Why the Last Set Is the Hardest: What 17,896 Set Puzzle Games Reveal
Published October 3, 2026
Set Puzzle 🎯 🟨9s 🟨9s 🟨11s 🟨13s 🟨19s 🟥34s
If you've played the Daily Set for a while, you know this feeling. Five Sets fall into place, you're on pace for a new personal best, and then Set 6 just… isn't there. You scan the same twelve cards over and over, convinced it can't exist.
You're not imagining it. We looked at every timed game on the leaderboard, 17,896 games and 107,376 Sets since March, and the last Set really is different.
What the numbers say
- Set 6 is the slowest Set in 47% of games. If every Set were equally hard, each would be the slowest about one game in six (17%).
- It eats 32% of the whole game, about twice its fair share of one sixth.
- In a typical game it takes 3.9× as long as the first Set. The median times for Sets 1 through 6 are 9s, 9s, 11s, 13s, 19s and 34s.
What shape is the slowdown?
There are two natural guesses. Maybe the slowdown is steady, with each Set a little harder than the last. Or maybe it doubles, with each Set twice as hard as the one before.
It's neither. The real times follow a harmonic curve, the same curve behind a famous probability puzzle: the coupon collector's problem.
Imagine every cereal box hides one of six prizes, picked at random. Your first box is guaranteed to give you something new. But once you have five prizes, only one box in six has the one you're missing, so on average you'll open six boxes to get it. The number of boxes it takes to find a new prize grows like this:
| Prize | 1st | 2nd | 3rd | 4th | 5th | 6th |
|---|---|---|---|---|---|---|
| Boxes needed, on average | 1 | 1.2 | 1.5 | 2 | 3 | 6 |
Fit that curve to the real Daily Set times (plus a small fixed cost for tapping the cards) and it matches almost perfectly: R² = 0.999 in Regular, and 0.998 in Diablo. A straight line, by comparison, scores 0.75.
Why cereal boxes? Re-finding
The connection is how most of us search. Pick two cards, picture the card that would complete the Set, and look for it.
A board of 12 cards has 66 pairs of cards. Every Set on the board contains three of those pairs, and a pair can only ever point at one card, so a board's six Sets account for exactly 18 pairs. The other 48 point at a card that isn't there.
Early in the game almost every hit is a new Set. But by Set 6, 15 of those 18 hits lead to Sets you've already found. Only 3 of the 66 pairs lead anywhere new, which is 1 in 22. Pick a pair at random and, most of the time, the third card isn't on the board. When it is, it's usually a Set you already have.
You're not missing the last Set. You're re-finding the first five.
Three ways to search, tested
We simulated three strategies on every Daily Set board since February, all 444 of them, counting one "check" each time a pair is tried.
| Strategy | How it works | Checks to find all 6 | Checks for the last Set | Set 6 is the slowest |
|---|---|---|---|---|
| Random pairs | Pick any two cards, look for the third | 54.5 | 22.5 | 54% of boards |
| Reading order | Every pair once, left to right, top to bottom | 48.7 | 12.1 | 29% |
| By number | Ones with ones, twos with twos, threes with threes, then one of each | 32.6 | 3.2 | 0.2% |
Random pairs reproduces the human curve almost exactly. Searching by number turns it upside down, for one reason: 77% of all Sets are one-two-three (one card of each number). An average board has 4.6 of them. Every board in the data had at least 3, and 94% had at least 4.
By number, you clear the same-number groups first. They are slow, with lots of checks for only one or two Sets. Then you pair each "one" with each "two", about 16 pairs that hold 4 or 5 Sets. Each one-two-three Set contains exactly one one-and-two pair, so nothing in that phase can be found twice, and the last Sets arrive about every 3 checks. It also has a finish line: once every one-and-two pair has been tried, you know you're done.
At a typical player's pace (about 1.4 seconds per check), that is roughly 96 seconds for random pairs versus 65 seconds by number, about 30 seconds saved.
Even the best players hit the wall
The ten fastest regular players, with a median game of 39 seconds, still slow down at the end. Set 6 is their slowest in 42% of games.
Their curve is flatter than everyone else's. Their last Set takes 2.2× their first, against 3.9× overall, which is what more systematic scanning looks like. But the shape is still there. A good player's first Sets aren't searched for so much as noticed: three reds in a row, two matching shapes. By definition, the last Set is the one that didn't jump out, so it gets the full search every time.
The good news is that practice works. Players with 60 or more games got about 25% faster over their first sixty, from a median of 1:53 to 1:24.
Try it tomorrow
Next time you're stuck on Set 6, stop picking pairs at random. Go by number: 1·1·1, 2·2·2, 3·3·3, then 1·2·3, and see whether Set 6 still gets you.
About the data: every timed game on the Daily Set leaderboard from March 6 to September 28, 2026 (games without per-Set timings, or over an hour long, are left out). Times are medians, because a few games left open for an hour would pull every average toward the last Set. The strategy figures come from a simulation run on every board played from February 19 onward. No individual player's data is shown.
Special thanks to Sarah Rosston for the idea behind this series.