Neuron Cortex MT5
- Experts
-
Olivier Nomblot
I look at math and computers as ways to eliminate the emotional ups and downs of investing. “I don’t want to have to worry about the market every minute. I want models that will make money “.
WE TRADE THE FUTURE, NOT THE PAST. IMAGINATION TO SOFTWARE . - Version: 6.6
- Updated: 1 October 2026
- Activations: 5
NEURON CORTEX
The swarm decides. The swarm learns. The swarm knows when to stay out.
Most neural EAs sell you one network and call it artificial intelligence. Neuron Cortex runs five independent recurrent networks in parallel — a swarm — and refuses to touch the market unless four of them agree. One net is an opinion. Five nets in agreement is a signal.
This is the distilled product of a decade of live trading and system building by an author who has held the #1 global ranking on this marketplace three times. Every mechanism inside it exists because a live account paid for the lesson.
THE ARCHITECTURE
- Five-member recurrent ensemble, each with its own memory state, reading a rolling window of market structure. Entry requires four-of-five consensus and a confidence threshold.
- Outcome-based training. The swarm is trained on whether a real trade would have resolved profitably — never on raw price movement. Every lesson is a settled trade.
- Confidence calibration. Predictions are graded by conviction against realized outcomes. A confidence band that stops delivering loses its authority to trade.
- Abstention gate. When rolling accuracy degrades, Neuron Cortex stops trading entirely and waits. Knowing when not to trade is the whole game.
- Persistent memory. Weights survive restarts, reboots and VPS migrations. Your instance becomes your instance — no other copy will have learned the same lessons.
THE PROTECTION MECHANISM
Neuron Cortex uses an internally-managed adaptive protective level. It is calculated live by the EA and deliberately not visible to the broker — a resting stop is a target, and this system is built to be on the collecting side of a liquidity sweep rather than the paying side.
The distance is not a fixed guess. The EA measures how far its own profitable trades were pushed against before they resolved, and parks the protective level beyond the envelope those winners actually travelled. It calibrates to the instrument and re-calibrates as the tape changes. Levels are persisted per position and survive any restart.
An optional server-side catastrophe stop can be enabled at a far wider distance for unattended operation.
THE RAILS
- No grid. No martingale. No averaging. No hidden recovery mode. Ever.
- Concurrent positions are hard-limited in code. A compiled constant, not a setting. It cannot be raised.
- Position size is hard-capped on a balance ladder. Also compiled. The discipline that makes this system work cannot be undone from the inputs tab.
- Trading stops before the session close. New entries are blocked ahead of the close, then all positions are flattened. No overnight exposure, no weekend gap.
- Independent virtual-equity drawdown kill switch, spread gate, margin-safe order handling, optional time-stop.
EFFICIENCY
Small size, few positions, no leverage headroom required, no floating losses held in hope. Precision over volume — a swarm of small disciplined strikes rather than one heavy bet. It does not need a large account. It needs to be left alone to learn.
RECOMMENDED SETUP
- Symbol: index CFDs (NDX/US100 primary), also runs on metals
- Timeframe: M5
- Account: any, ECN/raw spread preferred
- Scale with additional instances on separate symbols — never with larger lots. The size cap is not a limitation, it is the design.
WHY THIS ONE
You are not buying a curve-fitted backtest. You are buying an adaptive swarm with a built-in conscience — a system that grades its own predictions, disqualifies its own weak signals, protects itself with a level it calculated rather than one it was handed, and shuts itself down when the regime turns against it.
Ten years of iteration. Five brains. One decision.
Neuron Cortex.
