INSTITUTIONAL WHITEPAPER 7 MIN READ RELEASED: 2026-08-15

Production Architecture & Empirical Benchmarks

Full specifications of the deployed breakout engine: 127M real broker ticks, a 1.52 out-of-sample profit factor, +0.297R gated expectancy, and 0% ruin across 10,000 paths.

PROVENANCE Walk-Forward Optimization RecordReal Ticks Benchmark Registry

EXECUTIVE SUMMARY · VERIFIED BENCHMARKS

127M: Real broker ticks. Jan 2023 to Jun 2026, live spread and slippage modeled.
29 / 32: Out-of-sample folds positive. Rolling 9-month windows, strict forward embargo.
+0.297R: Gated expectancy. Two-stage meta-labeling lifts the +0.088R raw pool by +237% on identical folds.
40.0%: Trailing brake, set by a 1,500-trial sweep. Without it, every long-horizon path ruins.

1. Verified System Specifications

The platform does not rely on subjective chart reading or backward-fitted bar-close simulations. Every benchmark below reflects automated execution across real broker tick data with spread and execution slippage accounted for.

Specification	Certified value	Basis
Asset Universe	XAUUSD, XAGUSD, GBPJPY, BTCUSD	Precious metals, major G10 cross, liquid cryptocurrency
Evaluation Dataset	127 Million Real Broker Ticks	Tick-by-tick real bid/ask liquidity (Jan 2023 - Jun 2026)
Out-of-Sample Profit Factor	1.52 (1.14 Un-Gated)	Meta-labeling gate lifts the raw breakout pool expectancy by +237%
Gated Net Expectancy	+0.297R per trade	Verified against the +0.088R un-gated pool on the same folds
Out-of-Sample Fold Reliability	29 / 32 Folds Positive	All 2025/26 walk-forward folds green on live-venue real ticks
BookGovernor Stack	4 Layers + 40% Trailing Brake	ATR vol-target, Kelly stream weights, drawdown governor, circuit brake; sweep-certified optimum
Position Concurrency	Max 2 per Sector · 8% Cap	Sector heat caps plus MFE breakeven ratchet (+2.0R excursion to BE+0.2R)
Monte Carlo Ruin Probability	0.00% (10,000 Bootstrap Runs)	The zero-ruin gate blocks every deployment candidate before it ships

2. Walk-Forward Out-of-Sample Verification

Standard backtests in the retail trading industry suffer from fatal survivorship and lookahead bias. The model runs an embargoed walk-forward protocol instead: training on rolling 9-month windows and evaluating strictly on out-of-sample forward periods that the fit never saw.

The gated book is compared against the raw breakout pool it was filtered from, on the identical folds and the same tick data. That comparison is the whole claim the gate makes, so the two are always published together.

Source: System Genesis, Section 4 (two-stage meta-labeling and the +237% amplification)

1.52 Gated profit factor 1.14 on the raw breakout pool, same folds

+0.297R Gated expectancy +0.088R un-gated on identical folds

29 / 32 Folds positive Rolling 9-month windows, forward embargo

40.1% Gated win rate 34.4% on the raw breakout pool

3. The 4-Layer BookGovernor Risk Stack

Alpha without risk governance is financial suicide. The deployed BookGovernor normalizes risk on every trade, weights each stream by verified out-of-sample contribution, de-risks continuously as drawdown deepens, and halts compounding at a sweep-certified brake line.

Layer	Scope	Mechanism	Effect
Layer 1 · Vol target	Per trade	ATR-normalized sizing gives every signal identical percentage risk regardless of instrument volatility.	Gold, silver, FX and crypto fire at the same controlled heat
Layer 2 · Kelly weights	Per stream	Each strategy stream is sized by its walk-forward Kelly contribution rather than flat.	3 to 7 points of compounded return over flat sizing
Layer 3 · Drawdown governor	Continuous	Sizing decays continuously as drawdown deepens - no steps, no cliff, no liquidation trigger.	f · max(f_floor, 1 - dd / dd_cap)
Layer 4 · Trailing brake	40% certified	Stops ratchet to breakeven after +2.0R excursions; sector heat caps at 2 positions and 8% open risk.	Set by a 1,500-trial sweep; 35-45% is the certified band

4. Monte Carlo Ruin Probability: 0.00%

Every deployment candidate must pass a bootstrap ruin gate before it ships: 10,000 randomized-order bootstrap equity paths are simulated with the full governor stack active, and any candidate that produces a single ruin event is rejected.

The deployed book recorded zero ruin events across all 10,000 paths. The same zero-ruin requirement blocks every future configuration change, which is why the brake line is a gate rather than a setting.

10,000 Bootstrap equity paths Randomized trade order, full governor stack active

0.00% Ruin probability Zero ruin events across every path

40.0% Certified brake line The sweep optimum the gate is armed to

1 Gate for every change No configuration ships without passing it

These figures describe the deployed book, not a candidate configuration.

No parameter above is a backtest optimum chosen after the fact: each one was fixed in advance, run forward on the next out-of-sample fold, and kept only when it survived.

See the seat ladder Inspect the tape

Where this architecture is used

This paper is the specification sheet. These pages show the book it produced.