AI Momentum
Machine learning-driven momentum classification with AI confidence scoring
Abstract
The AI Momentum strategy represents the convergence of traditional momentum analysis with modern large language model capabilities. At its foundation, the system uses Gemini 2.5 Flash to perform real-time analysis of market structure, news sentiment, and cross-pair correlation for each monitored instrument. The AI generates a structured assessment including directional bias, confidence score (0-100), key support/resistance levels, and a risk commentary. This AI layer acts as the first filter -- only setups where the AI confidence exceeds the strategy threshold proceed to the technical momentum engine.
The technical momentum engine operates on M15 to H1, combining rate-of-change analysis, volume-weighted momentum scoring, and ADX trend strength filtering to classify the current momentum regime (strong trend, weak trend, ranging, or transitioning). Trades are initiated only when the AI directional bias aligns with a strong or accelerating momentum regime, creating a dual-confirmation system that reduces false signals in choppy or transitioning markets. The result is a strategy that captures genuine momentum moves while avoiding the whipsaws that plague traditional momentum approaches.
Mechanism
AI analysis cycle: Every 15 minutes, submit current market state (OHLCV data, recent price action, open positions, recent trade history) to Gemini 2.5 Flash with a structured prompt. Receive a JSON response containing: directional bias (LONG/SHORT/NEUTRAL), confidence score (0-100), key levels, timeframe assessment, and risk notes.
Confidence filtering: Only proceed with setups where AI confidence >= 70. Setups with confidence 70-80 require full technical confluence. Setups with confidence > 80 may proceed with reduced confluence requirements. Neutral bias signals pause new entries for the pair.
Momentum regime classification: Calculate 14-period and 28-period rate of change on H1. Compute volume-weighted momentum score using tick volume as a proxy. Read ADX(14) to assess trend strength. Classify regime: Strong Trend (ADX > 25, ROC aligned), Weak Trend (ADX 20-25), Ranging (ADX < 20), Transitioning (ADX crossing 20-25 zone).
Entry signal: Generate an entry signal when AI bias is LONG or SHORT, confidence >= threshold, and the momentum regime is classified as Strong Trend or Transitioning-to-Strong. Enter on the close of an M15 candle that confirms the direction with a body-to-range ratio > 60%.
Dynamic position sizing: Adjust lot size based on AI confidence level and current account equity. Higher confidence scores allow marginally larger position sizes within the risk framework limits. All sizing remains subject to the 12-layer risk filter.
Adaptive exit: Primary exit at the AI-identified key level in the profit direction. Trailing stop activates after 1R using an ATR-based ratchet. If AI confidence drops below 50 on a subsequent cycle while the position is open, tighten stops to breakeven regardless of current P&L.
Multi-timeframe confluence
Every entry requires alignment across multiple timeframes. No single timeframe can generate a trade independently.
| Timeframe | Role |
|---|---|
| H4 | Macro trend context for AI analysis input |
| H1 | Momentum regime classification and ADX/ROC analysis |
| M15 | Entry signal generation and confirmation |
| M5 | Entry timing refinement |
Risk profile
64%
Win Rate
1:1.6
Avg R:R
1h 20m
Avg Hold
4
Max Consec. Loss