Rank puzzle levels by how often they get skipped without being fooled by small numbers

A level skipped by three of three people tops any list sorted by percentage, yet tells you almost nothing. This entry shows how to rank levels so thin data cannot jump the queue ahead of levels with solid evidence.

Why this works

A percentage from three people is a rumour, one from three hundred is evidence. Ranking by the cautious end of the plausible range lets a level climb only as far as the data can honestly support, so thin levels cannot jump the queue.

When not to use it

Do not bother when every level has thousands of plays, since raw rates then barely differ from careful ones. And do not read the order as a verdict when two levels are close. Treat the top few as a shortlist to play yourself.

In the full entry

• What to build, step by step
• Starting values and the reason for each
• 3 common mistakes and the fix
• 25 lines of example code with 4 checks that pass
• How your AI shows you that it works

Say this to your AI

Use the Spielerando entry "Rank puzzle levels by how often they get skipped without being fooled by small numbers" (read_entry puzzle-skip-rate-small-sample-ranking) in my game. Build it step by step, and show me how you checked that it works.

Or skip this and connect the MCP: your AI then finds all of this by itself.Connect the MCP