About the par model

Live coefficients are loaded from handicaps.json at the bottom of this page.

Every daily and weekly puzzle in GregSweeper has a par time — the time the model expects a typical player to need on that specific board. Your personal handicap shifts par up or down based on your past plays, golf-style.

What's in the model

The model is a Bayesian mixed-effects regression of completion time on board features:

time ~ moveType_coefficients + boardShape_coefficients + gimmickCellCount_coefficients + (1 | user)

The fixed-effect coefficients are global (every player shares them). The random intercept (1 | user) is your personal handicap — the average residual between your times and predicted par.

The fit, in detail

Your personal handicap

If you've played at least 30 daily puzzles, your handicap is fitted directly as the random intercept on the model. Below that, the game shows a provisional handicap computed from the mean residual of your last few plays (with bomb-hit time cost subtracted). The provisional handicap converges to the fitted one as you accumulate plays.

Where the code lives

Current coefficients

Loading…

← Back to GregSweeper