Prospective ledger
The implemented ledger stores immutable, content-addressed pre-event forecasts. Corrections append new outcomes and scores rather than overwrite prior evidence. Implementation is not proof of deployment or forecast accuracy.
Know what backs the call.
A forecast helps you weigh a move. It cannot promise the result. Check the data, how recent it is, and the assumptions before acting.
We are still collecting forecasts made before games are played. Pick Pilot is not publishing an accuracy or calibration claim yet.
We compare those forecasts with what actually happened, keep corrections visible, and separate a good decision from a lucky result. The full methods and publication requirements are below.
How we record a forecast
A forecast counts only when its receipt was recorded before the factual event, with a database-issued timestamp, model version, input hash, and exact source snapshots. Recomputed history is labelled reconstruction and never promoted to prospective evidence.
Internal factual scoring
Win, tie, and loss remain separate. Internal scoring uses Brier and log loss for probabilities, unit-labelled absolute error for numeric targets, and coverage only for predictive intervals. These are methods, not published results.
No composite score
Internal evaluation distinguishes eligible, matured, scored, void, missing, and corrected rows by source, league format, phase, surface, model, metric, and stage. No measured public scorecard is published until the gates pass.
What is not evidence
Monte Carlo sampling-error bands are not predictive intervals. Producer self-confidence is not empirical calibration. AI narration quality is evaluated separately from forecast accuracy. A licensed-source credential does not activate a feed without accepted legal, security, and product reviews.
Current product facts
- Pick Pilot is a fantasy-basketball decision layer in web beta, not GA. It does not host leagues or submit provider transactions; you complete moves on your league platform.
- Yahoo league access is blocked pending Fantasy API approval. OAuth sign-in alone does not grant league access; enabling Yahoo also requires official attribution assets and release checks. Sleeper is conditional beta for compatible NBA leagues. Fantrax is private-beta/request access with no guaranteed acceptance. ESPN is planned, with no public connector.
- Rolling form uses 7/15/30 played-game windows, not calendar days. No licensed rest-of-season projection feed or active dynasty age curves are offered. Bring-your-own rankings, tiers, ADP, and supported user-supplied projection imports are implemented, not a promise of automatic external imports or production availability. Upload only data you have the right to use.
- Free is $0 for a one-league starting experience. Pro reservations are $9/month or $84/year; Elite reservations are $19/month or $180/year. Reservations record intent and price, not a charge, subscription, automatic trial, or access guarantee. Billing remains reservation-only pending Vandly acceptance.
- Native mobile is not publicly available. AI is budgeted rather than unlimited, and exports are limited to supported surfaces and plans. Forecast evidence is collecting; no measured accuracy or calibration results are published.
Public reference directory
These canonical production links describe the product and its limits. The same facts and links are in llms.txt, an optional experimental discovery file. It does not guarantee crawler use, search rankings, or AI citations.