Guide · Thinking method
How to Find the Best Chess Move Yourself
An engine finds the best move by searching millions of positions. You can't, so you need a method that points your limited attention at the moves that matter. The four steps below are the ones coaches teach, and every example position on this page was checked with Stockfish.
If you want suggestions while you practice, ChessHelper (chesshelper.ai) can show moves at a strength you choose, from 800 to 3200 Elo, which keeps them close to what you could find yourself. The method here is what you do without any help, and it is the part that transfers to over-the-board play.
The method in one table
Use the four steps in order on every move that is not completely forced. With practice the first step takes seconds.
| Step | Question you ask | What it catches |
|---|---|---|
| Forcing moves | What checks, captures and threats exist, for both sides? | Tactics you can play, and tactics you are about to suffer |
| Candidates | Which two to four moves deserve real calculation? | Stops you from playing the first move you see |
| Compare | After each candidate and the best reply, who is better? | Moves that look active but lose material |
| Blunder check | After my chosen move, what is the opponent's best check, capture or threat? | The one reply you did not look at |
Start with checks, captures and threats
Forcing moves limit your opponent's choices, so they are the moves you can calculate furthest. List every check, every capture and every move that makes a direct threat, first for yourself and then for your opponent. Asking "what does my opponent want?" is the half most players skip.
The finder's Show the threat button does the opponent half for you: it gives the move to the other side and asks the engine what they would play. In your head, the same question is "if I passed, what would they do?"
Here is a position every beginner meets. After 1.e4 e5 2.Bc4 Nc6 3.Qh5, it is Black's move.
The threat question finds the answer immediately: if Black passed, White would play Qxf7, which is mate, because the queen is protected by the bishop on c4. White's queen also attacks the e5 pawn. Any good move must stop the mate first.
Natural-looking moves fail this test. 3...Nf6?, attacking the queen, is the classic mistake: 4.Qxf7# ends the game. The knight on f6 blocks nothing and does not defend f7.
Pick two to four candidate moves
A candidate is a move that does something useful in the position: it meets the threat, wins material, improves your worst piece or prepares a plan. Write them down mentally before you calculate any of them. Calculating as you go tends to lock you onto the first idea.
In the diagram above, the candidates that stop Qxf7# are 3...g6 (attacks the queen and blocks the h5-f7 diagonal), 3...Qe7 and 3...Qf6 (both defend f7 and e5), and 3...Nh6 (defends f7 with the knight). Stockfish at depth 20 rates them as follows.
| Candidate | Eval (White's side) | Comment |
|---|---|---|
| 3...g6 | −0.38 | Gains time on the queen; the bishop goes to g7 next |
| 3...Qe7 | −0.21 | Solid; guards f7 and e5 |
| 3...Qf6 | −0.20 | Also fine, though the queen can be chased later |
| 3...Nh6 | +0.40 | Stops mate but puts the knight on the rim |
| 3...Nf6? | Mate | 4.Qxf7# |
Notice how close the top three are. In real positions there is often no single "best move" by a wide margin, just a few good ones and some bad ones. Your job is to avoid the bad ones and pick a good one you understand.
Compare the candidates honestly
For each candidate, find your opponent's best reply, not the reply you hope for, and stop when the position is quiet: no more checks, captures or direct threats. Then count material and ask who has the better pieces. This is the step where a capture that looked free often turns out to cost a piece.
The Petrov Defense shows it well. After 1.e4 e5 2.Nf3 Nf6 3.Nxe5 Nxe4 4.Qe2, Black has to move.
Run step one for White: the queen on e2 attacks the knight on e4. Retreating it looks obvious, so 4...Nf6 is a natural candidate. Compare it properly, though, and you find White's reply 5.Nc6+. The knight leaves e5, which opens the e-file so the queen gives check, and the knight itself attacks Black's queen on d8. Black must deal with the check, and White takes the queen next move. Stockfish rates the position after 5.Nc6+ at about +8.5 for White.
The engine's choice is 4...Qe7, which meets the threat indirectly. If White takes with 5.Qxe4, then 5...d6 attacks the knight on e5, and the knight cannot step away because Black's queen on e7 would then capture White's queen on e4. Black wins the piece back and the position stays close to level (+0.64). The difference between the two candidates only shows up when you look one reply deeper.
Finish with a blunder check
Before you touch a piece, look at the position after your chosen move and repeat step one for your opponent: every check, every capture, every threat. This takes ten seconds and catches most of the mistakes that lose games below 1500.
The Blackburne Shilling trap is a good test. After 1.e4 e5 2.Nf3 Nc6 3.Bc4 Nd4, Black has left the e5 pawn hanging.
The four best moves for White, according to Stockfish at depth 20:
| Move | Eval | Why |
|---|---|---|
| 4.Nxd4 | +1.86 | Removes the active knight; after 4...exd4 5.O-O White is clearly better |
| 4.O-O | +1.43 | Calm development; ...Nxf3+ is answered by Qxf3 |
| 4.c3 | +1.10 | Kicks the knight |
| 4.Nxe5 | about +0.4 | Wins a pawn but allows 4...Qg5, and only accurate play keeps the balance |
4.Nxe5 is the move most players choose, and a blunder check shows why it is risky. After 4...Qg5 Black attacks both the knight on e5 and the pawn on g2. White's best answers are 5.Bxf7+ or 5.O-O. The greedy 5.Nxf7?, forking queen and rook, runs into 5...Qxg2 6.Rf1 Qxe4+ 7.Be2 Nf3#, a knight mate with White's king boxed in by its own pieces.
The lesson is not "never take pawns". It is that the check of the opponent's forcing replies has to happen after every capture, especially when an enemy piece is already near your king.
Use the finder to check your thinking, not replace it
The best way to train this method is to do it first and check it second. Set up a position from one of your finished games in the move finder, write down your candidates and your choice, and only then press Find the best move. If the engine's top move was not on your list, ask which of the four steps would have found it. If it was on your list but you rejected it, find the reply you were afraid of and see why it does not work.
Practice on fixed positions too. The practice positions page has eight with hidden solutions, and the tactical patterns guide covers the shapes that step one should make you notice.
Common mistakes with this method
Stopping at the first good move. When a capture wins material, look for a check that wins more before you play it. Strong players call this "sitting on your hands".
Calculating only your own moves. Every line you calculate must include the opponent's best reply, including replies that ignore your threat and create a bigger one.
Ending a line too early. Stop calculating only when the position is quiet. If the last move in your line is a capture, the opponent may simply recapture.
Trusting evaluation numbers over ideas. A +0.3 difference between two moves rarely matters in your games. Pick the move whose plan you understand.
Fair play. Chess.com, Lichess, FIDE and US Chess forbid outside help during games. Do the method in your head during a game, and use engines only afterwards.