Research · validation · deployment

Promotion is earned, not assumed.

Multi-regime testing, multiple in-sample/out-of-sample evaluations, walk-forward analysis, stress testing, forward observation. Governance continues after release. Every live system survived this exact pipeline.

Tick backtestingIS/OOSWalk-forwardStress + forward testing
TPC / PSeven-stage promotion path
No skipped gates
RESEARCH ENVIRONMENT01–05 · CANDIDATE QUALIFICATION
  1. 01Observe
  2. 02Classify
  3. 03Challenge
  4. 04Qualify
  5. 05Harden
LIVE SUPERVISION06–07 · GOVERNED RELEASE
  1. 06Isolate
  2. 07Govern
Candidate logic Governed release

The Gauntlet · seven governed stages

A defined control at every handoff.

From market intelligence to live supervision, exact thresholds, parameters, scoring, and implementation details stay proprietary.

Validation depth

Robustness must persist beyond the development window.

Each layer attacks a different model risk; none substitutes for the sequence.

01

Tick-level backtesting

Execution behavior at the resolution each candidate requires, across market regimes.

Why it matters · Engineering fidelity across regimes
02

Multiple IS/OOS boundaries

Development-versus-holdout separation repeated across market history.

Why it matters · Generalization beyond fitted data
03

Walk-forward analysis

The research window moves through time; results must stay stable.

Why it matters · Stability through changing conditions
04

Stress + forward testing

Adverse assumptions first, then observed forward performance before live promotion.

Why it matters · Resilience under adverse assumptions

Release discipline

Research output does not enter production directly.

Approved logic passes review, compilation, release checks, and controlled deployment; only the versioned EX5 crosses the boundary. The validated library runs deeper than the roster. Only the strongest current bots hold capital.

01MQL5Reviewed source
02BUILDCompilation gate
03EX5Versioned artifact
04VPSControlled promotion
Capitalgoverned allocation
Promote Monitor Reduce Retire

Controlled rotation · the living ecosystem

Adaptation happens at the portfolio layer.

Live monitoring, regime alignment, and validation history drive a governed loop: capital concentrates where fit is strongest; the rest are reduced, quarantined, or retired.

Rotation never lets a live bot rewrite its own logic. Every promoted candidate already cleared validation and release.

View the current algo-bot roster

Survivors of the gauntlet

Meet the roster this process produced.

View the current roster