Does the skill rating really predict who wins, or is it guessing?

If a rating says one side should win nine times out of ten, the logs should show close to that. This entry explains how to compare promise and outcome in groups without counting every match twice from both sides.

Why this works

A rating is a promise about how often the favorite wins. The way to check a promise is to group matches by what was promised and see whether the favorites win that often. Gaps that grow with the promise show the ratings are spread too wide or too narrow.

When not to use it

It cannot judge a rating when most matches are between players who are very unevenly matched or when there are few games per group. And it says nothing on whether matches feel fun. Gather more games first, and ask players how their games felt.

In the full entry

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

Say this to your AI

Use the Spielerando entry "Does the skill rating really predict who wins, or is it guessing?" (read_entry rating-prediction-versus-real-results) 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