Six minutes for an easy grid. Twelve for a medium, twenty-two for a hard, thirty-five for an extreme. Those are the estimates I put on the timed challenges in my own Sudoku app, and if your times land within a few minutes of them, there is nothing wrong with your solving.
The word doing all the work up there is the grade, and a grade is a house style rather than a unit. Nobody agrees on where hard starts. So a time has to travel with the name of the puzzle set it came from, or it means nothing: yours against yours, on one source, is the only comparison that survives contact with a second publisher.
What does carry across is the shape of the clock. A hard puzzle is not slow because it has more cells to fill. It has barely any more. It is slow because of what happens in the gaps between the digits, and that is the part of a solve almost nobody measures.
Four reference times, and what they hide
Here are the four grades in the puzzle set I ship, with the givens counted out of the files themselves and the challenge estimate next to them.
| Grade | Givens | Cells to fill | Estimate | Seconds per digit |
|---|---|---|---|---|
| Easy | 36 | 45 | 6 minutes | 8 |
| Medium | 31 | 50 | 12 minutes | 14 |
| Hard | 26 to 28 | 53 to 55 | 22 minutes | 24 |
| Extreme | 21 to 27 | 54 to 60 | 35 minutes | 37 |
Now read the two middle columns together. Going from easy to extreme, the amount of writing you do goes up by about a quarter. The clock goes up by nearly six times.
Nobody writes a digit five times slower on an extreme puzzle. The writing was never the cost. What grows is the search you have to run before each digit becomes writable, and on a hard grid that search stops paying out long before the grid is full.
Want your own numbers instead of mine? Sudoku Master times every game and keeps a best and an average for each difficulty, so you are measured against yourself rather than against a stranger's screenshot. Free, works with no signal, and no ads while you solve.
An easy grid, timed by the sweep
The unit of time in a Sudoku is not the digit. It is the sweep: one pass where you take the digits 1 to 9 in turn and, for each one, check the nine boxes for one with a single legal home left. Cross-hatching, in other words. Every solver does this, and on easy puzzles it does all the work.
Here is an easy puzzle from that set. Thirty-six givens.
..7.5.23.
..193.45.
8.3..2.1.
.2.8.317.
6.847....
1..5.9..6
...387..2
.42....9.
...2...61
Start on the 1s. Box 5 is the middle box, rows 4 to 6 and columns 4 to 6. Row 4
already has a 1 at r4c7 and row 6 already has one at r6c1, so both of those
rows are struck out inside the box. That leaves r5c6, and the 1 goes in
without you writing a single candidate.
Keep going in that order and the first sweep places 25 digits. The second places 13. The third places 6, the fourth places the last one, and the puzzle is finished. Four sweeps. No pencil marks anywhere, no technique above a hidden single, nothing to hold in your head between one digit and the next.
Your split will differ a little if you take the digits in another order. The shape will not. Easy grids pay out on the first sweep and keep paying, and six minutes is mostly the time it takes to write 45 digits and look at the grid four times.
The same method on a hard grid
Now a hard puzzle from the same set, 26 givens.
9.8..4...
4.16..23.
....2....
8....2...
.6...519.
..9.....8
..2..194.
...9.3...
....5.7.1
The method that just finished the easy grid still works. Box 1 is the top left.
Row 2 has a 2 at r2c7 and row 3 has a 2 at r3c5, so the 2 in box 1 has to
sit in row 1, and r1c1 and r1c3 are already taken. Write 2 at r1c2.
That is the good news, and it lasts for six digits. The first full sweep places 6. The second places 1. The third places 1. The fourth places nothing at all.
Then sweep the rows and the columns for hidden singles as well, and check every empty cell for a naked single. Both come back empty. Here is the grid the method leaves you with.
928..4...
4.16..23.
....2....
8...92...
.6...519.
..9.....8
..2..1943
...943...
39425.7.1
Eight digits placed. Forty-seven cells still empty, which is more than half the board. You have spent four sweeps of a technique that is only fast while it is working, and the puzzle has not been touched in any real sense.
That is what a hard grade actually buys you. Not more digits. A wall that arrives after eight of them.
