Autonomous Execution
16 Signals · 13 Gates · Honest Ledger
Autonomous ES/MES execution on a live market feed — 16 signal modules, 13 independent risk gates, multi-timeframe consensus, and direction-lock so the engine does not fight itself. Manual signals available separately at $100/mo.
/link your@email.com to unlock subscriber channels.
Walk through the module stack, risk gatekeepers, and session logging model — the same architecture you deploy on Sovereign. The 6-month direction-locked backtest below is the full published replay — 790 trades, Oct 2025 through Apr 2026.
Full direction-locked replay on ES/MES — Oct 2025 through Apr 2026. Every fill, day-by-day expandable ledger.
View backtest →2-of-8 TF consensus, direction lock, HMM regime, VPIN conviction, alpha monitor, and the full gatekeeper stack.
Signal stack & gates →Layer-by-layer breakdown of how each module votes, how consensus forms, and what gets blocked before execution.
Module deep dive →Every fill logged to session vault on your deployment — entry, exit, P&L, running equity. Public site shows the UI shell only.
Engine showcase →The engine writes session logs to operator storage on hardware you control — not to this public marketing site. The expandable day log below is an empty UI shell showing how reviews work after you deploy.
No session data on the public site. Logs live in your deployment vault only.
Starter locks launch access at $500 one-time preorder. Pro is full autonomous execution at $499/mo when you're ready to run live.
Manual signals only? Signals — $100/mo →
Aggregate ledger from the full 6-month direction-locked backtest on real ES/MES market data — Oct 2025 through Apr 2026. Raw signals filtered through multi-TF consensus and direction lock. No curve-fit, no survivorship bias, no cherry-picked windows.
Figures verified against raw tick data in Python backtests. Every fill, commission, and slippage estimate is preserved. Want the CSV + equity curve? Grab the Backtest Report ($150) or preorder Trading Engine Starter ($500) — report included.
Most trading bots fail because they fire on noise. Black Label Bots is different. Every single trade passes through 13 independent risk gatekeepers and a 16-module signal stack before execution — no emotional decisions, and a system that adapts faster than any human trader.
Cut false entries dramatically. A trade fires when at least 2 of 8 timeframes confirm.
Eight independent timeframe bots run simultaneously. A trade fires when at least 2 of those 8 timeframes align across a 5-voter signal stack: StepGMA, EMA 9/21, CVD, VWAP, and Multi-TF agreement. Most retail bots fire on a single timeframe. Black Label requires 2-of-8 timeframe agreement before any capital is deployed.
Stop losing $200+ per day to hedging fees. One direction, full conviction.
When any position is open, conflicting entries are blocked at the gate level. This eliminates the self-canceling commissions that quietly destroy retail bot returns. One direction, full conviction, until the trade is resolved.
Trade the regime, not the noise. Automatically switches strategies.
A 3-state Hidden Markov Model classifies every bar into LOW_VOL_TREND, HIGH_VOL_BREAKOUT, or MEAN_REVERTING. Sub-strategies are gated by regime — Sniper is blocked in chop, Reversal is blocked in trends. The engine adapts instead of forcing.
Follow the smart money. Boost signals when institutions move, suppress when retail does.
Volume-Synchronized Probability of Informed Trading (Easley, López de Prado, O'Hara 2012) measures whether current flow is driven by informed institutions or uninformed retail. High VPIN boosts signal confidence. Low VPIN suppresses entries. You only trade when the smart money is moving.
Dead signals die fast. The engine auto-disables broken strategies in real time.
Rolling 20-trade expectancy is tracked per signal type. If any signal's edge drops below −0.5R, the monitor disables it automatically. After 50 bars and a regime change, the signal gets a fresh trial window. The engine detects and routes around its own decay — something no human trader can do consistently.
Lock profits every 1-2 points. Let winners run, protect against reversals.
As each trade moves in your favor, the stop ratchets forward through six locked tiers — +1pt protected at +3, +2 at +4, +3 at +5, +5 at +7, +7 at +10, and the full 25-point target at hit. Winners convert to locked capital instead of reversing to flat.
The engine analyzes market flow through 16 independent signal modules, organized by function.
After a signal fires, 13 risk gatekeepers validate execution. Any one red light blocks the trade.
Ready to see how this executes live?
See the full 6-month backtest →The logging shell shows entry, exit, P&L, and running equity per day. Your member deployment writes to vault on your hardware — the public storefront shows layout only, not operator fills.
No session data on the public site. Logs live in your deployment vault only.
Every entry is the result of an alignment check across eight independent intelligence layers. If any one of them disagrees, the trade does not fire.
Gradient momentum acceleration. Measures the rate-of-change of a step-filtered moving average to detect institutional order-flow shifts before they appear in price.
The engine requires StepGMA to be aligned with the intended direction on the active timeframe and on the next timeframe up.
Short-vs-medium term trend alignment. Not a crossover signal — a directional gate. The 9 EMA must be above the 21 EMA for a long and below for a short.
The engine checks 9/21 on the active TF and 21/50 on the timeframe above. If they disagree, the vote fails.
Tick-level buy vs sell pressure computed from the raw tick feed. The running delta tells you who is pressing the tape.
For a long, CVD must be positive and rising. A bearish CVD with a rising price is absorption — the engine blocks longs even if the chart looks strong.
Above VWAP is a long bias, below is short. The engine uses session VWAP for intraday and anchored VWAP from prior week's open for longer timeframes.
VWAP bands (1σ, 2σ) serve as dynamic S/R. A long above the upper band requires HMM to be in a trending regime.
All four voters must align on the active TF and the next TF up. Cross-timeframe agreement is the primary filter against single-TF noise.
Most retail bots fire on a single timeframe. Black Label requires at least 2 of 8 timeframes to agree before any bot acts.
Measures whether flow is driven by informed institutions or uninformed retail. 500-contract buckets, 50-bucket rolling average.
High VPIN (>0.70) = 1.25× multiplier. Spike above +0.15 = 1.50×. Below 0.50 = weighted down to 0.85×.
Three hidden states: LOW_VOL_TREND, HIGH_VOL_BREAKOUT, MEAN_REVERTING. Features: log return, volume ratio, normalized range. Auto-retrains every 200 bars.
The engine gates which sub-strategies fire based on regime. Falls back to rule-based classifier if the library is unavailable.
Net gamma per strike from the SPY chain. Positive GEX walls = market makers sell rallies (tighten targets). Negative GEX voids = market makers chase (extend targets).
Refreshes every 5 minutes on a background thread. Strikes above $5M absolute GEX are mapped as significant levels.
Detects large resting orders absorbing market-order flow without letting price through. 500+ contracts in a 30-second window.
Absorption at the ask = bullish. Absorption at the bid = bearish. Signals decay after 5 minutes.
Rolling 20-trade expectancy per signal type. If expectancy drops below −0.5R, the monitor disables that signal and fires an alert.
After 50 bars + regime change, the signal gets a fresh trial window. The engine cannot get stuck firing a broken signal.