Adaptive Weighted Algo — Full Guide

Overview

Adaptive Weighted Algo is an ensemble trading system built on 8 specialized expert modules, each representing a distinct market analytical approach. Rather than using a fixed-weight combination, it employs a Multiplicative Weights Update (MWU) algorithm to dynamically adjust how much each expert's opinion counts based on recent performance. The result is a self-adapting consensus signal that reflects which analytical approaches are currently working.

Settings Reference

The 8 Expert Modules

Each module contributes a bullish or bearish vote. The final signal reflects the weighted consensus:

1. Trend Expert: Multi-period trend direction and strength analysis
2. Momentum Expert: Rate of change and momentum oscillator signals
3. Mean Reversion Expert: Deviation from statistical norms — identifies overextended moves
4. Volatility Expert: Volatility state and regime detection
5. Volume Expert: Volume confirmation and accumulation/distribution analysis
6. Pattern Expert: Candlestick and chart pattern recognition
7. Structure Expert: Support/resistance and market structure alignment
8. Sentiment Expert: Intraday sentiment indicators and participation analysis

How MWU Works (Plain English)

The Multiplicative Weights Update algorithm works like a voting system where past accuracy determines future voting power. An expert module that has been frequently correct in recent bars gets more voting weight. An expert that has been frequently wrong gets its weight reduced. This means the algorithm automatically de-emphasizes approaches that are not working in the current market regime (e.g., mean reversion in a strong trending market) and amplifies those that are.

Visual Elements

• Buy/Sell Signals: Plotted on the chart with the consensus score displayed
• Consensus Strength Indicator: Shows the count of experts in agreement on each signal
• Expert Weight Panel (if visible): Shows the current weighting of each expert — useful for understanding which analytical lens the algorithm is currently emphasizing

Recommended Timeframes

• Day trading: 1-min to 5-min for execution timing
• Swing trading: 1H to 4H for confirmation
• Minimum recommended: 1-min. Below this, noise dominates and expert accuracy degrades significantly.

Tips & Best Practices

• High consensus (7–8 experts agree) = strongest signals. When nearly every expert module is aligned, it reflects a multi-dimensional confluence that the MWU algorithm has determined is currently reliable.
• Low consensus (4–5 split) = stand aside or reduce size. A split vote means different analytical approaches disagree. This typically reflects a transitional or choppy market condition where the edge is unclear.
• The system adapts automatically — resist the urge to manually override its weights. If the algorithm is currently de-weighting the trend expert because trending strategies have been losing recently, that is information. Don't fight the adaptation.

Common Mistakes

• Overriding the algo because 'it doesn't feel right.' The MWU algorithm has no emotional bias. If you feel compelled to override its consensus, ask yourself whether your override is based on additional objective data or on confirmation bias.
• Using it on timeframes below 1-min where noise dominates. The expert modules require a sufficient number of bars to establish reliable performance estimates. Below 1-minute, the lookback contains too much random noise for the MWU weights to be meaningful.