Building Legacy Through Systematic Arbitrage.

Raw expectancy means nothing if it cannot survive live market friction, platform execution costs, and statutory tax obligations. The XTNL pipeline applies an adversarial filter stack to the theoretical dataset before computing any metrics.

The Haircut filter compresses winners and expands losers to model spread widening. The Operator Efficiency filter caps theoretical yield at 85%. The Toxic Streak filter compounds psychological tilt penalties on each consecutive loss. All projected capital growth is taxed annually at the ATO's maximum rate of 47%. The 0.982 R expectancy reported by the system survives this entire matrix — it is a pessimistic floor, not an optimistic ceiling.

Statistical significance testing across N=308 non-overlapping observations. The null hypothesis — that returns are random noise — is rejected with a joint probability of approximately p = 3.48 × 10⁻³³. The System Quality Number (SQN) of 4.253 on the primary session-filtered core (N=106) places the system in the upper tier of quantitative systems by the Van Tharp classification.

A 95% confidence interval lower bound of 0.529 R means that even at the pessimistic extreme of the statistical estimate, the system is expected to generate positive expectancy. The t-statistic of 4.251 confirms this is not attributable to random variance.

Walk-Forward Optimisation (WFO) is the standard methodology for detecting curve-fitting: the model is trained on historical data, then tested on the immediately following period it has never seen. If the edge is real, out-of-sample performance remains consistent. If it is curve-fitted, OOS performance collapses.

The XTNL WFO engine runs an expanding-window validation across 4 folds. SESSION_FILTERED yields aggregate OOS SQN 1.787 and OOS expectancy 0.904 R — STABLE. Fold 3 showed 77.5% degradation, which is precisely why the 19:00 AEST temporal cluster was identified as a structural toxicity and permanently excised. The system does not ignore its own failures — it routes them into architectural improvements.

Any specific edge will eventually decay. Market inefficiencies close as participants adapt. The XTNL architecture resolves this through strict structural decoupling between the Tangible Asset Chassis and the Intellectual Property Payload.

The chassis — the analytics pipeline, risk engine, WFO validator, execution firmware, and capital allocation logic — is permanently owned. The IP (the specific edge definition) is treated as a replaceable payload. If WFO signals critical edge decay, the pipeline halts deployment and the chassis remains intact, ready to validate and load a successor edge. The business never depends on any single market observation surviving indefinitely.