The Engine Is An Oracle

I was listening to the How to Chess podcast and heard Chris Chabris say, “The engine is not an oracle.” I agreed with almost everything else Chris said, but the more I thought about it, the more I wanted to say the engine is an oracle.

In stories, taking action based on a prophecy often leads to disaster. Voldemort tries to kill Harry Potter because of Professor Trelawney’s prophecy, but this ultimately leads to his own undoing. (If you hadn’t gotten around to reading Harry Potter, uh, spoiler alert. Voldemort loses.) The thing is, the prophecy wasn’t wrong. But Voldemort’s actions based on the prophecy were naive and ultimately self-destructive.

Likewise, the engine is almost never wrong. There are still a few kinds of positions where engines make mistakes, but these are getting rarer all the time. And of course different engines will disagree on the finer points of complex positions. But for all practical purposes in human games, the engine is right. If it says you blundered, you blundered. So the issue isn’t that the engine is wrong. It’s that in misunderstanding or misapplying what it’s saying, you’ll lead yourself into disaster.

The engine is one of those weird things that’s both overrated and underrated at the same time. Commentators are understandably annoyed with spectators who watch the games of top grandmasters with the engine on and crow over every “obvious” mistake. But I also see players who dismiss almost every engine comment on their game as an “engine move.” There seems to be a sense that Stockfish plays a completely different kind of chess that’s incomprehensible to humans.

But to repurpose a phrase originally applied to Kasparov, Stockfish’s pieces move the same as everyone else’s. It’s playing the same game we are, just better. In a sense, “engine move” is just another way of saying “the best move.” The best moves can be hard to find, but if you want to be really good at chess, there’s only so much you can compromise here. Your skill at chess is largely a function of your ability to find the best moves.

On an average day I play four blitz games and analyze them with the engine. I honestly can’t remember the last time the engine said something that I couldn’t make sense of. Usually it’s, “Yeah, I didn’t see that, but it’s obviously right.” Occasionally it’s, “Huh, I don’t get that one.” But in those cases I can almost always figure out what’s going on by playing a few moves on the board and seeing how the engine reacts.

In general, the stronger the player the easier it is to get something useful from the engine, so your mileage may vary. Having said that, I believe players at all levels can get something useful out of working with the engine. It’s one of the most powerful tools at your disposal, because used correctly, it allows you to answer your own chess questions.

Given that prompt and accurate feedback is considered an essential part of deliberate practice, and you probably can’t afford to have a human coach at your side 24-7, this is an incredibly powerful ability. In fact, I would consider using the engine effectively to be a core skill of a chess player. Yes, it also has pitfalls, and you can avoid those by refusing to use the engine altogether; but you’ll be outpaced by players who do the work of learning to use it effectively.

If you’re struggling to get started with using the engine, one tip is to focus on big swings in the evaluation. If there are a bunch of moves within 0.5 points of each other, the differences are going to be pretty subtle and not worth worrying about. Additionally, differences above a certain point don’t matter much. For example, if the position goes from +5 to +8, that’s a difference of 3, but in human terms both positions are “clearly winning.” Doesn’t matter. But if the position goes from 0 to +5 in one move, the engine is saying it went from equal to completely winning. That’s a huge swing and you should be able to figure out what’s going on. (The exact numbers may change, especially because the latest release of Stockfish fundamentally changes the evaluation standard. But you get the idea: focus on big swings.)

All of this gave me an idea for a kind of exercise. Normally any coach worth their salt would tell you to solve exercises with the engine off, and for good reason: in a game you don’t have access to the engine (beads notwithstanding) so it’s best to practice using only your own brain. But what if the goal of the exercise is to learn to use the engine effectively? You can think of it sort of like an open-book test.

So here’s the exercise. Consider the position above. First, analyze it without the engine. I’d suggest setting a timer for five minutes and do the best you can within that window. The point of the exercise isn’t to see everything, it’s to use the engine effectively, but it’s helpful to get a sense of the position on your one first. I’ve selected this position to be tricky enough that it should be hard to work out all the details on your own, but if you manage to do that, all the better. Second, analyze it with the engine on. Here’s a link to the position in Lichess. The Lichess in-browser Stockfish will work fine for this exercise. In this part of the exercise your task is to use the engine’s help to craft a coherent explanation of what is going on in the position. This should include:

• Best play for both sides and why it’s best.

• Plausible alternatives and why they’re worse than the best line.

Understanding what the engine is saying isn’t easy, but it’s worth practicing, because at the end of the day it’s one of the most powerful ways to get better at chess.