Playing an engine is the easiest chess practice to arrange and the easiest to waste. Most people either pick a level that crushes them, which teaches nothing, or one they beat comfortably, which teaches almost as little. A bit of structure changes that.
What the difficulty levels actually do
The engine on this site searches ahead using alpha-beta with a quiescence search, meaning it will not stop calculating in the middle of a sequence of captures. The difficulty setting changes two things: how many moves ahead it looks, and how often it deliberately plays a weaker move.
| Level | Looks ahead | Behaviour |
|---|---|---|
| Beginner | 1 move | Takes free material, misses most two-move tactics, sometimes plays a random quiet move |
| Casual | 2 moves | Defends single threats, occasionally walks into a fork |
| Club | 3 moves | Sees most short tactics, rarely hangs a piece |
| Strong | 5 moves | Punishes loose pieces and calculates through captures |
The practical consequence: beginner and casual are for practising plans without being punished for every slip. Club is where tactics start to matter. Strong is where you find out whether your pieces are actually defended.
Pick the level where you win about a third of the time
That is the rough sweet spot for learning anything. Winning most games means you are not being tested; losing most means you are not getting to the part of the game you are trying to practise. If you win five in a row, go up. If you lose five in a row, go down without embarrassment.
Four exercises that beat just playing games
- The opening drill. Play the first ten moves only, then stop and assess: are all my pieces out, am I castled, is my king safe? Reset and do it again. Ten repetitions build the habit faster than ten full games.
- The endgame conversion. Set up a won ending — rook and king against king, or king and pawn — and play it from the winning side against the strong level. If you cannot convert it, you have found something worth studying.
- The blunder replay. When you drop a piece, use undo, then find the move that would have kept it. Play it and continue. You are training the specific check that failed.
- The no-undo game. Once a week, play a full game at your level with no takebacks and no rushing. This is the one that measures real progress.
Play both colours
Most improving players play White far more often and never learn to handle being on the back foot. Alternate deliberately — half your games as Black.
What an engine will not teach you
Engines do not play like humans. They rarely fall for the swindles and traps that win games between people, and they defend better than a person under pressure. If your entire practice is against a computer, you may find human opponents both easier and stranger than expected.
They also give no feedback. The engine never says why your move was bad; it just wins. That is why the replay habit matters — you have to supply the explanation yourself.
Using the move list and undo well
The move list beside the board records the game in standard notation. After a loss, read back through it and find the move where the evaluation actually turned — usually one specific capture or one missed defensive move. That single move is worth more study than the rest of the game combined.
Undo steps back past the engine's reply as well as your own, so taking back once returns you to the position you actually faced. Use it to explore alternatives, not to grind out a win you did not earn.
Two players on one device
The local two-player mode turns off the engine entirely and lets two people share a screen. The rules, notation and legality checks all still apply, which makes it a good way to teach someone — the board simply will not let an illegal move happen, so you can focus on explaining ideas rather than policing the rules.