Which Sudoku Techniques Slow Players Down? Data from 243,104 Moves

When a sudoku board still has a naked single on it, our players place their next correct digit in a median of 3.6 seconds. When the only moves left are hidden singles, that wait grows to 14.6 seconds. When the board has no single at all and needs a pointing pair, it grows to 43 seconds, and almost two in three of the digits players eventually place there cannot be justified by any logical technique we check for.

Those numbers come from 243,104 correct placements by 212 players in solo Sudoku Royale games between 6 July and 24 September 2026. Below we add Dojo drill logs, a grade of 457,048 puzzles, 5,362 timed solves and notes data. Everything is an aggregate. The CSV is free to reuse with a link back to this page.

Download the aggregates (CSV, 61 rows)

How long players stall, by what the board needs

For every correct digit a player places, our server logs how long it has been since their previous move. It also works out the cheapest technique that would make progress on the board they were looking at. Grouping the waits by that cheapest technique shows how hard each rung of the ladder feels in practice.

"Hidden single" here means no naked single was left, and "pointing" means no single of either kind was left. The last column needs a word of explanation. Each placed digit is checked against 13 techniques, from naked singles up to simple colouring. If none of them proves the digit on the board the player could see, we log it as a guess. On boards that still offer a naked single, only 2.7% of placements are guesses. On boards that need a pointing pair, 62.5% are. Some of those will be sound reasoning our checker does not cover, and some come from the player's own notes, but most players facing a pointing position do not find the pointing pair. They wait, then place a digit they cannot yet prove.

Boards that needed claiming, naked pairs or anything harder had fewer than 20 players each in this window, so we left them out.

Naked singles vs hidden singles

Looking at the digits players chose to place, a naked single took a median of 2.5 seconds and a hidden single 6.7 seconds. The tails differ more than the medians: 21.7% of hidden singles took longer than 20 seconds to place, against 7.1% of naked singles.

We also have placements classified as needing pointing, pairs, triples, X-Wing, swordfish, XY-Wing and colouring, and they look implausibly fast (a median of 4.0 seconds for X-Wing placements). The catch is that for 95% to 100% of those placements an easier move existed somewhere else on the board. Players were not hunting for an X-Wing. They were most likely placing a digit they had worked out another way, or guessing. The CSV includes these rows, flagged, and we do not read them as find times.

Practice halves the time to find a hidden single

To measure practice without mixing in who keeps playing, we compared the same players with themselves. For each player we took the median time of their first ten finds of a technique and the median of their finds after number 500.

Both techniques get about twice as fast. The gain starts early: among 161 players with at least 20 hidden-single finds, finds 11 to 100 took a median of 9.6 seconds against 11.3 seconds for their first ten, and 65% of those players got faster. For naked singles the same comparison over 205 players went from 5.4 to 4.4 seconds.

Drills: speed and accuracy when you know what to look for

Dojo drills show a board, tell you the technique, and time each answer. That isolates recognition from the question of which technique to try.

One in four hidden-single drill answers is wrong, against one in twenty for a full house. Drill times run slower than in-game times because every drill board is new: in a real game you have already been scanning the grid before the move. With practice, the 29 players who did at least 20 naked-single reps went from 12.1 to 9.5 seconds (reps 1 to 10 vs 11 to 50), and 27 hidden-single players went from 13.7 to 10.9 seconds. Their accuracy did not move (89.7% to 89.8%, and 72.2% to 72.9%), so drill practice buys speed rather than fewer mistakes. Drills for pointing, pairs and the advanced patterns had fewer than 20 players each, so we do not report them.

Which techniques puzzles need at each difficulty

Our grader solves every puzzle in the library the way a careful human would: at each step it uses the cheapest technique that makes progress. We rate a puzzle by the hardest technique it needed and group puzzles into four tiers. The table shows how often each technique is used inside each tier, from 457,048 active graded puzzles.

