How the desk works

The full methodology — what gets scored, what gets filtered, what gets assumed, and the rules the record itself obeys. If something here is vague, that's a bug; tell us.

The morning read: five ICT signals

Every weekday the desk scores each instrument's daily bias from five signals. The score lands between roughly −5 and +5; the sign is the lean, the magnitude is the conviction.

structure — higher highs/higher lows (bullish) vs lower highs/lower lows (bearish): the trend of the swing points

BOS — break of structure: a close beyond the most recent swing, confirming intent

premium/discount — where price sits in its recent dealing range; discount favours longs, premium favours shorts (this is entry quality, not direction)

displacement — an unusually large, energetic candle showing institutional intent

PD-break — a close beyond the previous day's high or low: liquidity taken

The cipher on each bias card — e.g. 2/2 | +structure +discount — reads: factors aligned / factors checked, then each signal with its direction. A minus sign means the signal argues against the lean.

The setups

Sweep + reclaim (reversion): price raids the previous day's high or low — taking the liquidity resting there — then closes back inside with conviction (a close in the top/bottom third of the bar). Entry is the reclaim close.
FVG continuation: a displacement leg leaves a fair-value gap; the desk enters the retrace into that gap, in the direction of the displacement.

Both are gated by the daily bias and sized to 1R targets, flat by end of day. A setup is tagged HIGH conviction when its entry thrust — the trigger bar's range divided by the average true range — is at least 1.2. When the 2-position correlation cap forces a choice, HIGH setups go first.

The filters (when the desk refuses to trade)

Inter-market regime: computed daily from the dollar (DXY direction) and volatility (VIX vs its 20-day average), always from the prior session's close — no look-ahead. Dollar soft + vol calm = risk-on (longs favoured); dollar bid + vol hot = risk-off. Index trades that fight the regime are skipped.
News stand-aside: no entries within ±30 minutes of high-impact USD events (FOMC, CPI, NFP and friends).
Correlation cap: at most 2 concurrent positions across the four indices — they're ~0.8 correlated, so more isn't diversification, it's leverage.

Execution assumptions — read this one skeptically

The forward record is a model record: real, timestamped signals resolved against real bars, but with assumed fills. Treat every number on the terminal with these assumptions attached.

Fills: entries are assumed filled at the 15-minute trigger bar's close. No slippage is modelled. Stops assume fill at the stop price.
Data: free, delayed public data (the same feed you can check). Bars can reach the desk minutes after they close; the alert pipeline adds its own minutes. Nothing here is co-located, and the terminal labels data ages honestly.
Costs: the prop-firm simulation deducts a flat $8 per trade. That is roughly right for 1–3 micro contracts and optimistic for larger sizes — a 13-micro alert costs more like $25+ in commissions plus spread. A per-contract cost model is on the roadmap; until then, assume the sim's costs are kind.
Resolution order: when a bar touches both stop and target, the desk books the loss. Ambiguity always resolves against the record.
End of day: anything still open settles FLAT at the session's last close, positive or negative.

The eval attempt

The live alerts run through Apex $50k EOD account rules as a dry-run: start $50,000, profit target $53,000, trailing drawdown $2,500 ratcheted at each day's close, ratchet locks once the floor reaches $50,100, minimum one trading day. A passing sim is necessary, not sufficient — a live eval with intraday trailing is strictly harder than the EOD-checked version modelled here.

The record's own rules

Alerts are logged at send time, with the Telegram message id kept as an independent timestamp where available.
The forward ledger is append-only and forward-only: every row carries a logged_at stamp, and rows dated before the newest existing row are refused by the writer — adding a new setup type later can extend the record, never rewrite its past.
The ledger is published to the public scoreboard repository, fingerprinted with a SHA-256 hash shown in the terminal footer. If the published copy and your terminal disagree, one of them is lying — that's the point of the hash.
Losses are displayed everywhere wins are. The current drawdown gets a chip, not a cover-up.
No performance marketing until the live record clears 40–60 trades. Until then the product is watching the record build, in public.

Glossary

R: One unit of risk — the distance from entry to stop. +1R means the trade made what it risked. All desk targets are 1R.
PDH / PDL: Previous day's high / low — the nearest pools of resting liquidity, and the levels the sweep setup hunts.
BOS: Break of structure — a close beyond the most recent swing high/low.
FVG: Fair value gap — a three-bar gap left by a fast move; price often retraces into it before continuing.
Displacement: A candle with unusually large range relative to recent volatility — interpreted as institutional intent.
Premium / discount: The upper / lower half of the recent dealing range. Buy discounts, sell premiums.
Killzone: The desk's trading window: 07:00–13:00 ET, covering the New York morning. Scans run hourly inside it.
Combine / eval: A prop firm's test account. Pass it (hit the target without breaching the trailing drawdown) and they fund you.
EOD trailing drawdown: A failure line that ratchets up with your end-of-day equity peaks. Fall to it, even once, and the eval fails.
Expectancy: Average R per trade across all outcomes — the number that tells you whether the edge is real once streaks wash out.

← back to the terminal