How to Talk to Engines
Zwischenzug is a newsletter for smart, curious chess players. To have it delivered weekly to your inbox subscribe here:
“Turn off the engine.”
This is what new chess players are often told. And yet it’s strange advice if you think about it. We have this magical technology that can instantly find the best move in any chess position. Okay, that’s not literally true, but it’s pretty close. The engines are at our beck and call 24-7, for free, to analyze our games and correct our mistakes. Are you really telling me we can’t find any way to use these things effectively?
Engines are like a mashup between two figures from Greek mythology: Cassandra and Medusa. Cassandra, you’ll recall, had the gift of prophecy but no one would believe her. Medusa turns anyone who looks at her to stone. Likewise, engines have all the answers, but (we’re told) we can’t understand them. And if we try - if we look at them - our human understanding will shrivel up and turn to stone.
One of the biggest differentiators between strong players and weaker players is knowing how to use the engine effectively. It’s not actually that hard to gain human insights from working with computers. It’s just that the standard interface for working with engines doesn’t guide us towards the most helpful patterns. This is a skill you can learn.
The key to making the engine work for you is to make it an active process where you’re asking questions. In this article I’ll talk about three of the most common questions and how to pose them to the engine so that you’ll get back an intelligible answer.
This guide will focus on Chessbase and Lichess, since they share many of the same shortcuts, but the same basic strategies can be used on chess.com or anywhere else that offers computer analysis (the keyboard shortcuts might be different).
“What would you do?”
This is one of the most basic things you can ask the engine. It can be one of the most helpful, but also one of the most dangerous. To play the best move - what the engine would do in the position - on Lichess or Chessbase you just press spacebar.
WARNING: This is the easiest feature to abuse!
The worst thing you can do with the engine is to spam spacebar to play out a long line of “best” moves. You’re unlikely to gain much understanding this way.
A better idea is to play against the engine - not for the whole game, just a few moves. Start by hiding the engine variations. On Lichess you can do this by clicking the “hamburger” icon at the bottom right of the moves window on the right, on Chessbase you can do it by dragging the moves window off the screen. With the variations hidden, make your move, wait a few seconds, and press spacebar to have the engine play its top response. Repeat this process for a few moves before checking the engine evaluation again.
Typically what you’ll find is that your position deteriorates very quickly. That’s fine, the point is not to outplay the engine, it’s to gain insight into the position. You’ll begin to see how typical human moves (aka, yours) are crushed. This may seem depressing, but keep in mind that you can use these same ideas against your unsuspecting human opponents in future games.
It’s a good idea to iterate this process several times to build up an understanding of the position. If your first moves turn out not to work, go back to the starting position and try again. You’ll get better at playing the position after seeing how your initial attempts are refuted. Keep an eye out for moves and maneuvers that occur over and over in many variations, because these are important ideas that are likely to be useful.
“What’s the threat?”
If the engine plays a move you don’t understand, it may have a hidden threat you haven’t seen. On both Lichess and Chessbase, the key for “show threat” is x. (On Lichess, you can also click the target icon in the upper right of the engine area to do the same thing.) What show threat actually does is pass the move back to the other side - in computer chess parlance, “pass a null move.”
If the engine has just played a move for White, normally of course it would be Black’s turn, but using show threat will make the engine tell you what White’s next move would be if it could move again immediately without giving Black a chance to respond. In many cases “mysterious” engine moves have a hidden threat that you can understand easily enough once you’ve see it. In others, the engine move may just be an improving move that has no particular threat.
You can tell the difference by how the evaluation reacts when you do show threat: if there was indeed a major threat that the opponent didn’t prevent, the evaluation should jump; if it was just an improving move with no particular threat, the evaluation will stay relatively stable. Either way, seeing what the engine intends to do next is often helpful for understanding the purpose behind its moves.
“What’s wrong with this?”
Depending on how you’ve adjusted the settings, the engine will usually only display 1-4 lines. These are the top lines according to its analysis. But often what seems to you the best or most natural move isn’t included in the lines the engine is showing. This probably means the move you want to play isn’t very good, but that information alone doesn’t help you much. You really want to know:
How much worse is your move than the computer’s top choices? Is it a disastrous blunder, or just another good move in a position with many equivalent moves?
What exactly is wrong with your move? Does it have a concrete refutation? Is it the beginning of a mistaken plan? Or maybe nothing much is wrong with it at all?
Getting the answers to these questions is really simple: just make your move on the board. Since the engine lines always start from the position on the board, now all the lines being displayed will start with your move. If your move was a blunder, the evaluation will immediately drop drastically. If it has a clear refutation, that will now be the engine’s top line in the new position. If on the other hand your move was fine, the evaluation will not change much.
To wrap it all up I made a video of how I would use these techniques to analyze a position from one of my games.
Of course this is only scratching the surface of what you can do with the engine, but as long as you remember to ask questions, you’ll get a lot more out of the process. Remember: you’re the boss.
If you liked this post, why not share it with a friend?