What Happens Across Five Woodpecker Cycles: Data From 1,568 Puzzle Sets
Among 795 Peck sets completing five Woodpecker cycles, median time was 40% of each set's first-cycle time and median accuracy change was +12 percentage points.
- By Peck
- 4 min read
- 1,183 words
Among the 795 Peck sets that completed cycle 5, median time per puzzle was 40% of each set's own first-cycle time, and the median accuracy change was +12 percentage points. These sets belong to 269 players. Both figures compare each set with itself.
The full study covers completed cycles on Peck up to 30 September 2026: 1,568 puzzle sets from 663 players. Fewer sets reach each later cycle, so the results describe those that continued. They show observed changes on repeated positions, not proof that repetition caused the changes or raised anyone's playing rating.
Solving time and accuracy by cycle
Raw medians for completed cycles 1–5. Each point uses a different sample, from 1,568 sets at cycle 1 to 795 at cycle 5; this is not a matched before-and-after comparison. The accuracy axis starts at 80%.
| Cycle | Players | Sets | Seconds per puzzle | Accuracy | Time vs own cycle 1 | Accuracy vs own cycle 1 |
|---|---|---|---|---|---|---|
| 1 | 663 | 1,568 | 17.8 | 85.0% | 100% | — |
| 2 | 461 | 1,182 | 12.9 | 91.0% | 76% | +4.0 pts |
| 3 | 361 | 992 | 9.5 | 94.0% | 61% | +7.3 pts |
| 4 | 300 | 855 | 7.5 | 96.3% | 48% | +10.0 pts |
| 5 | 269 | 795 | 6.4 | 98.0% | 40% | +12.0 pts |
| 6 | 57 | 193 | 5.6 | 98.8% | 24% | +17.5 pts |
| 7 | 54 | 174 | 5.2 | 98.8% | 21% | +17.4 pts |
All figures are medians. Seconds per puzzle and accuracy describe the sets that completed each cycle. The last two columns first compare each set with its own first cycle, then take the median of those paired comparisons. For example, +12.0 points at cycle 5 is the median within-set accuracy change; it is not the difference between the raw 85% and 98% medians.
Three things stand out.
The cycle-5 paired medians show faster solving and higher accuracy. The median time ratio is 40% and the median accuracy change is +12 points among those 795 sets. These separate medians do not establish that every set improved both measures. Track your own accuracy alongside speed.
Every paired comparison uses cycle 1 as its baseline. Cycle-2 completers have a median time ratio of 76% of their own first cycle; cycle-5 completers have a ratio of 40%. Because these rows include different sets, dividing one row by the previous one does not tell us the typical improvement between adjacent cycles.
Solving time and calendar time are different measures. The Woodpecker Method asks you to finish each cycle in half the calendar time of the previous one. This study measures solving time per puzzle, not days between the start and finish of a cycle, so it cannot tell us how closely players followed that schedule. See the cycle schedule in our guide.
Paired time changes by puzzle rating
Median time per puzzle as a share of each set's own first cycle, by target puzzle rating. Each cycle includes only the sets that completed it; the sample changes across points.
| Target rating | Sets (cycle 1) | Cycle 1: seconds, accuracy | Cycle 5: seconds, accuracy | Cycle 5 time vs cycle 1 |
|---|---|---|---|---|
| Under 1000 | 71 | 15.1s, 90% | too few sets | 77% by cycle 3 |
| 1000–1399 | 937 | 13.6s, 90% | 6.1s, 97.6% | 50% |
| 1400–1799 | 249 | 24.6s, 77% | 9.3s, 96% | 38% |
| 1800+ | 311 | 39.5s, 61% | 5.9s, 99% | 15% |
Among sets completing cycle 5, the median paired time ratio is 50% for target ratings 1000–1399, 38% for 1400–1799, and 15% for 1800+. The raw time and accuracy columns use different samples at cycles 1 and 5, so they should not be read as changes in the same cohort.
