The Woodpecker Method, and the week most people quit

The method in one paragraph

Solve a fixed set of tactical puzzles from start to finish. Then solve the same set again, trying to finish in half the time. Then again, and again — seven cycles in total, each one faster than the last. That is the whole method. It comes from Axel Smith and Hans Tikkanen's book *The Woodpecker Method* (Quality Chess, 2018), which supplies about 1,100 puzzles in three difficulty bands.

The claim is not that you learn new tactics. It is that you make known patterns automatic. On cycle one you calculate a smothered mate. By cycle five you see it and move on. The target is recognition speed, not solving ability.

Two rules matter more than the schedule. Solve the full line before you check, including the opponent's best defence, and do not move the pieces. If you guess the first move and read the answer, you are running a different experiment.

The evidence, stated honestly

The headline anecdote is Tikkanen's: he ran the training and then scored three GM norms in seven weeks in 2010. That is one player, one season, no control condition, and he was already extremely strong. It is a good story. It is not evidence in the sense that word usually means.

The underlying mechanism — repeated retrieval of the same material, spaced out over time, produces more durable recall than reading it once — is well established in learning research generally. What has not been tested, as far as I can find, is the chess-specific version: that seven halving cycles over one fixed puzzle set produces measurable rating gain compared with solving the same number of new puzzles. Nobody has run that trial properly. Anyone who tells you the effect size is guessing.

So treat it as a plausible training structure with a strong logic and weak proof, not as a protocol. That is enough reason to try it. It is not enough reason to defend it when it does not work for you.

Where your errors actually are

Before you commit fifty hours, it is worth knowing what you are aiming at. Across 166,747 engine-classified moves from 5,341 analysed games on Mated, 3.8% of moves played are outright blunders and another 4.2% are mistakes — roughly one move in twelve is a real error, before you count the 18.5% classed as inaccuracies. That is the raw material tactics training works on.

The same set shows the average move gives away 95 centipawns against the engine's best, while the median is far lower. A small number of very bad moves carry most of the damage. This is the strongest argument for the Woodpecker approach that I know of: your rating is not being held down by a slow accumulation of half-pawn drift, it is being held down by a handful of catastrophes per game. Pattern fluency attacks catastrophes.

Those same 166,747 moves also show errors clustering by phase — 9.1% of middlegame moves are a mistake or worse, against 2.3% in the endgame. Puzzle sets are overwhelmingly middlegame positions, which is at least the right target. But note what that also means: if your losses are endgame conversions, a thousand mating nets will not touch them, and you should be honest with yourself about that before starting.

The week-two drop-off

Almost everyone completes some of cycle one. Very few people start cycle two. That is the part that gets dropped, and the reasons are predictable.

The first reason is arithmetic. A 1,000-puzzle set at four minutes a puzzle is roughly 66 hours. At an hour a day that is over two months before you reach the repetition — which is the only part of the method that is actually the method. By the time you finish, you never want to see puzzle 412 again.

The second reason is that cycle two feels like cheating. You recognise the position, you play the move in six seconds, and nothing feels like work. Solving new puzzles feels productive; re-solving old ones feels like admin. So people quietly switch back to a puzzle stream and tell themselves they are still doing tactics. They are doing something, but it is not this.

The subjective feeling of "too easy" is the condition the method needs. Automaticity is supposed to feel like nothing. If cycle two is still hard, the set was too difficult, which is the other common failure and the subject of the next section.

Size the set so cycle two happens

Pick the set size backwards from the time you will actually spend. This is the only decision in the whole method that matters.

Say you have 45 minutes a day and average three minutes per puzzle. That is 15 puzzles a day. A 300-puzzle set takes 20 days for cycle one, around 10 for cycle two, five for cycle three. You reach the fourth repetition inside two months. A 1,000-puzzle set on the same budget does not reach cycle two until March.

Difficulty follows the same logic. Choose puzzles you can solve in under four minutes on the first pass with roughly 80% accuracy. If cycle one accuracy is 50%, you are studying, not drilling, and the halving schedule will not survive. Smith and Tikkanen suggest working within a band you can handle, and the practical version of that is: it should be uncomfortable, not baffling.

• 300–400 puzzles is a realistic first set for most club players.
• Time cycle one honestly, then set the cycle-two target at half that number, written down.
• Shuffle the order each cycle so you recall the motif rather than the sequence.
• If you cannot halve the time, do not extend the deadline — cut the set and restart.

What repeating puzzles will not fix

A puzzle tells you a tactic exists. Your game does not. The Woodpecker Method trains the second half of the skill — once you look, you see it — and does almost nothing for the first half, which is noticing that this is a position worth looking at. Plenty of players with sharp puzzle ratings still hang a rook on move 23 in a quiet position, because they never switched into checking mode.

That trigger is trained differently: on your own games, in the positions where you personally stopped looking. This is the gap Mated is built for — it pulls your recent Chess.com or Lichess games, runs Stockfish over every position, and rebuilds the moments you went wrong into a short daily session, so you are re-solving your own blind spots rather than someone else's puzzle book. It is a complement to a Woodpecker set, not a replacement, and if you already keep a disciplined log of your own blunders you do not need software to do it.

The other thing repetition will not fix is opening disasters and endgame technique. Tactics sets are silent on both.

How you will know whether it worked

Do not use rating over eight weeks. The noise is larger than any effect you are likely to produce, and you will read a good week as proof and a bad week as failure.

Measure two things instead. First, cycle time — the actual minutes per cycle, recorded. That tells you whether the training itself is happening. Second, your blunder rate per game from engine review over a fixed sample: count blunders across your last 30 games before you start and your next 30 after cycle four. Thirty games is still a small sample and a difference of one blunder in ten games proves nothing, but a drop from four blunders per game to two is visible above the noise.

Third and unmeasurable: notice whether you are seeing motifs sooner in real games, before you have consciously started calculating. That is the thing the method is actually trying to produce. It is subjective, it is unreliable, and it is still worth paying attention to.