Stockfish Depth Explained: How Deep Should You Analyze?

Every analysis tool puts a depth setting in front of you, and none of them explain it. Slide it up and the analysis takes longer. Slide it down and it finishes fast. Somewhere in between sits a number that determines whether the engine's verdict on your queen sacrifice can be trusted.

The short answer: depth 12 catches the tactical mistakes that decide club games, and past depth 20 you are refining evaluations rather than discovering new ones. The longer answer explains why the same engine can call a move a mistake at one depth and the best move at another, and when that difference matters for your games.

Depth Counts Half-Moves, Not Moves

Depth is measured in plies. One ply is a single move by one side, so depth 12 means the engine has searched the game tree twelve half-moves ahead, six of your moves and six replies.

The number understates what the engine sees. Stockfish does not examine each branch to the same distance. Its search discards unpromising moves after a shallow look and follows forcing sequences well past the nominal limit. Checks, captures, and recaptures get extended, so a depth 12 search chases a forcing tactic twenty or more plies before it settles on an evaluation. The depth number describes the search's foundation, and the sharpest lines get read far beyond it.

Each depth also builds on the last. The engine searches to depth 1, uses that result to order its candidates, searches to depth 2, and climbs one ply at a time. A depth 20 analysis has already produced the depth 12 analysis on the way up. Deeper never means a different method, only a longer look.

Depth 12 Already Plays Above Any Human

Stockfish restricted to a shallow search remains stronger than any grandmaster who has ever lived. The engine evaluates positions with a neural network trained on hundreds of millions of them, so even its quick look rests on pattern recognition no human can match. By depth 12 it has found the refutation of a hung piece, a missed fork, or an unsound sacrifice many times over.

For grading your games, this covers the mistakes that decide them. Below 2000, games swing on blunders and two-to-three-move tactics, and a shallow search catches those with room to spare. A free game review at the default depth 12 assigns the same Brilliant-to-Blunder grades to the vast majority of moves that a much deeper search would. Our guide to move classifications covers what each grade means.

What Deeper Searches Add

Raise the depth and three kinds of position start to resolve.

Quiet positions. With no tactics on the board, the difference between candidate moves is a long-term plan: a knight rerouted over five moves, a pawn break prepared behind it. A shallow search sees little difference between the candidates. A deeper one sees where each plan lands.

Endgames. A won king-and-pawn endgame can require twenty accurate moves, and a shallow search reads a decisive position as a draw because the winning plan lies past its horizon. Endgame evaluations shift more between depth 12 and depth 25 than any other phase.

Sacrifices. Compensation that arrives eight moves later sits at the edge of a shallow search. This cuts both ways: shallow analysis can flag a sound sacrifice as a blunder, and it can miss the slow refutation of an unsound one. On Chess It Up, a candidate brilliant move must survive a second look at depth 18 before it keeps the label, because sacrifices are where shallow verdicts flip most often.

An evaluation that swings as depth climbs is itself information. A move rated +0.5 at depth 10 and -0.3 at depth 16 sits in a sharp position the engine needed sixteen plies to untangle. You had one minute on the clock and none of those plies. Treat swinging evaluations as a marker of the positions worth studying, and read our guide to analyzing your games for what to do once you find them.

Each Ply Costs More Than the Last

The search tree grows with every ply, and the time to search it compounds. Going from depth 12 to depth 20 multiplies the work many times over, and the moves whose grades change number a handful per game. The strength gained per added ply shrinks as depth climbs: the jump from depth 6 to 12 transforms the analysis, the jump from 18 to 24 polishes it.

Depth also interacts with where the engine runs. Chess It Up runs Stockfish in your browser, so your own hardware sets the pace, and a depth that finishes in seconds on a desktop takes longer on a phone. Premium analysis runs on our servers at depth 25 with a dedicated engine pool, which buys both the deeper search and hardware that reaches it fast.

Which Depth to Pick

Checking games for blunders (depth 8 to 12). Sorting your wins and losses by mistake count, finding the move where the game turned, reviewing a blitz session. The default depth 12 with fast results serves this well, and it is where most of your reviews should happen.

Studying a serious game (depth 16 to 20). A long rapid or classical game you want the truth about, an endgame you misplayed, an opening line you keep reaching. Slower, and the quiet-position evaluations firm up.

Settling a specific question (depth 20 to 25). Was the sacrifice sound? Is the endgame at move 40 won or drawn? Give the engine one position and let it run deep, rather than analyzing the whole game at maximum depth for the sake of two moves.

One habit beats a high slider setting: analyze more games at moderate depth instead of a few games at maximum depth. Your improvement comes from the pattern across twenty reviews, and the mistakes that repeat are visible at depth 12.

Frequently Asked Questions

Does depth 20 mean the engine sees 20 moves ahead?

It means 20 half-moves, so ten full moves, and only along the main line. Forcing sequences get searched deeper and dismissed candidates shallower. Read it as "at least ten moves ahead where it counts."

Will my accuracy change if I change the depth?

By a little. A deeper search regrades the odd move in a sharp or quiet position, which shifts accuracy a point or two. Compare scores across your own games at the same depth and the trend stays meaningful.

Why does the engine's evaluation keep changing while it runs?

You are watching the depth climb in real time. Each completed depth updates the score, and sharp positions swing until the search resolves the tactics. The number to trust is the final one at the target depth.

Is depth 25 enough to trust for correspondence or opening prep?

For grading your own play, more than enough. For deep opening novelties, correspondence players run engines for hours per position, which no game-review tool aims to replace. Depth 25 tells you the truth about the games you play; it does not exhaust chess.

See the Difference Yourself

Take one sharp game you lost and run the review at the default depth, then raise the slider and run it again. Most grades hold. The ones that change mark the positions that were beyond you at the board, and studying those two or three positions teaches more than the other thirty-five combined.