The rating groups can also differ in player experience, practice habits, and which sets continue. These figures do not show that choosing harder puzzles causes larger gains, or distinguish recall from fresh calculation. The 1800+ group starts with 311 sets from just 74 players, so players with several sets contribute more heavily. When repeating positions, keep checking why the move works.
Set size
| Set size | Sets (cycle 1) | Cycle 1: seconds, accuracy | Cycle 5 time vs cycle 1 | Cycle 5 accuracy change |
|---|---|---|---|---|
| 50–99 puzzles | 339 | 22.9s, 78.8% | 32% | +20.0 pts |
| 100–150 puzzles | 1,087 | 16.3s, 87% | 42% | +11.0 pts |
| 151–500 puzzles | 119 | 19.2s, 86% | 49% | +10.0 pts |
Among cycle-5 completers, smaller sets have lower median paired time ratios and larger median accuracy changes. The groups differ in more than size, so this does not establish that reducing set size causes more improvement. The cycle-1 figures cover all first-cycle completers; the cycle-5 comparisons cover only those that reached cycle 5. Sets of 20–49 puzzles did not have enough completed cycles to report.
Most players use 100–150 puzzles, Peck's default starter size. The book's benchmark still applies: choose a set whose first cycle you can finish in a few weeks. See how to size your set.
How many players keep going
Of the 1,568 sets that completed a first cycle, 795 (51%) have completed a fifth. By player, 269 of 663 (41%) have reached cycle 5 on at least one set. Some of the rest are still in progress, so these are floors, not final completion rates.
Few go beyond five: 174 sets have completed a seventh cycle; five is Peck's default target. Among those 174 sets, the median time ratio is 21% of their own first cycle. This comparison with cycle 1 does not tell us their change from cycle 6 to cycle 7.
If you stall, review your set size, difficulty, and daily schedule. Our guide to how many puzzles to solve a day helps you set a pace you can keep.
What this data does not show
- Rating gains. We measured performance on repeated positions only. Getting faster at puzzles you have seen before is expected; whether it carries over to your games depends on the patterns and on everything else you study. Test yourself on fresh puzzles and your own games.
- Cause and effect. Players who complete five cycles may differ from those who stop. The "vs own cycle 1" columns compare each set with itself, which helps, but later cycles still come only from sets that got that far.
- Individual results. These are medians. Your first cycle may be slower or faster, and so may your improvement.
Methodology
- Data: every completed cycle on Peck through 30 September 2026 in UTC (completed before 1 October 2026 at 00:00 UTC), from sets of 20 puzzles or more, cycles 1–7. Unfinished cycles are excluded.
- Time per puzzle: a cycle's total solving time divided by the puzzles in the set. Accuracy: puzzles solved correctly divided by puzzles in the set, so skips count as misses.
- Medians across sets. A player with several sets contributes each one. Raw medians use all sets completing that cycle, so their samples differ. "Time vs own cycle 1" is the median of each set's own time ratio; "Accuracy vs own cycle 1" is the median of each set's accuracy difference in percentage points. Neither is calculated from the raw medians. The paired columns also use different samples in each row and cannot be divided or subtracted to infer typical adjacent-cycle changes.
- Rating band is the set's target puzzle rating (Lichess scale), not the player's playing rating.
- Minimum sample: a figure is published only when it covers at least 20 players and 30 sets. Cycles 6–7 are shown overall only for that reason.
- Privacy: only aggregates left the database; no player can be identified. See our privacy policy. The query is in Peck's repository at
sql/studies/woodpecker-cycles.sql. The published figures are preserved insql/studies/woodpecker-cycles-2026-09-30.published.json; the original query export is not included in the repository. Peck's own internal accounts were not filtered out.
Try it yourself
The fastest way to see the curve is to run a set. Pick a theme from the puzzles by theme pages or try three puzzles without an account. The Woodpecker Method guide covers the rest.