I first played correspondence chess a long time ago. Maybe in the 80s or 90s. I agreed to play for Surrey in a county match. This was before the time of chess engines. The rate of play was very slow. One move at a time, sent by postcard. I still remember the game, dimly. I was White. It was a French Winawer. The only resource I had was the 1975 Batsford classic ‘The French Defence Main Line Winawer’ by John Moles, and that was written from the Black point of view. I followed theory as far as I could (move 21 or thereabouts) then had to think for myself. I was outplayed in the middlegame and lost. And the game seemed to take forever. After that I decided I would never play correspondence chess again.

Fast forward to 2022. I have returned to the correspondence world. In less than nine months I have started 120 correspondence games and finished 70 of them, with a score of +33 =33 –4, or about 71%. My opponents have been rated 1866 on average (that’s not FIDE, but ICCF, or the International Correspondence Chess Federation). I started with a default rating of 1800 and this has been rising steadily by over 100 points a quarter. I now like correspondence chess very much. How and why has this happened?

It started with a mailshot from the Surrey captain almost a year ago inviting interest in the 2022 round of county matches. He explained how modern correspondence chess worked. Instead of sending moves by post, games were played online on a server. Although the rate of play was still rather stately (10 moves in 50 days is standard), players could move as quickly as they wanted. My completed games have ranged from 1 day to 156 days in length, at an overall average of 34 days. Apart from one game (ongoing) where there is bad blood between me and my opponent, I always play very quickly. My opponents tend to play the opening at a decent pace, continuing in a timely fashion when the game is still quite balanced, but slowing down considerably when they realise they’re losing.

The more significant innovation is that the use of chess engines to analyse during the game is actively encouraged. Engines have improved tremendously over the past 20 years or so. The top engines are rated about 3500 – well above world champion human level. They play technically perfect chess. They also play beautiful chess, developing concepts that go beyond human experience. They are not afraid to enter mind-boggling complications. And finally, they have started to play like humans, incorporating intuitive factors such as the value of the initiative and prioritising piece activity over materialistic considerations. Since this is the one chess format where it is allowed, I simply could not resist enhancing my game with riches as wonderful as this.

I will have more to say about correspondence chess in later posts. For now, here is one of my first correspondence games from early in 2022.

White: Rodney Barking (1800, Surrey), Black: Ken Clow (2099, Essex).
B31: Sicilian, Rossolimo.
1.e4 c5 2.Nf3 Nc6 3.Bb5 g6 4.Bxc6 bxc6 5.0-0 Bg7 6.Re1 Rb8 7.e5
7.h3 is the theoretical move. I rather like 7.e5, cramping Black’s position and forcing him to weaken his pawn structure in order to develop.
7…f6 8.d4 cxd4 9.Bf4!

Sacrificing a pawn for active play is the computer’s top choice. I preferred this to retaining the material balance with 9.exf6 Nxf6 10.Qxd4, when Black has open lines and a dynamic position.
9…Rxb2?!
Greedy. He should have clarified the position in the centre with multiple exchanges on e5 so as to develop and castle. This was for me the first indication that my opponent was either playing without a chess engine, or with an engine that wasn’t as strong as mine. The evaluation is now +1.36 despite the temporary two-pawn deficit.
10.Qxd4 Qb6 11.Qd3! Ba6?! 12.Qa3 fxe5?!
Black is again two pawns up and his developed pieces are superficially active. But the computer soon demonstrates that White has a very strong initiative.
13.Be3 Qb7 14.Nc3 d6?! 15.Rab1 Rxb1 16.Rxb1 Qc8?!
The position looks messy, but Black is behind in development and his king is still in the centre. White’s advantage has now risen to +4.71 which tells me that I am now winning. Black cannot hold his queenside together.
17.Bxa7 Bb7?! 18.Qb3 Ba6 19.Ng5 Nh6 20.Nd5!
Not really a piece sacrifice. 20…cxd5 21.Qa4+ Kf8 22.Qxa6! wins immediately.
20…Bf6 21.Nxf6+ exf6 22.Ne4 Nf7! 23.Qa3! Ke7 24.Rb8 Qxb8! 25.Bxb8 Rxb8

Black’s queen sacrifice has produced a very interesting endgame with an unusual material balance. Material is dynamically equal: I have queen for rook, bishop and pawn. On the face of it, Black should be able to hold this position. But chess engines are very good at exploiting the limitless potential of the queen – especially in combination with a knight. The evaluation remains at between +4 and +5 for quite a long time.
26.h3! Bc8 27.Nd2! f5 28.Qa5 Bd7 29.Nc4! Rb7 30.a4 e4! 31.Kh2 f4?!
Black’s pieces are awkwardly-placed and he finds it hard to co-ordinate. In fact there is now no defence to the simple plan of advancing the white a-pawn.
32.Nb6 f3 33.gxf3 exf3 34.Qa6 Nd8 35.a5 Rc7 36.Qa8 Rb7 37.Qxb7! Nxb7
Sacrificing queen for rook to force the pawn through to promotion. The black knight on b7 is spectacularly badly placed and cannot stop the pawn. This is an elementary endgame tactic and you don’t need a chess engine to see it.
38.a6 1–0

Black resigned. On the face of it, this was a crushing victory. But the computer’s summary is interesting: “White loses 0 pawn per move (no errors), Black loses 0.17 pawn per move (8 inaccuracies).” Computer analysis shows that Black did not make any blunders or obvious errors. It shows that at the very top level, you have to play perfectly to survive.