Published 6 September 2026 · 7 min read · Playing

How to Beat the Computer at Chess

A browser engine searches a fixed depth ahead with a simple evaluation. That makes it strong tactically and weak somewhere quite specific — and knowing where is most of the battle.

Dense cluster of interlocked pawns filling the frame at a low angle

Engines do not get tired, do not get impatient, and do not miss one-move tactics. That sounds unbeatable, and at the top level it is. But the engine in a browser game is searching a fixed number of moves ahead with a simple evaluation, and that creates specific, exploitable weaknesses.

The one habit that matters most. Before every move, check what your opponent's last move threatened. Most losses against a browser engine are one-move blunders played quickly, not strategic defeats.

1. Slow down — the engine's speed is a trap

The engine replies in about a second. That sets a rhythm, and you unconsciously match it. Then you hang a knight.

There is no clock. Take fifteen seconds before every move and run the same three checks a strong player runs automatically: what is attacked that was not attacked before, what checks are available to either side, and what captures are available to either side. This alone will improve your results more than any opening knowledge.

2. Steer for closed positions

This is the biggest structural weakness of a search-based engine at shallow depth.

In an open position with loose pieces, an engine calculates its way to an advantage — that is exactly what it is good at. In a closed position, where pawn chains lock the centre and there is nothing to capture, the advantage instead comes from slow manoeuvring: repositioning a knight over four moves to reach a good square, preparing a pawn break, improving your worst piece.

Those plans take longer than the engine can see. It will shuffle, make small concessions, and gradually end up worse without ever having made an identifiable mistake.

Openings that tend to close the centre are your friends here. A locked pawn structure with pawns on d4 and e5 facing d5 and e6 is the kind of position where a patient human outplays a shallow engine.

3. Take the endgame seriously

Search-based engines are weakest in endgames, and the reason is structural. Endgames need very deep calculation — a king walk can take twelve moves to matter — and simple material evaluation misses concepts like opposition, zugzwang and the value of a passed pawn until far too late.

If you reach an endgame level with the engine, you have a real chance even if you were slightly worse earlier. That is exactly why endgame study pays: it is the phase where a modest amount of knowledge converts directly into results.

Three worth knowing cold: king and rook mate (and there are puzzles on that page to drill it), king and pawn endings where the opposition decides everything, and the basics of rook endgames, which are the most common of all.

4. Do not try to out-calculate it

The single most common losing plan is to start a sharp tactical melee because it feels like the aggressive, winning approach. You are picking a fight on the engine's home ground.

Sacrifices that "look" strong get refuted instantly. Complications where you are relying on your opponent missing something will fail, because it will not miss it. If your plan requires the engine to blunder, you do not have a plan.

5. Use its randomness against it at lower levels

At levels one and two, the engine has deliberate noise added to its evaluation — that is what makes it beatable. Practically, it will sometimes hang material for no reason.

So at those levels, the winning approach is simply do not blunder and wait. Develop sensibly, castle, keep everything defended, and take the free material when it appears. You do not need to create anything. This is also excellent practice, because "do not blunder and wait" is a genuinely strong strategy against weak human opponents too.

6. Play the same opening every game

Against a human, repeating an opening lets them prepare. Against an engine, repetition is pure benefit. Play the same first six moves twenty times and you will start to understand the resulting middlegames properly — which pieces belong where, which pawn breaks work, what the typical plans are.

Our five openings for beginners is a small repertoire worth drilling this way. The London System in particular tends to produce the semi-closed positions discussed above.

A realistic ladder

Level	What beats it
1	Not hanging pieces. Nothing else required.
2	The above, plus taking free material when offered.
3	Basic opening principles and checking threats every move.
4	Closed positions, patient manoeuvring, careful endgames.
5	All of the above played accurately for the whole game.

Move up a level when you are winning more than half your games. If you are losing nearly all of them, move down — that is not defeat, it is calibration.

The honest caveat

None of this works against a serious engine. Stockfish on decent hardware will beat every human alive, in closed positions, in endgames, and at any speed. The advice here works because a browser engine searches shallowly with a simple evaluation, and those limits are exploitable.

That is not a criticism of it. A practice partner that loads in a second and can be set to beginner strength is far more useful for improving than one that crushes you flawlessly every time. If you want to know exactly how it picks its moves, we explain the search algorithm here.

Go and try it

Questions about beating the engine

How do you beat a chess computer?

At the levels found in browser games, three things work: avoid tactical shots by checking every capture and check before you move, steer for closed positions where long-term planning matters more than calculation, and play the endgame carefully because search-based engines at shallow depth handle endgames poorly.

Why does the computer always beat me?

Usually because you are moving too fast. An engine replies in about a second, which pulls you into replying at the same speed. Most losses to a browser engine come from missing a one-move threat, not from being outplayed strategically.

What is a closed position and why does it help?

A closed position is one where locked pawn chains block the open lines, so pieces cannot fly around and immediate tactics are scarce. Engines that rely on searching a few moves ahead are weakest here, because the advantage comes from slow manoeuvring that lies beyond their horizon.

Is it cheating to use undo?

Against a computer, no — there is nobody to cheat. Undo is a learning tool. The useful habit is to take the move back and work out what you missed, rather than taking it back and immediately playing something else.

Should I play the hardest level to improve fastest?

No. You learn most from games you nearly win. If you lose every game without understanding why, you are collecting losses rather than lessons. Play the level you beat about one time in three.