How an edge earns its place here.
Most signal services publish results. We publish the process that produces them — because a win rate without its methodology is just a screenshot. Every rule below is checkable against the catalog, research log, and ledger.
Backtest → walk-forward → permutation → live shadow
Backtest, cost-adjusted
Every candidate edge is tested on up to 7 years of 1-minute MNQ data (171,000+ bars on the dense era). All results are net of costs — we charge every simulated trade 2 points for commission and slippage. Fills on higher-timeframe setups are simulated bar-by-bar on the 1-minute path, not on the bar close, so intrabar stop-outs count against us.
Walk-forward on held-out data
The data is split into sequential regime windows (typically 6). An edge must stay profitable in windows it was never fitted on. In-sample-only results are never published as validation — when you see 'walk-forward' on a premium card, it means every fold was out-of-sample.
Permutation controls
A surviving edge is tested against matched random entries (same days, same session, same exit rules). If random entries perform similarly, the edge is an artifact of market drift, not a real effect — it dies regardless of its win rate. Placebo-level tests killed our own legacy volume-profile edges in July 2026; the retraction is public on the research log.
Live shadow tracking
Validated edges fire in shadow first: real-time signals, tracked to resolution, no live positions and no subscriber alerts until live behavior matches the backtest. The shadow record accrues on the ledger like everything else.
Premium is a claim, not a price
Walk-forward validated: held-out test positive, permutation-tested where applicable, live shadow-tracked. This is the only tier we describe as validated.
Backtest-positive with honest in-sample numbers shown; walk-forward audit pending. Explicitly labeled as such on every card.
Small samples, experimental filters, or booster logic. Shown with caution flags because hiding work-in-progress would be its own kind of lie.
Retractions stay public
When a re-audit kills a published number, we don't quietly delete it — we retract it in public. This page once linked to an edge showing 69.5% WR; a five-month re-audit on live bars measured 33.5% for that configuration. The edge was re-scoped, demoted, and the retraction note lives on its catalog card and the research log. An entire family of volume-profile edges was closed the same way after placebo-controlled testing. If a service has never retracted anything, it isn't because they've never been wrong.
The tamper-evident alert log
The oldest trick in this category is posting the alert after the move. Our answer: every published signal is stamped into a hash chain at the moment of emission. Each signal's SHA-256 integrity hash covers its trade-defining fields (setup, symbol, direction, entry, stop, target, emission timestamp) plus the hash of the signal before it. Editing or deleting any historical alert breaks every hash after it — visibly and permanently.
Lines we don't cross
- Every outcome publishes — wins, losses, and expiries. The ledger computes from the database; there is no editorial step.
- No earnings claims. We publish per-trade statistics with sample sizes, never income promises.
- No front-running: the system that emits alerts holds no position ahead of them. Signals are generated and published by the same automated pipeline.
- Cancel anytime — subscriptions are managed by Whop, not by talking to us.
- Not financial advice, and not a substitute for your own risk management. Futures trading involves substantial risk of loss.
- Not affiliated with ICT / Inner Circle Trader or any educator. If a Discord claims to be us and isn't linked from this site, it isn't us.