Five portfolio workstreams, one evidence standard

An August 2026 public update on PBZ, Deriv licensed-strategy research, the native 12-profile book, FTMO demo observation, and Darwinex reconstruction.

20Deriv demo configurations

Licensed-strategy research book; composition withheld.

12native validation profiles

Diversified Engine research book; no live authority.

213reused broker-realistic ledgers

Seventy-one Darwinex components across three periods.

< 0.8%maximum reconstruction difference

Calibration against the deployed 58-component book.

The evidence-led verdict

MSRLab now has several portfolios at deliberately different evidence states. Demo observation, historical reconstruction, and live deployment are reported separately. No new book is presented as validated merely because it has been packaged or started on a demo account.

What the work established.

Each section states the public conclusion while keeping reconstructive details out of the publication pipeline.

PBZ · from paused to prospective evidence

PBZ has moved into demo forward testing and lifecycle data gathering under a newly built safety and observation layer.

Five native components compiled cleanly, thirteen automated checks passed, and controlled tester smoke runs covered more than 22,000 generated ticks.
Durable restart state, causal lifecycle logging, concentration controls, and demo-only authorization were verified.
The package proves research readiness—not challenger profitability or live-capital readiness.

Deriv · a new licensed-strategy demo book

MSRLab continued ingesting, normalizing, validating, and portfolio-engineering licensed strategy configurations and selected a new 20-configuration Deriv research book.

Intake verifies provenance, schema compatibility, duplicates, and exact engine behavior before economic testing.
Broker-realistic periods, costs, stability, concentration, and portfolio interaction determine whether a configuration remains eligible.
The book is now in demo testing; its contents and construction evidence remain confidential.

Diversified Engine · a 12-profile native book

The general MSRLab engine retains a 12-profile validation portfolio while the search for a genuinely distinct additional edge continues.

New functionality and many candidate families were tested without changing the book when economic admission evidence failed.
Dependence is assessed through exposure, timing, drawdown, regime, and concentration—not ordinary correlation alone.
No live approval follows from the current research package or historical evidence.

Prop-funding · FTMO 2-Step demo observation

The independent prop-funding program has resumed and one governed candidate is now being observed in a two-stage demo challenge environment.

Official macro sources and many price-based families were screened; extensive negative findings remain part of the record.
The candidate is judged on survival, loss limits, execution, and evidence consistency—not a lucky fast pass.
Forward reconciliation is still required; there is no paid-account or live-capital authority.

Darwinex · reconstruct before replacing

A 71-component candidate universe was revalidated and then used to reconstruct the economics and concentration of the deployed 58-component licensed-strategy book.

Seventy-one components across three periods produced 213 reusable broker-realistic ledgers.
The offline reconstruction matched exact deployed-book net results to within 0.8% across the tested periods.
Redundant exposure and position-stacking risks were identified; compact alternatives remain shadow candidates and the deployed 58 was left unchanged.

Why the evidence states differ

A portfolio can be technically packaged, historically attractive, or operationally observable without being ready for capital.

Backtest and reconstruction answer historical questions; demo and shadow evidence answer operational and prospective questions.
Live monitoring compares realized activity, risk, concentration, and outcomes with a frozen expectation envelope.
Replacement, scaling, and deployment remain explicit human decisions rather than automated consequences of a score.

PUBLIC RESEARCH CONCLUSION

Progress is a decision,
not just a curve.

The common achievement is a repeatable path from intake to evidence: authenticate, normalize, test, reconstruct, challenge concentration, package safely, observe prospectively, and preserve the decision. The current decisions are intentionally mixed—continue, observe, hold unchanged, or wait for evidence.

Next evidence gate

Accumulate prospective demo and shadow evidence, reconcile it against frozen expectations, and make any capital or replacement decision only after the relevant monitoring gate has matured.