About RatingForge

RatingForge teaches the mathematics of skill measurement the way a statistician meets it: by building a rating system and watching it work. You give a league of bots hidden true strengths, play a seeded season, and watch Elo converge on the true order — then you change the parameters and see what breaks. It is a statistics-and-simulation lab for ages 15–18 (logistic curves, Monte-Carlo convergence, parameter sensitivity, and the idea of a confidence band).

The cast (a text-forward case, no illustrations)

RatingForge is an older-teen app, so its story is delivered as realistic case personas — no cartoon cast, no chapters (that is deliberate for this band).

• Devi, the tournament statistician — embodies the season as a measurement: “one game is an anecdote; a season is data.” She asks what the numbers can and cannot tell you.
• Rourke, the rating-system designer — embodies the trade-off: every choice of K, every draw rule, every uncertainty model buys one thing at the cost of another. He asks “what are we optimising for, and who does it hurt?”

The motivating case

Online game platforms constantly field the complaint “why did I only get 9 points for beating someone way above me?” The honest answer is the logistic curve: you were expected to win that game only rarely, so the system had already priced most of the credit in. RatingForge lets a teen see that isn’t unfair — it is K × (result − expected), applied identically to everyone.

What it does NOT do

• No AI near the statistics. Every number is a deterministic, seeded computation you can reproduce — never a guess from a model.
• No claim that a rating measures worth. A rating is a running prediction about games; that is all.
• No accounts, no data leaves your device. The whole simulation runs in your browser; the only thing stored locally is how many seasons you have run.
• No leaderboards, streaks, or pressure. The lab is calm — run a season, change a knob, run again.

Honest yield

RatingForge is practice with real statistical ideas — expected value, convergence, and sensitivity — presented through a system teens already care about. It is not a claim that using it raises test scores or makes anyone “smarter.”

Part of the Spark & Anvil portfolio · Program #113 (Board-Game Platform & Together). Web-pioneered — the iOS app is a separate, founder-gated build.