Engine depth is the number of plies, half-moves, the search looked ahead before reporting its result. A depth of 20 means the engine examined sequences 20 plies long, which is ten moves by each side. Higher depth generally means a more reliable assessment, but it takes longer and the reliability does not improve in a straight line.
Depth is the dial that trades time for confidence. Understanding it prevents two opposite mistakes: trusting a shallow search as if it were final, and waiting minutes for a depth that adds nothing to the decision in front of you.
Plies, not moves
A ply is one half-move. When White plays e4 that is one ply, and Black's reply is the second. Depth is counted in plies, so a depth of 16 is eight moves by each side.
| Depth | Plies ahead | Typically enough for |
|---|---|---|
| 8 | 4 moves each | A quick sanity check on a position |
| 12 | 6 moves each | Casual review and spotting obvious errors |
| 18 | 9 moves each | Serious game review in a reasonable time |
| 24+ | 12 moves each | Sharp tactical positions and deep endgames |
| 35+ | 17 moves each | Positions where a single tempo decides everything |
Those figures describe the shape of the trade-off, not a rule. A quiet, closed position may be fully understood at depth 15 while a forcing line with sacrifices needs far more before the score stops moving.
Why depth does not scale linearly
Each extra ply multiplies the work, because every reply at ply one has its own replies at ply two, and so on. This is the reason a search that reaches depth 20 in a second can take a minute to reach depth 25, and why the last few plies are usually the least valuable per second of waiting.
Practical engines soften this by pruning: they search promising lines deeply and abandon others early. That is what makes depth 25 possible at all, and it is also why depth alone is not a complete description of the work done. Nodes searched and positions per second are the other half of the picture.
What depth cannot tell you
- That the score is correct. A deep search of the wrong position is still wrong.
- That a move is easy or hard. Depth measures the engine's work, not a human's difficulty.
- That a difference in score is meaningful. Some positions settle early and stay settled.
- That one analysis is comparable to another. A score at depth 12 and a score at depth 30 are not the same kind of claim, which is why analysis settings belong in the record.
Choosing a depth
- For a full game review where you want to move quickly, something in the mid teens finds most real mistakes.
- When examining one critical position, go deeper. The position is where the value is, and waiting is worth it.
- When the score is still visibly changing as depth rises, keep going. That instability is the signal that the position is not yet understood.
- When the score has been flat for several depths, stop. More depth is not going to change your understanding.
- If you are on a laptop on battery or a phone, lower depth is a reasonable trade. The engine runs on your own hardware, so the cost is heat and waiting rather than money.
Where the engine runs matters for how long you wait
In a browser-based analysis board the engine runs on your device, so depth 22 on a fast desktop might take seconds and the same depth on a phone might take minutes. That is a hardware difference, not a bug, and it is why a depth control belongs in front of the user rather than being fixed by the site.