Leaderboards
Ranked by verified outcomes on published setups. Nothing here is self-reported — and until publishing opens and records resolve, nothing here is ranked at all.
| Author | Receipt Score | n resolved | Hit rate | Avg R | 90d trend |
|---|---|---|---|---|---|
No ranked records yet. A row appears here when an author reaches n=20 resolved public setups. Publishing opens in invite waves — this table filling up slowly, with real names earning real numbers, is the launch working as designed. | |||||
Boards will break out by market · timeframe class · 90d / 1y / all-time, with minimum-n gates displayed.
The Receipt Score
Computed nightly, from resolved public setups only. Realised R against the stated plan per setup, clamped to a band per timeframe class so scalpers and swing traders are not compared apples to oranges; weighted by recency with a 90-day half-life so nobody coasts on 2019; damped at low volume so one lucky call is a data point, not a legend.
The conviction normaliser is the anti-cherry-picking term: publishing only your surest calls, rarely, scores no better than publishing consistently. The incentive gradient points at showing your actual work.
ReceiptScore = 100 · Σ(wᵢ · outcomeᵢ · conviction_norm) / Σ(wᵢ) outcomeᵢ realised R vs stated plan, clamped per TF class wᵢ recency decay (half-life 90d) × sample damping n < 20 resolved → "Unranked — building a record" inactive 180d → score greys as stale (never deleted)
What boosts do — and don't
Boosts and comments exist because a community needs a way to say good read. They feed the discovery ranking, damped and log-scaled so a boost ring buys almost nothing. They have no path into the Receipt Score or these boards — which also makes engagement fraud a cosmetic attack on a surface that does not matter.
Badges are computed, never purchased and never streak-based: Calibrated, Iron Record (one hundred resolved, never archived one), Pack Author, Desk Pick. No XP, no levels, no login streaks — this is a scoreboard, not a slot machine.