PunditPoorly

Pundits make calls. The market answers eventually. We do the bookkeeping in between — translating every meaningful public-market statement into a synthetic portfolio position, marked to market, with full audit trail.

Each card below is a pundit or pundit-cluster we've catalogued. The methodology is shared across every product: the same conviction rubric, the same deterministic portfolio engine, the same correction layer where the audience help us get attribution right. Different pundits, same receipts.

1. Transcribe. Every episode of every covered show, end to end. Whisper-base on CPU, RSS-sourced audio.
2. Score. Each financial-instrument mention gets scored on two dimensions: claim strength (how forcefully the view was stated) and specificity (how testable the call is). The combination drives position size and which conviction tier the call lands in. We surface a qualitative tier (HIGH / MID / LOW) on every receipt; the numeric weights are operational and not published.
3. Ledger. Predictions flow into a deterministic engine (v5.4 SPY-as-cash + 5 user-selectable methodology presets via BYOB) that opens, sizes, holds, decays, and exits positions over time per the v2.3 rubric. Real prices via yfinance, real holidays, real next-open execution.
4. Audit. Every record has a 🚩 button. Spot a wrong speaker, ticker, direction, sarcasm-missed? Flag it. Corrections get voted on, reviewed by mods, and applied with full edit history.

The methodology page has the full rubric, the anti-gaming rules, the inter-rater reliability targets, and the changelog. Everything is published, including the misses.