The Calibration Gap We Explained Away
PlayerGem's calibration review found a persistent confidence-accuracy gap for one quarter — and the shop spent most o...
Read itLatest Articles
No Single Night May Dominate the Slate
How PlayerGem's concentration cap on single-night ratings was written, what it cost in apparent precision, and why we...
Read itIf You Cannot Grade It, Do Not Publish It
PlayerGem's rule against publishing ungraded ratings — what broke without it, what the shop tried, and why the rule s...
Read itRules We Made Up That Turned Out to Be Wrong
Four internal rules PlayerGem ran on for months — each one written after a real failure, each one eventually wrong in...
Read itDistrust Any Result You Are Pleased About
Bad results get investigated because they demand it. Good ones supply their own explanation, which is exactly why they need scrutiny more.
Read articleNever Delete a Rating, Only Grade It
Once a rating exists it stays and gets graded however it turns out. The rule sounds like theatre until you see what deletion does.
Read articleWrite the Threshold Down Before You See the Data
The single most useful rule we have, and the one that took the longest to accept, because following it means being wrong on the record more often.
Read article