Where the extra sixteen minutes go
From the stall, the only way forward is to write down what each empty cell can still hold. Count that job before you start it: 47 cells, and between them 158 candidates. Nine of the cells are already down to two, which is where the next deduction will come from.
So the honest budget for a hard puzzle looks like this.
| Phase | What you are doing | Where the time goes |
|---|---|---|
| Opening sweeps | Cross-hatching, 8 digits here | Fast, and it stops early |
| The dead sweeps | Sweeping again because it worked last time | Pure loss, a minute each |
| Markup | Writing 158 candidates into 47 cells | Slow, silent, and it places nothing |
| Deductions | A pair, a pointing pair, a box-line cut | A few seconds of thinking, a few digits each |
| The cascade | Singles again, now that a candidate is gone | Fast, and it is why hard puzzles finish suddenly |
Three of those five phases place no digits at all. That is the whole answer to where sixteen extra minutes come from, and none of it is about writing speed.
Time yourself on one cell of markup if you want your own figure. Pick an empty cell in the grid above, work out its candidate list, and see what it cost you. Multiply by 47. Most people are surprised, and it explains why marking up feels like it takes forever: it does, and it is still the cheapest route through.
How to actually get faster
Speed on a Sudoku is almost entirely about not wasting sweeps. Five things cut real minutes.
Stop sweeping when a sweep stops paying. On the grid above, the second and third sweeps placed one digit each and the fourth placed none. The rule that saves the most time is the bluntest: if a full pass gives you fewer than three digits, that phase is over. Mark up.
Match the tool to the grade before you start. On a 36-given easy puzzle, writing candidates is wasted effort you will rub out unused. On a 26-given hard one, marking up after the first stall instead of the fourth saves two or three minutes on its own.
Hunt the digits that are nearly done. A digit already placed seven times has two boxes left to live in, so it takes seconds to check and often gives you a placement. A digit placed twice has seven boxes and will usually give you nothing. Sweep in that order rather than 1 to 9.
Work the tightest unit, not the prettiest. A row with two blanks is a one-second job. A box with six is a project. When you are looking for somewhere to spend attention, count blanks first.
Learn the next rung. This is the big one, and it is not really a speed tip. A grid built around a technique you do not know costs you the whole session. The same grid, once you know it, costs ninety seconds. Most people who feel slow on hard puzzles are not slow. They are running one technique short, usually the pair of shapes where a box meets a line.
About fast times and speed runs
Times get posted, and they are worth almost nothing without the grid attached. A time is a statement about one specific puzzle, and two puzzles wearing the same grade name can differ by a factor of four in the work they demand. Somebody else's 3 minutes and your 11 may be the same solving.
There is a second reason to be careful with them. Fast solving is pattern recognition, and it is trained by volume on one source, not by hurrying. The person doing 4-minute grids has seen that generator's habits a thousand times. They are not reading faster than you. They are reading less, because they already know where the puzzle keeps its hidden singles.
If you do want to race, race on a grade below your usual one. Speed comes from solving well within your technique, and it collapses the moment the grid needs a move you have to think about.
The timer: keep it or kill it
The timer is worth having for one job and is actively harmful in another.
Keep it on if you use it as a record. Best and average times per difficulty, over a few weeks, on one source, tell you honestly whether you are improving, and they catch the thing that a single lucky grid hides.
Turn it off if it makes you place a digit you cannot justify. Under pressure a guess feels like a shortcut, and it is a coin flip on a grid that has one answer and a route to it. The cost is not the flip. It is that a wrong digit stays legal for twenty more placements before the grid dies on it, and by then you are rebuilding the whole board. That failure is the most expensive thing that can happen to your clock, and it is worked through in why some puzzles feel impossible.
There is a middle setting most people miss: leave the timer running and stop looking at it. You get the record without the pressure.
Measuring yourself honestly
Four rules make your own times mean something.
- One source at a time. Compare a paper against itself, or an app against itself. Cross-publisher comparison is noise.
- Use the median, not the best. Your fastest time is a story about one friendly grid. The middle of your last ten is the number that actually moves when you improve.
- Count mistakes next to the clock. A 9-minute solve with three corrections is worse than a 12-minute clean one, and it is the one that will bite you on a harder grade.
