Meridian 23

Meridian: proprietary neural decision-making for Gold

Meridian is a scalping system built for XAUUSD on 15 minute. It uses a proprietary neural architecture to read several market inputs and decide where a trade should be placed. The architecture is inspired by research around quadratic and higher-order neurons, where a neuron can learn interactions between its inputs instead of only adding them together. This matters because market conditions are rarely independent.

A signal can mean something different depending on volatility, momentum, range and what happened before it. Meridian is designed to learn these relationships directly rather than relying on a long list of fixed rules. The model also looks at the market across multiple time horizons, from short-term movement to broader recent structure. The result is a compact model with only a few parameters.

The model also maintains positive results across different timeframes, providing evidence that its behaviour is not tied exclusively to M15 and suggesting that it captures underlying dynamics of the market rather than isolated timeframe patterns.

How Meridian trades

Meridian does not enter immediately at market. When the model identifies an opportunity, it selects an entry price and places a pending stop order. If price reaches the selected level, the trade is activated. If it does not, the order expires.

At placement, Meridian also places the stop loss and take profit using the current volatility of the market.

Once the trade is placed:

• The system does not average down.
• There is no grid.
• There is no martingale.
• Positions have a defined maximum lifetime.

The strategy therefore always knows where it wants to enter, where it is wrong, and where it wants to exit.

Volatility based decisions

Meridian does not use one fixed distance for every market condition. Entry distance, stop loss and take profit are expressed relative to the current volatility. This allows the same model to work with both quiet and fast moving Gold without manually changing the model's parameters for each regime.

Risk management

The model decides if and where to trade. Account-level settings decide how much to trade and when trading should be blocked.

Input controls include:

• Fixed lot sizing
• Balance-based lot sizing
• Percentage risk per trade
• Daily loss limit
• Daily profit limit
• Maximum drawdown
• Maximum spread
• Trading hours
• Friday protection
• Optional trailing stop
• Optional kill switch

Defaults are the tested configuration. The model layer has no adjustable parameter — the network, its entry logic and its exit rules are fixed, by design, so that what you run is what was measured.