For example, suppose it's game 12 of the Carlsen-Caruana match last year. Caruana has white, and he wants to win before the tiebreaks. Now if instead of Caruana playing white, Stockfish were playing, one can practically guarantee that Stockfish will win the game. Therefore, in principle, Caruana can also win the game, regardless of what opening line Carlsen chooses.

So Caruana can pick an opening by inputting what Carlsen is likely to play to Stockfish. For example, suppose Stockfish plays 1.d4 and Carlsen is likely to respond with 1...Nf6. Caruana can give that to Stockfish and see what Stockfish comes up with. Regardless of what line Carlsen chooses, Stockfish should be able to reach a line that has winning chances. In the same way, this technique would beat well-known drawing lines such as the Petroff or the Berlin. This technique would even work to achieve a win with black, since Stockfish is also very likely to beat Carlsen with black.

Questions: does this scheme work? If so, do human GMs already use it? If so, why aren't GMs aiming to win with black as well as with white?

4 Answers 4


You pretty much described modern day preparation...

This is exactly what GMs are doing, except that they also try to add some of their own creativity into the process. Nevertheless, the majority of the heavy lifting is done by engines.

However, the process doesn't come close to guaranteeing a win due to the sheer number of possibilities. For example, assume that on each of your opponent's moves, he has 3 good moves to choose from. This means:

  • After his 1st move, you have to prepare for 3 outcomes.
  • After his 2nd move, you have to prepare for 9 outcomes (since each of the 3 outcomes for his 1st move then have 3 follow up outcomes when its his turn on the 2nd move).
  • After his 3rd move, you have to prepare for 27 outcomes.
  • 4th move ==> 81 outcomes.
  • 5th move ==> 243 outcomes.
  • 6th move ==> 729 outcomes.
  • 7th move ==> 2187 outcomes.

You can see where this is going, and this is based on the generous assumption that your opponent only has 3 acceptable choices. After the early stages of the opening, this number is often much higher.

But going off 3 choices, the horrible function for the approx. number of reasonable outcomes is f(x) = 3^x, where x is number of moves your opponent has made. However, GMs often still manage to prepare for 20 moves, which means their opponent has moved 10 times. 3^10 = 59,049. So GMs are clearly very clever at predicting what's going to happen, and have excellent memories, but even this can barely get to 20 moves.

  • reminded me a certain Soviet grandmaster who had an endgame position on his board in the hotel room and his teammates were shocked to see the position next day on the board during the next round of the tournament game. i believe it was about Lev Polugaevsky Jan 19, 2019 at 7:20
  • Polugaevsky was definitely renowned for his deep preparation. Jan 19, 2019 at 17:28

Doesn’t make sense to me. Stockfish even running on a super computer will not be able refute an established opening. Furthermore, the other side can do the same preparation.

It’s not as simple as that. You can’t just memorise by heart engine lines and hope to win. You don’t have memory to hold all the possible lines.


That a human uses engine techniques to beat their opponent is practically impossible. Even if the human has an excellent understanding of all of the evaluation techniques used by the engine, the human would never be able to prune all lines as perfect as the engine.

The human is a human, while the engine is a silicon brain. Engines do win drawish openings against humans because they can punish every small inaccuracy made by the human. The human can never do this. If he/she does so, he/she will be making their own mistakes, and pressing for a win in such scenario is likely to result in a loss.

In your example, Caruana can only win the game if he can play a perfect game and penalise every mistake Carlsen makes. Otherwise, his only chance is to be the second-to-last to make a mistake, which he won't likely be able to achieve intentionally.

  1. Yes it works against unprepared opponents to get a superior position.
  2. Yes, most super-gms use engine prep for certain must win scenarios.
  3. Because both sides are using engine prep and white inherently has an advantage to being with.

In fact in the previous WCC, most advantages came from engine use exclusively.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.