ForumHelpSearchMy ProfileSite InfoGuests InfoRepertoireLinks
Over the last quarter of a century we have become accustomed to the presence of computer technology in the chess domain. ChessBase, analysis engines, e-mail chess, and online Blitz are part and parcel of the chess scene these days. A further development in the last couple of years or so is the study of top level computer vs. computer games. Even Magnus Carlsen admits he has been influenced by AlphaZero.

Download PGN of May ’20 1 d4 d5 2 c4 games

>> Previous Update >>

There are some differences in the style between the machines and us:

  1. They have an ability to uncover and then put into practise new concepts in the opening.
  2. They sacrifice material for long term benefits as the initiative seems very important in their list of priorities.
  3. They play on drawish positions a lot longer than us, as they don't tire!

With few over-the-board events going on at present I've even included four engine encounters. They can play all they like, as they aren't affected by the same viruses as us!

Chigorin Defence 3.Nf3 Bg4 4.cxd5 Bxf3 5.gxf3 Qxd5 6.e3 e5 [D07]

It doesn't get many outings at top human level, but here it was interesting to see some of the best engines play a couple of Chigorins.

In Stockfish - Lc0, the main line occurred: 3.Nf3 Bg4 4.cxd5 Bxf3 5.gxf3 Qxd5 6.e3 e5 7.Nc3 Bb4 8.Bd2 Bxc3 9.bxc3 Qd6 after which White needs to choose between 10.Rb1, and 10.Qb3 as in the featured game:

There are three approaches in reply: 10...b6 (solid but passive), 10...0-0-0 (dynamic, but perhaps not 100% correct for top engines), and 10...Nge7 (an inferior gambit, as Black soon runs out of ideas) which was the game choice. I don't think Lc0 had a good day at the office as his play was lacking in inspiration and his pawn sacrifice in the endgame was suspicious. With reverse colours, Stockfish varied with 9...exd4 10.cxd4 Nge7 and held the draw. This has theoretical relevance as well, as it shows that the 'blockading' attempt might well be a good way of making the Chigorin into a solid weapon after all.

Queen's Gambit Accepted 3.e4 e5 4.Nf3 Bb4+ 5.Nc3 [D20]

In Lc0 - Stockfish (match game 23) Black opted for quite a solid line where the main feature of the early play is the presence of asymmetric majorities. White's early f2-f3 indicates his intention to hold onto his e-pawn and not seek any early expansion. In general, he has more space to deploy his forces and create at least modest pressure due to his (small but relevant) space preponderance. Practical experience suggests that Black is not unduly pressed, but there is the feeling that White keeps his nose ahead.

Queen's Gambit Accepted 3.e4 e5 4.Nf3 exd4 5.Bxc4 Nc6 6.0-0 Be6 [D20]

In Lc0 - Stockfish (match game 149, if that makes any sense) Black chose a sharper reaction. Here there was not just a novelty, but a whole new plan introduced by Lc0.

Previously 12.Bb2 and 12.Qc2 had been played. Here, the retreat 12.Ne1 is dismissed as basically inferior by my main engine, but ultimately proved to be sound. I'm not sure how this modern generation of computers function, but the critical lines are often deep into the game (so in the old days there would have been an issue with the horizon). I'm amazed that these engines can come up with such an inspired choice. Maybe I'm being naïve, but perhaps chess is moving into a new age where some well hidden opening ideas are about to be uncovered by artificial intelligence.

QGD Semi-Tarrasch 6.e3 [D42]

This opening is not usually cutting-edge, but more an attempt to outplay the opponent from a position where both sides have lots of options. The first key choice in Balog, I - Juhasz, Kr was 9.Qe2 after which Black has to decide when to capture on d4:

The game continuation 9...Ndb4 10.Be4 b6! (a novelty, it seems, and probably better in practical terms than grabbing the pawn) turned out quite well and Juhasz certainly obtained at least equality. Later on however, somewhat passive play left White with all the chances. A curious feature is that after Black decided against grabbing White's d-pawn, White then had four or five opportunities (of which, two or three were quite strong) to snatch Black's, but never did!

Semi-Slav Latvian Bayonet 7...dxc4 [D45]

The diagram position from Komodo - Fire is an illustration of how differently computers play to us:

The g-pawn is hanging, so what did Komodo do? He played 12.Be3 inviting Fire to capture on g4 with tempo, which he promptly did! How many humans would have played like that? Surely 12.Rg1 or 12.g5 would have been our candidates moves, but not Komodo's choice!

The middegame then carried on with White a pawn down, but with the more harmonious set-up. I find it hard to judge if this is really enough compensation, but in the game just when Black seemed to be getting his act together to play for an advantage Komodo was able to bail out into a slightly inferior but drawish endgame.

Anyway, 7...dxc4 8.Bxc4 b5 9.Be2 h6! looks fine (see the game and notes), and certainly worked better than in the reverse pairing where 9...b4 10.Na4 Bb7 11.e4 caused Black some difficulties.

Semi-Slav Anti-Meran 8...dxc4 9.Bxc4 b5 10.Be2 Bb7 [D46]

Alexei Shirov is one of the World's leading experts on this particular line of the Semi-Slav and he came to the board well-armed. In Sarana, A - Shirov, A he introduced a strong novelty:

Here he played 15...Qb6+ (instead of 15...Bc5+) which hangs together nicely after 16. Kh1 Nf2+ 17.Rxf2 Qxf2 18.Nf3 Rfe8! when I can't find any advantage for White. Best then might be the odd-looking 19.Nd1!? which you can check for yourself. Even so, it might be that White should settle for 15.g3 when he is less at risk from falling for a tactical motif. Shirov rather ruined his great preparation by blundering later which put paid to any winning chances.

Classical Queen’s Gambit Declined with ...Nbd7 and ...h6 [D61]

A trend in the Classical Queen's Gambit Declined is to play with an early ...c5, especially after first kicking the bishop back with 7...h6. I've been investigating the consequences this time in a couple of encounters (one with 8.Bf4 and the other with 8.Bh4) and have reached a conclusion that Black's plan is perfectly sound. Even if we pick White's optimal sequence we often find that he obtains a minimal pull at best.

I don't rate the 8.Bf4 of Petrosyan, M - Lobanov, S because of 8...c5, as the pressure is reduced and Black happily counters in the centre. After 9.dxc5 Nxc5 White should now keep the tension when he doesn't risk being worse, for example with 10.Be2. Instead 10.cxd5 (?!) just opened up the game for his better mobilized opponent and White was soon in trouble.

In Narayanan, S - Gukesh, D White reacted to 7...h6 with the main move 8.Bh4 and then 8...c5 was Black's reaction:

Superficially, Black's problem piece on c8 would seem to be a good enough reason for White to be better (somewhere in these lines), but this is far from evident to demonstrate. In the following play, some sort of big cull in the central zone tends to occur and the bishop finds a way to get into the game, thus largely negating White's opening pull. Black might not be able to claim full equality in all lines but I can't see anywhere where he is struggling. I suppose the main problem for ordinary players is that some of these marginally worse endgames with an IQP might be a basic draw for 2700s but lesser mortals might not have the same technique. It's not all a case of being on the back foot however, as there are a number of cases (including the main game) where Black can get the better of a tense middlegame.

Till next month, Glenn Flear

>> Previous Update >>

If you have any questions, then please post a message at the 1 d4 d5 Forum, or subscribers can email