+ 23.Bg3 b5 24.b3 Bg5 25.Qc5 [1.10/30]

click for larger view

◦ 16...g6 - SF6's preference - 17.Rc1 Rxc1 18.Rxc1 Bg7 19.Qa7 f5 20.exf5 Qd7 21.Qb8+ Kf7 22.Qc7 Rd8 23.fxg6+ hxg6 24.b3 b5 25.Qxd7 Rxd7 > [0.91/35]

click for larger view

Different approaches, close evals again. Here SF6's way is the simplest. In the final SF6 position White's doubled pawns IMO nullify his material advantage.

Sorry, there's no separate opinion. A mix up occurred, I think, and the game in my test db that is tagged Rybka apparently got overwritten by some Komodo 6 analysis (both were done at 22 ply).

I do have a line DR41 produced in Infinite Analysis while pondering an alternative to 15...e5:

<15...Qe6 16.Rc1 Rd8 17.Qc4 Qxc4 18.Rxc4 g6 19.Rc7 Rd7 20.Rfc1 Bg7 21.f4 Rg8 22.Kg2 b5 23.R7c6 Rd6 24.Rxd6 > [0.59] d=25 05:05:53 1652mN

Lines produced by IA are not as reliable as those in DPA, IMO, but five hours of processing time on a Core 2 Duo 2.4 MHz is still nothing to sneeze at, heh.

After the obvious 24...exd6 with get this position:

click for larger view

So, another variety of the 16.Rc1 line. After 25.Rc6 White will be up at least a pawn, but Rybka seems to think that the kingside development was worth it.

TL;DR.

Black made several mistakes
- 10...Bxf3 (⌂ 10...a6)
- 14...f6 (⌂ 14...Qd6)
- 15...e5 (⌂ 15...Qe6)

None of them were critical, but combined together they gave White a winning game. Black's last chance was 16...Bc5, shedding the queenside pawns in order to castle ASAP. After losing the exchange his position was hopeless.

Thank you for the kind words! Yes, this gradually turned into something of a pet project over the past month, even though this game is outside my main interests (early Tal and Korchnoi, Soviet championships of the 50s and 60s and the Alekhine Defence). What sparked my interest (besides Howard's questions) was that at 10.Be3 the engines don't give White more than a tiny edge, yet Black (a world-class GM) lost this game and improvement are quite hard to find.

20.Qb8+ Bd8 21.Rc8 Ke7 22.Bb6.

Does 22...Bc7 win, as in the above-mentioned line in which White first checks with Bc5 and Black replies with Ke6, and THEN White deploys the bishop to Bb6 ?

In other words, why would White "bother" checking with Bc5+ ? Why not just move the bishop straight to Bb6 in one move, and let Black's king stay on e7 ? Does that make any difference ?

20.Qb8+ Bd8 21.Rc8 Ke7>

click for larger view

This line was included in the annotations merely to illustrate the trap, I think. Of its four moves so far, two are already inferior.

After 19...Qd7 the engines estimate White's advantage to be worth better than a whole minor piece [3.81/34 Komodo 9.1]. <20.Qb8+?!> throws half of this advantage away, while <21...Ke7?> gives White most of his advantage back. The engines recommend the mysterious-looking <21...a5> instead, when White has no better course of action than to revert to the game continuation (a tempo down now) with <22.Qa8> (<22.Bxa5?> allows the same <...Bc7> trick as in the main line, although here Black "only" gets a ). Then after <22...0-0 23.Bc5 Re8 24.Bb6 Bxb6 25.Rxe8+ Kf7 26.Rf8+ Kg6 27.Qe8+ Qxe8 28.Rxe8 Bd4>

click for larger view

the mystery is revealed: compared to the game continuation, White doesn't have an immediate b4 push. Black has used the extra tempo granted by <20.Qb8+> to establish a blockade on the queenside and in the center. The position is still won for White [2.18/43 Stockfish 6] but not as easily as in the game after 25...Bd4 [5.34/42 K9.1].

Now to answer your first question. <22.Bb6> (see the first diagram) is not best for the same reasons - <22.Qa8> is still the clearest way to win. However, here <22...Bc7> doesn't work and leads by force to a White advantage that is even more overwhelming than what Portisch got in the game:

<20.Qb8+ Bd8 21.Rc8 Ke7 22.Bb6 Bc7 23.Bc5+ Ke6 <<not 23...Kf7, of course, as after 24.Rxc7 Rxb8 25.Rxd7+ White is a clean piece up>> 24.Qxb7!>

click for larger view

The white rook is taboo:

a) <24...Qxc8?? 25.Qd5#> b) <24...Rxc8?? 25.Qb3+ Qd5▢ 25.Qxd5#>

so Black has no choice but to go into a clearly lost ending:

<24...Qd1+ 25.Kg2 Rxc8 26.Qxc8+ Qd7 27.Qxa6+ Kf7 28.a4> [9.39/36 SF 25015, that's more than an extra queen]. White will have no trouble exchanging the rest of the pieces off and promoting his connected passed pawns on the queenside.

9...Rc8 [0.41/41]
10.Be3 [0.13/42]
10...Bxf3 [0.47/50]
11.gxf3 [0.55/48]
11...a6 [0.29/47]
12.Rd1 [0.35/47]
12...Qc7 [0.41/48]
13.Bxc6+ [0.33/47]
13...Qxc6 [0.30/43]
14.Qd4 [0.36/42]
14...f6 [0.96/44]
15.O-O [0.89/42]
15...e5 [2.27/46]
16.Qa7 [2.27/43]
16...Be7 [3.31/49]
17.Rc1 [3.24/49]

missing a pair of moves there, sorry; that should read

Korchnoi vs Ljubojevic, 1978

10.Bf4 is only a further improvement if you consider it useful to prevent Black from playing ...Qc7, otherwise there is not much difference. I suspect that the whole line with 5...cxd4 is reckoned dodgy these days.

Also note this post-Gelfand game where 10.Be3 was played:-