Opening Acts

I’ve written before about switching from the Sicilian to the Caro-Kann. In fact, I don’t remember the Sicilian well. It’s a sharp opening, and you just sort of have to memorize the right move in many positions. The Caro-Kann is easy to play over the board or elsewhere if you’re used to playing Queen-Pawn openings. Black’s second move is practically always d5, and this line has a lot in common with the Slav.

I’ve started playing ICCF correspondence tournaments, where engines are allowed. I suspect this is just due to the inability to police well. USCF correspondence does not allow engines. Endgame tablebases are generated by engines, but they are always permitted in correspondence here and elsewhere. These are just considered known positions. You’re allowed to look them up and refer to them, even claim a favorable theoretical result once there are no more than seven pieces remaining in USCF correspondence and ICCF. But engines are not permitted. You are not allowed to discuss games in progress with anyone (fundamental rule of correspondence), but for the most part you are also not allowed to feed your position into an engine and ask it for the answer. Except in ICCF.

I was not very interested in this arrangement at first. It seems boring to just let the engine play another engine, and how does it improve my play if I’m not really playing? At this point I’m playing on mostly so I can get an actual rating and see if I want to continue. But for one thing, it makes the tournaments low-stress. How badly can things go when an engine is backing you?

And that’s the sixty-four-square question. It is entirely possible to lose Chess when playing engine vs engine, if you don’t have good insight into how to use the engine. It’s also quite possible, essentially, to lose a game within the first few moves if you think you just know the right way to play your opening without consulting the engine.

Sometimes opponents apparently aren’t using engines and lose all their games. (Word to the wise: If your opponent is allowed to use any resource, then you must absolutely make use of it.) Once my current games are finished, I expect to have 6.5/11, which given engines is a record I’ll be happy to accept. Aside from two Polish openings in two different tournaments where two different opponents were not using engines and got quickly destroyed, I’ve had two results, both Caro-Kanns (Caros-Kann)? I had a win in the Main Line and a loss in the Advance.

I play the Caro-Kann a lot, and the Advance Variation is one of my favorite lines. I smile whenever my opponent plays 1. e4 c6 2. d4 d5 3. e5. I immediately play 3...c5 knowing this gets hard for White to play, and if you’re not familiar with this line you might think, but 4. dxc5 e6? Are you joking? You moved the same Pawn twice (after White did), you gave the bloody Pawn up the second time you moved it, and while one big point of the Caro-Kann is getting the Bishop from c8 out early, now you’ve decided to block it. But this gives White two weaknesses to contend with on e5 and c5, and for a person this can be hard to play. This may be similar to the way in the Nimzo-Indian Black gives up the good Bishop on c3 immediately after developing it in order to produce Pawn weaknesses that are hard to manage. (And parenthetically regarding changes of repertoire, the Catalan is a nice alternative to the Nimzo that I’ve started playing.)

I know you’ll be shocked, but for an engine the Caro-Kann Advance line with 3...c5 is not confusing, and you’re likely to lose it without ever committing an error. For one thing, engine evaluations slide gradually over time as they evaluate to greater depth in the progress of a game. Stockfish 14 evaluates the position after 4...e6 at +0.4, but after 5. Bd3 Nd7 6. Bb5 Ne7 7. Bd4 a6 8. Ba4 Qc7 (Stockfish’s own line), the evaluation degrades to +0.5 or 0.6. After 22 moves, with dead material equality and no inaccuracy from either player, I was in a position evaluated at +1.0, which against an engine is a flat loss for Black.

So I’ve started reevaluating my repertoire against 1. e4 yet again. To 43 moves, Stockfish offers the best responses to 1. e4:

1...e6 (+0.3)
1...e5 (+0.3)
1...c5 (+0.4)

The evaluations of 1...e6 and 1...e5 are close, but 1...e6 is slightly better for Black. The move 1...c6 rarely shows up in the top three variations except at lower depth of evaluation.

So I started studying the French. And I might say it is in a way the other half of the Caro-Kann. The two openings together form a sort of Queen-Pawn complex in answer to 1. e4. Both bear resemblance at times to each other and the QGD, especially the Slav and Semi-Slav. It is very common to end up with Black Pawns on c6, d5 and e6 in all these lines. It’s largely a question of move order and your opponent’s choices.

These ICCF games also provide a fine reminder that you are always from the beginning playing for a draw. Only when your opponent fails to play the best move in the position are you playing for a win. Assume your opponent will always play perfectly and play for a draw. Assume your opponent is always an engine. Remember there is no human element in Chess and that your only task every move is to solve the puzzle and find the best move in the position, and you only lose games by failing at that task. Don’t try to do things that are novel. Don’t try to surprise or sneak up on your opponent. Your goal is to find the right move every move and, basically, play a perfect game. Anything else will win some games against humans but isn’t likely to result in consistently good games. And for better or worse, this means learning your openings.

Learning an opening doesn’t always mean memorization. I prefer the Caro-Kann to the Sicilian because the plan doesn’t vary too much from one line to another. You’re usually playing Bf5, sometimes Bg4, Nc6, Qc7 or Qb6, maybe Qa5. It’s easier for me to determine the right move when I understand the plan and don’t have to memorize. If you’re better at memorization, play the Sicilian or other sharp lines. But improvisation in Chess is unlikely to get you far.

And switching openings can be a breath of fresh air. I have switched from:

• The Sicilian to the Caro-Kann
• The Nimzo-Indian to the Catalan
• The QGD to the Slav
• The Caro-Kann to the French

Every time I switch, I at least learn something and broaden my game and understanding. It’s worth looking at other options, at any rate.

While they’re in progress, I play these correspondence tourneys in studies. With USCF I disable the engine in the study. With ICCF it gets a workout. Afterward I do server-side evaluations and eventually make the studies public.

Some relevant public studies:

Caro-Kann: https://lichess.org/study/69sXMWEM
French: https://lichess.org/study/In7z2MJC
Completed ICCF tourney: https://lichess.org/study/RMEBzFYz