Three things stand out. First, 52% of singles-only puzzles never need a hidden single; naked singles alone finish them. Second, pointing is the workhorse of every harder tier: the median puzzle in the pairs tier uses it four times. Third, the hardest step is usually needed once. In the top tier the median puzzle uses its XY-Wing or colouring step exactly one time, so a hard puzzle is mostly an ordinary puzzle with one wall in it. X-Wing and swordfish are rarer than their reputation: 11.4% and 2.9% of top-tier puzzles.

Because our grader stops at simple colouring, puzzles that need chains or uniqueness arguments are rejected and never reach players. The top tier here is hard, but it is not the hardest sudoku that exists.

Clue count does not separate the harder tiers

A puzzle with 35 or more clues is almost certainly easy. Below 27 clues, though, a pointing puzzle and an XY-Wing puzzle have the same clue distribution, median 25. Counting givens tells you nothing about which of the harder tiers you are in. Our guide to difficulty levels explains how the techniques set the rating.

Solve time by hardest technique

The biggest jump is the first one. Needing a single pointing pair adds about six minutes to a median solve, more than the step from pointing to pairs and from pairs to wings combined. Per empty cell, the time almost doubles between the first and last tier.

Notes usage climbs with the technique

Only 37% of singles-only puzzles get any notes; 82% of pairs puzzles do, which makes sense since a naked pair is hard to see without candidates written in. The top tier dips to 60%. That tier had a small, busy group of 24 players, and this is our shortest data window (12 days), so treat that dip with care.

What this suggests for your own solving

If you want to get faster, hidden singles are where the time goes. They are the most common move you will struggle with, the drill with the lowest accuracy, and the technique where practice paid off most. After that, learn to find pointing pairs: they show up in 80% to 95% of the puzzles in every harder tier and are the point where most players stall and start guessing. Our technique guide covers the whole ladder, and the naked pairs and hidden pairs pages cover the next step up.

Methodology

• Source. Read-only SQL against the Sudoku Royale production database on 24 September 2026. No player-level data leaves the database; this page and the CSV hold aggregates only.
• Who counts. We used the same player filter as our internal analytics: no bots, no test or synthetic accounts, no accounts created from development or TestFlight builds. Automated test runs that delete their accounts afterwards are excluded too, since only accounts that still exist count. Solo-only players are included. Any group with fewer than 20 players is suppressed.
• Move timing. Each correct placement in solo Practice and Daily games logs the milliseconds since the player's previous event (a placement or a mistake) and the first move of each game is dropped. Live matches are excluded because an opponent can fill a cell for you and end your think early.
• Technique labels. The technique for a placement is the cheapest of 13 techniques that proves that digit on the board the player saw. "Board needs" is the cheapest technique that makes any progress anywhere on the board. Both are computed from the given digits and placed digits, not from the player's own notes.
• Drills. Each Dojo rep logs its time and whether the answer was right. Times are for correct answers. Locate-style drills (tap the solvable cell) are excluded. Drill data runs from 30 May to 24 September 2026.
• Puzzles. 457,048 active puzzles, each graded by applying naked single, hidden single, pointing, claiming, naked and hidden pairs and triples, X-Wing, swordfish, jellyfish, XY-Wing and simple colouring, always cheapest first. Technique counts reflect that order: a cell that is both a naked and a hidden single is counted as a naked single.
• Solve times. Completed Practice and Daily puzzles, excluding tutorial puzzles with fewer than 40 empty cells and any solve faster than 1.5 seconds per empty cell (a timer artifact). Tier comes from the grader, not the level the player chose.
• Speed of the queries. Each aggregate ran in 0.03 to 2.4 seconds on the production database. The full SQL is in the pull request that added this page.

How to cite this

Sudoku Royale (2026). "Which Sudoku Techniques Slow Players Down? Data from 243,104 Moves." https://www.sudoku-royale.com/resources/sudoku-technique-study. Data licensed CC BY 4.0.