pippen¶
Reliability-adjusted NBA player impact estimates, with calibrated uncertainty.
Pre-release
The data layer and the metric are under construction. No results have been published yet, and nothing here is stable.
The idea in one paragraph¶
Public NBA impact metrics disagree with one another, and none reports how much to trust any single number. This project treats those metrics as noisy measurements of one quantity nobody observes directly: a player's true contribution to point differential. That turns player evaluation into a measurement-error problem, which is a solved area of statistics. Each metric's reliability is measured rather than asserted, the metrics are combined by inverse-variance weighting, and every player gets an interval rather than a bare number.
The claim under test¶
Does PIPPEN predict next-season team net rating better than any single input metric does, out of sample?
If the answer turns out to be no, that will be published here. A falsifiable claim stated before the experiment is what separates research from a demo.
Where to start¶
Licensing in one line¶
Code is Apache-2.0, published data is CC BY 4.0, and nothing paywalled ever enters a release. The full position is on the data licensing page.