The Grading Window We Trusted Too Long
PlayerGem's grading window held for two seasons before a quiet structural shift made it a source of flattering noise ...
Read itLatest Articles
The Window We Trusted Because We Set It First
PlayerGem set its grading window before seeing results — and found that prior authorship of a rule is a surprisingly ...
Read itGrading the Window Instead of the Rating
PlayerGem's grading desk on the week we discovered our hit rate looked fine because we had quietly optimized the wind...
Read itThe Grading Window We Opened Early
PlayerGem opened a grading window before the fixed deadline and found that curiosity about a live rating is not a neu...
Read itA Column That Passed Every Check and Said Nothing
PlayerGem built a grading column that cleared every internal check, hit its accuracy target, and still told us nothin...
Read itThe Grading Window We Agreed On and Ignored
PlayerGem set a fixed grading window, documented it, and then quietly abandoned it the first time a rating looked bad...
Read itThe Grading Column We Forgot to Check
PlayerGem's grading sheet had a calibration column that went unchecked for six weeks — here is what that silence actu...
Read itThe Grading Column We Added to Feel Thorough
PlayerGem added a secondary grading column to its rating sheets and spent months confusing the appearance of rigor wi...
Read itGrading a Window We Closed Too Early
PlayerGem closed a grading window before a game finished and marked a rating correct — then the final minutes changed...
Read itThe Grading Window We Set After Seeing Results
PlayerGem set its grading window after already knowing the outcomes — and then had to reckon with what that did to ev...
Read itThe Grading Window We Kept Moving
PlayerGem's grading window for one basketball rating shifted four times before we admitted we were protecting a numbe...
Read itWhat We Do With a Rating We Can No Longer Grade
PlayerGem's method for handling ratings that can't be graded after the fact — and why unresolvable records are the mo...
Read itWhen Instinct Outscored the Corpus
PlayerGem's grading desk on the week a colleague's unaided instinct produced better-calibrated ratings than the shop'...
Read itShrinking Our Own Confidence on Principle
PlayerGem applied shrinkage to its own stated confidence levels and found the shop had been systematically overclaimi...
Read itThe Grading Window Has to Be Fixed in Advance
PlayerGem's grading process only works if the window closes before we see the outcome — here is what happened when we...
Read itOur Most Accurate Month Was Our Worst Result
Our best hit rate ever came from a month when the pipeline was broken. It took a fortnight to work out why the number was good.
Read articleAccuracy and Calibration Are Different Things
Being right often is one property. Knowing how often you will be right is a different one, and it is the one that makes a shop useful.
Read articlePublishing the Error Rate Changed How We Argued
The decision to make our own miss rate public was meant to be about honesty. Its largest effect was on how disagreements inside the shop got settled.
Read articleAll Keeping Articles
- The Grading Window We Trusted Too Long
- The Window We Trusted Because We Set It First
- Grading the Window Instead of the Rating
- The Grading Window We Opened Early
- A Column That Passed Every Check and Said Nothing
- The Grading Window We Agreed On and Ignored
- The Grading Column We Forgot to Check
- The Grading Column We Added to Feel Thorough
- Grading a Window We Closed Too Early
- The Grading Window We Set After Seeing Results
- The Grading Window We Kept Moving
- What We Do With a Rating We Can No Longer Grade
- When Instinct Outscored the Corpus
- Shrinking Our Own Confidence on Principle
- The Grading Window Has to Be Fixed in Advance
- Our Most Accurate Month Was Our Worst Result
- Accuracy and Calibration Are Different Things
- Publishing the Error Rate Changed How We Argued