Why Diablo Times Lie: Simpson's Paradox in the Daily Set Puzzle
Published October 3, 2026
You've got two cards picked. One more and that's a Set. Then the clock hits 0:30, and every card on the board flies to a new spot.
That's Diablo mode: every thirty seconds of game time, the whole board shuffles. It should be the harder mode. But look at the leaderboard:
The typical Diablo game takes 1:47. The typical Regular game takes 2:09. Diablo looks 17% faster. We dug through 6,291 Diablo games and 11,605 Regular games to find out why.
Who actually plays Diablo?
Start with the players. Each dot below is a signed-in player with at least five Regular games, placed by their typical Regular time.
- Players who also play Diablo finish Regular in a median of 2:08.
- Players who skip it take 4:01.
So the Diablo board is packed with fast players. 49% of Diablo games come from players who usually finish Regular in under two minutes, against 33% of Regular games.
It's like a race where only the pros take the hard course. The hard course ends up with the fastest average times, not because it's easier, but because of who is running it.
Same player, same day
The fair comparison is the same person on the same day. Across 4,150 days where a player finished both modes (306 players), Diablo was 9% slower: a median of 6 seconds a game.
It's not a sure loss. Diablo still came out ahead in 44% of those pairs, because daily boards differ and so do days. But on average, it costs you.
Split players by skill and it's the same story:
| Typical Regular time | Diablo vs Regular, same player, same day |
|---|---|
| under 1:00 | +9% |
| 1:00–2:00 | +15% |
| 2:00–4:00 | +5% |
| over 4:00 | −1% |
| everyone pooled | −17% (Diablo "faster") |
Faster overall, but slower for almost every group of players. That's a famous statistical trap called Simpson's paradox: a trend that holds in every group can reverse when the groups are pooled, because the groups aren't mixed evenly.
The shuffle tax
So where do those six seconds go? Diablo shuffles at exactly 0:30, 1:00, 1:30 and so on, on the same clock the leaderboard measures, so we can line up every Set found against the moment of the last shuffle and compare it with Regular games at the same point on the clock.
- In the first second after a shuffle, players find Sets at 17% of their normal pace.
- It takes three or four seconds to get back up to speed.
- Added up, that's about 1.75 seconds lost to every shuffle.
Your eyes had a plan: the rows you'd already checked, the card you were hunting for. The shuffle throws it all away.
A typical Diablo game has three shuffles: 3 × 1.75 ≈ 5.2 seconds, nearly the whole 6-second penalty. The dip looks the same early, mid and late in a game, and for the last Set too. There's no last-second rush before a shuffle, either.
One twist
Players who usually take over four minutes show no Diablo penalty at all (−1%), even though their games go through many more shuffles. We don't know why yet. One possibility is that when you're truly stuck, a fresh layout helps you see the board differently. That's a guess, not a finding. We may come back to it.
How to beat the shuffle
- Finish before 0:30, and it never happens. That's easier said than done: only 4.5% of Diablo games finish before the first shuffle.
- Search by number. A one is still a one wherever it lands, so a plan built on groups (all the ones, then the twos, then one of each) survives the shuffle. A plan built on positions ("I've checked the top two rows") does not. See why the last Set is so hard for why searching by number is faster anyway.
- Hold on to your picks. Cards you've selected stay selected through a shuffle, in the app and on the web, so finish the Set rather than starting over.
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). Skill groups use each signed-in player's median Regular time over at least five games. The shuffle figures compare when Sets are found relative to each 30-second mark in Diablo against the same clock points in Regular, which has no shuffles. No individual player's data is shown: each dot and line is anonymous.
Special thanks to Sarah Rosston for the idea behind this series.