- Expect a bad week when you level up. Moving from medium to hard makes your times worse for a fortnight, on purpose. That is what learning a new technique looks like on a graph.
What I built, and what it will not tell you
I built Sudoku Master, and the honest position is that it measures your time without explaining it. The statistics screen keeps games completed, games lost, best time, average time and your lowest mistake count for each of the four difficulties, plus streaks and an activity heatmap, so the comparison this page recommends is the one it makes. Each of the 4,000 puzzles ships with its solution inside the app, so none of that needs a signal.
What it does not do is tell you which minute you wasted. There is no technique-by-technique commentary in it: the solver returns a finished grid, not a lesson. The settings that genuinely help the clock are duller than a hint. Highlighting the row, column and box of the cell you touch draws the cross you were holding in your head. Clearing notes automatically when a digit goes down stops you re-reading marks that stopped being true four placements ago.
Common mistakes
Comparing your time to somebody else's grade. Their hard and your hard are two different words. The comparison that works is you against you, on one source.
Sweeping a fifth time. If two full passes have produced nothing, the grid is not hiding a single from you. It is asking a different question, and only markup answers it.
Marking up an easy grid. Forty-five cells of candidates on a puzzle that falls to cross-hatching is four minutes spent to save none.
Racing on a grade that is too hard. The clock makes you guess, the guess makes you wrong, and the wrong digit costs more than every minute you saved.
Reading a slow session as a verdict on you. It is far more often a verdict on one grid, or on the fact that you are solving it at 11pm.
Questions people ask
Does a faster time mean you are better at Sudoku?
Not on its own. A time measures the fit between one grid and the techniques you own, so an easy puzzle solved in four minutes says less about you than a hard one finished cleanly in half an hour. Improvement shows up as a falling median on one source, and as fewer corrections.
Why do some easy puzzles take longer than hard ones?
Because the grade is a label, not a measurement, and one label covers a range. An easy grid with an awkward digit distribution can cost you two dead sweeps, while a hard one that hands you an early pair falls over in ten minutes. Judge a grade by ten puzzles, never by one.
Should I turn the Sudoku timer off?
Turn it off if watching it makes you place digits you cannot justify. Keep it on if you look at it only after the grid is finished. The record is useful and the pressure is not, and you can have the first without the second.
Is it bad to spend an hour on one puzzle?
No. An hour on an extreme grid is a normal amount of time, and the puzzle does not expire. If you have stopped placing digits entirely, put it down for ten minutes rather than staring: attention on a fixed pattern degrades, and the cell you cannot see is usually one you have looked straight at.
Do pencil marks make you slower?
They cost time up front and save more than they cost on anything above medium. The mistake is the order. Sweep until a full pass stops paying, then mark up, and start with the unit that has the fewest blanks so the shortest list is also the first one you write. An easy grid never gets there.
Why has my average solving time got worse?
Usually one of three things. You moved up a grade, which makes times worse for a couple of weeks by design. Or you changed source, and the new one grades harder. Or you are counting sessions where you put the phone down mid-grid, which the clock cannot tell from thinking.
Does a puzzle with fewer clues take longer to solve?
Not reliably. The clock follows the gaps between placements rather than the number of blanks, so a 21-given grid that keeps handing you singles finishes sooner than a 28-given one that goes quiet twice. That is why the givens column in the table above overlaps between the hard and extreme rows.
How many Sudoku puzzles should I do in a session?
However many you enjoy. One a day on a grade that stretches you is worth more practice than six easy ones, because the technique you are missing only shows up on the grids that need it. Nothing about the timing changes if you do three.
Keep reading
- What actually makes a Sudoku easy, medium, hard or extreme, for what the grade on your puzzle is really measuring
- Cross-hatching: how to read a Sudoku grid quickly, the sweep that eats most of the clock
- How to use pencil marks without drowning in them, for markup that costs three minutes instead of ten
- Stuck on a Sudoku? Work this checklist in order, for when the clock runs and the grid does not move
- A daily Sudoku habit that actually sticks, for fitting one into a real day
- Sudoku: rules, techniques and solving, for how all of this fits together
Get it: Sudoku Master, free on iPhone and Android. 4,000 classic puzzles, timed and tracked per difficulty, and no ads while you solve.
