The Puzzle rating rules: a Glicko-2 module, its ADR and the glossary terms #441

New issue

Closed

opened 2026-10-16 08:19:34 +00:00 by root · 9 comments

root commented

2026-10-16 08:19:34 +00:00

(Migrated from gitlab.core-project.space)

Parent

#440

What to build

The rules of the Puzzle rating, written down and computable, before anything calls them.

A pure Glicko-2 module in the API. One function takes a Puzzle rating (rating, deviation, volatility), a Puzzle's rating and deviation, and a result (1 solved, 0 failed), and returns the new Puzzle rating. The volatility step is the iterative algorithm from Glickman's paper. Each result is its own rating period. A second function says whether a Puzzle rating is provisional (deviation above 110). Defaults for a new Account: 1500 / 350 / 0.06.
A new ADR, the Puzzle rating rules: Glicko-2 updated after every result; Puzzle ratings fixed at their imported values; only the Attempt counts; a Hint, the solution or giving up fails it; no deviation growth for inactivity in v1; provisional while deviation > 110; the system constant τ and why. Added to the ADR index.
Glossary: Puzzle, Theme (and curated Theme), Puzzle rating, Attempt, Replay, Hint, and a flagged ambiguity "Attempt vs Replay", worded as in #440.

Nothing calls the module yet; #443 does.

Seam: S2 (the Glicko-2 module). Covers stories 14-17 and 19 of #440.

Acceptance criteria

The update reproduces the worked example in Glickman's Glicko-2 paper to its published precision.
Tests show a new Player (1500 / 350) gains far more than a settled one (1500 / 60) from the same solve, and that a failure lowers the rating.
Deviation never grows from a result.
The provisional function is true above 110 and false at or below it.
The module has no I/O and no framework import.
The ADR is written, accepted, and listed in the ADR index (the index check passes).
GLOSSARY.md holds the six terms and the flagged ambiguity.
The gate is green (CONTRIBUTING.md, "The five commands").

## Parent #440 ## What to build The rules of the Puzzle rating, written down and computable, before anything calls them. - **A pure Glicko-2 module in the API.** One function takes a Puzzle rating (rating, deviation, volatility), a Puzzle's rating and deviation, and a result (1 solved, 0 failed), and returns the new Puzzle rating. The volatility step is the iterative algorithm from Glickman's paper. Each result is its own rating period. A second function says whether a Puzzle rating is provisional (deviation above 110). Defaults for a new Account: 1500 / 350 / 0.06. - **A new ADR, the Puzzle rating rules:** Glicko-2 updated after every result; Puzzle ratings fixed at their imported values; only the Attempt counts; a Hint, the solution or giving up fails it; no deviation growth for inactivity in v1; provisional while deviation > 110; the system constant τ and why. Added to the ADR index. - **Glossary:** Puzzle, Theme (and curated Theme), Puzzle rating, Attempt, Replay, Hint, and a flagged ambiguity "Attempt vs Replay", worded as in #440. Nothing calls the module yet; #443 does. Seam: S2 (the Glicko-2 module). Covers stories 14-17 and 19 of #440. ## Acceptance criteria - [ ] The update reproduces the worked example in Glickman's Glicko-2 paper to its published precision. - [ ] Tests show a new Player (1500 / 350) gains far more than a settled one (1500 / 60) from the same solve, and that a failure lowers the rating. - [ ] Deviation never grows from a result. - [ ] The provisional function is true above 110 and false at or below it. - [ ] The module has no I/O and no framework import. - [ ] The ADR is written, accepted, and listed in the ADR index (the index check passes). - [ ] `GLOSSARY.md` holds the six terms and the flagged ambiguity. - [ ] The gate is green (`CONTRIBUTING.md`, "The five commands").

root commented

2026-10-16 08:20:07 +00:00

(Migrated from gitlab.core-project.space)

mentioned in issue #443