Playing Chess Against the Computer
An opponent that is always available, never tilts, and can be set to any strength. Here is how the engine decides its moves, which level to start on, and how to get real practice out of it.
Playing a computer is how most people get their practice. It is available at three in the morning, it never resigns in a huff, and it will repeat the same position as many times as you need. The trick is picking a level that actually teaches you something.
The five levels, and who each is for
| Level | Plays like | Use it if |
|---|---|---|
| 1 | Overlooks material, hangs pieces | You have just learned how the pieces move |
| 2 | Sees one move ahead, takes free material | You want to practise not blundering |
| 3 | Roughly a casual club player | You beat levels 1 and 2 comfortably |
| 4 | Calculates several moves deep | You want a real fight |
| 5 | Plays the best line it finds | You can already beat most people you know |
The temptation is to jump straight to the hardest level, lose forty games, and conclude you are bad at chess. Resist it. You learn from games you nearly win, not from games you never understood.
A rule that works. Play the level you beat roughly one game in three. Below that you are being flattened and learning nothing; above it you are winning without being tested. Move up a level when you start winning more than half.
How the engine actually works
There is no mystery to it, and understanding the mechanism makes the engine's behaviour much less frustrating.
The engine uses negamax search with alpha-beta pruning — the same family of algorithm that powered every strong chess program before neural networks arrived, and still the backbone of most engines today. Three steps:
- Generate every legal move. In a typical middlegame position there are around thirty.
- Look ahead. For each move, assume you reply as well as possible, then it replies as well as possible, and so on to a fixed depth. The difficulty setting is essentially how deep this goes.
- Score the result. Each final position is evaluated by material count plus a table of values for where each piece stands — knights are worth more in the centre than on the rim, pawns are worth more as they advance, and so on. The move leading to the best worst-case is played.
Alpha-beta pruning is the optimisation that makes this practical: once a branch is proven worse than one already examined, the engine stops looking at it. It changes nothing about the move chosen but it cuts the work enormously, which is why the reply arrives in about a second rather than a minute.
At the lower levels, controlled randomness is injected into the evaluation. That is what makes level one hang pieces — not incompetence, but deliberate noise, so a beginner has something beatable to play against.
Everything runs on your device
The engine is JavaScript running in your browser. No move is sent to a server, nothing about your game is stored, and no account exists to hold a history. Once the page has loaded, the game continues even if your connection drops.
That is also why it will never be as strong as Stockfish. A serious engine is a large native program running on hardware you do not have in a browser tab. Ours is designed to be a good practice partner that loads instantly, not to compete with a supercomputer.
How to actually improve while playing an engine
Most people play the computer badly, in the sense that they get very little out of it. Three habits change that:
Slow down deliberately. The engine replies in a second, which tempts you to reply just as fast. Do not. Take fifteen seconds before every move to ask what the engine's last move threatened. Most losses against an engine come from answering quickly rather than from not knowing enough.
Use undo as a learning tool, not a rescue. When you lose a piece, take the move back and work out what you missed. That is worth more than starting a fresh game and making the same mistake again. If you undo without diagnosing, you are just deleting evidence.
Play the same opening repeatedly. Against a human, variety keeps things interesting. Against an engine, repetition is the point — play the same first moves twenty times and you will start to understand the resulting middlegames properly. Our five openings for beginners gives you a small repertoire worth drilling.
What an engine cannot teach you
Worth being honest about. An engine gives you no experience of an opponent who is also tired, also under time pressure, and also capable of panicking. It never falls for a trap out of impatience, and it never plays a speculative sacrifice because the position feels dangerous.
Those are real parts of competitive chess. If you eventually play in a club or a tournament, expect the human experience to feel different — in some ways harder, in some ways much easier. Engine practice sharpens your tactics; it does not prepare you for someone thinking across the table.
Where to go next
If you are being beaten by the computer more than you would like, our guide to beating the computer at chess covers the specific weaknesses engines at this level have. If you would rather play a person in the same room, two-player mode on one device needs no second account.
And if your tactics are the weak point — which for most people below 1500 they are — the engine-verified puzzles are a faster way to improve than playing more games.