RESEARCH OPERATING SYSTEM

A strategy is not a backtest.

It is a hypothesis, an implementation, a chain of evidence, a role in a portfolio, and an ongoing comparison between what was expected and what actually happened.

THE MSRLAB RESEARCH OS

Every gate can return the work upstream. Rejection is information, not a broken process.

01Question

02Specify

03Baseline

04Challenge

05Portfolio

06Promote

07Observe

IDEA SPACE↻ FEEDBACK / REFRAME / REBUILDLIVE EVIDENCE

01—07

Seven gates between an idea and capital.

The work moves forward only when the question, implementation and evidence remain coherent.

Historical signal

Does the behaviour exist at all?

Temporal survival

Does it persist across rolling periods?

Implementation stress

Does it survive costs, slippage and broker reality?

Portfolio value

Does it add a distinct return driver?

OUTPUTSUPPORTED / WATCH / REJECT

Question before code

State the market behavior, why it might persist, what would invalidate it, and what evidence would change our mind.

Implementation with provenance

Record source, assumptions, data, broker, indicator version, costs, and every material rule.

Baseline before tuning

Establish whether the idea has any signal before searching parameters. Beautiful optimization surfaces cannot rescue an incoherent premise.

Challenge the result

Use rolling periods, untouched data where possible, topology, stress, costs, Monte Carlo, and broker-native MT5 real ticks.

Portfolio test

Ask what return driver the candidate adds, what it duplicates, and how it changes drawdown, concentration, and latent exposure.

Promotion with an expectation

Store expected drawdown, frequency, holding time, costs, correlation, and regime notes before capital is involved.

Observe and govern

Compare live behavior with expectation. Agents diagnose and propose; research validates; the human Research Director approves material changes.

Infrastructure is part of the method.

Reproducibility depends on more than a testing script.

01VPS research and execution environments

02MetaQuotes MetaTrader 5 platform

03MQL5 strategy and indicator engineering

04Python backtesting and analysis

05Codex, Claude and ChatGPT research assistance

06Broker-specific MT5 research lanes

07Reusable exporters and indicator buffers

08Real-tick validation runners

09Shared portfolio risk governors

010Account-history attribution

011Cron-driven live monitoring

012Expected-vs-observed decision support

A living portfolio needs a nervous system.

Cron-driven monitoring turns live behaviour into decisions while parallel research keeps the edge pipeline replenished.

EXPECTED ENVELOPELIVE TRACE

NORMALWATCHINVESTIGATEACT

Observe

Scheduled cron jobs and on-demand pulls collect MT5 history, account state and strategy attribution.

Compare

Live frequency, execution, drawdown and sleeve contribution are compared with historical expectation.

Decide

The monitoring layer produces evidence for keep, reduce, pause, retire or investigate recommendations.

Replenish

Edges are not assumed to last forever. Degradation triggers review while the lab continues developing independent replacements.

What we are suspicious of

Parameter fishing · repeatedly inspected OOS · buffer leakage · huge winner searches · hidden recovery tails · small samples · PF-only selection · correlated strategy collections.