Why "write every candidate in every cell" is making your hard puzzles harder
Quick context: I've been building a sudoku web app for the last few months and went deep on pencil-mark UX. Wanted to share what I learned because I see a lot of "I'm stuck on hard" posts here where the grid is buried under candidates.
Most apps (and most players) use full notation — every empty cell gets every possible candidate written in. It's complete, it's safe, and on hard puzzles it turns the grid into visual noise. You're scanning a wall of tiny numbers trying to find the one that matters.
Thomas Snyder — multiple-time World Sudoku Champion — uses something almost opposite. The rule:
**Only mark a candidate when it's restricted to exactly two cells inside a single box.**
That's it. Not three, not "everywhere it could go" — just the two-cell pairs inside boxes.
Why it works:
- Pointing pairs become obvious instead of hidden
- X-Wings, hidden pairs, naked pairs all light up because the marks ARE the patterns
- The grid stays readable so your eyes can actually scan it
- You stop "solving by candidate elimination" and start solving by pattern recognition (which is faster)
The tradeoff: you miss some easy elimination wins that full notation catches automatically. But on Hard/Expert/Evil, where eliminations aren't the bottleneck anyway, it's a clear win.
What changed for me when I built this into the app:
- Players who were bouncing off hard difficulty started finishing them
- Solve times on Expert dropped meaningfully when I A/B tested the modes
- The "hint" button got used way less, which I think is the real signal
If anyone wants to mess with it, the app is at sudoku247online.com — has Snyder mode + full notation as a toggle, plus the harder tiers (Master/Evil) where this technique actually starts to matter. Free, no signup.
Curious if anyone here already uses Snyder or a variant of it — what works for you on the really nasty puzzles?