WikiPG Scoring SystemRegime-Aware Adaptive Weights (P4)

Regime-Aware Adaptive Weights

PulseGrid v2.0 introduces a regime detection module that classifies the current market environment and dynamically adjusts the PG-MIM sub-score weights. This addresses a fundamental limitation of static-weight models: the relative importance of event impact, macro conditions, and uncertainty shifts dramatically between calm and turbulent markets.

Theoretical Foundation

The regime-switching approach is grounded in the work of Ang & Timmermann (2012), who demonstrated that asset return dynamics exhibit distinct regimes with different volatility, correlation, and mean-return characteristics. The Two Sigma research team (2021) further validated that machine learning-based regime classification can improve portfolio construction by 5-10% in risk-adjusted terms.

Regime Classification

PulseGrid classifies markets into four regimes based on observable indicators:

RegimeVIX RangeCharacteristicsWeight Emphasis
Crisis> 30Extreme fear, correlation spikes, deep drawdownsEvent Impact (40%), Uncertainty (25%)
High Volatility20-30Elevated uncertainty, sector rotationBalanced with Uncertainty boost (20%)
Steady State12-20Normal conditions, trend-following worksMacro Compatibility (30%), Exposure (25%)
Recovery< 20 + uptrendPost-crisis rebound, mean reversionExposure (30%), Sentiment (25%)

Regime Indicators

The regime detector uses four quantitative signals:

1. VIX Level (weight: 0.35): The CBOE Volatility Index as a direct measure of implied market fear

2. Cross-Asset Correlation (weight: 0.25): When correlations spike above 0.7, it signals systemic risk

3. Drawdown Depth (weight: 0.25): Maximum peak-to-trough decline across major indices

4. Trend Strength (weight: 0.15): 20-day vs. 50-day moving average relationship

Composite Regime Score

The regime is determined by a weighted composite:

> Regime Score = 0.35 VIX_normalized + 0.25 Correlation_spike + 0.25 Drawdown_depth + 0.15 (1 - Trend_strength)

  • Score > 0.7: Crisis
  • Score 0.45-0.7: High Volatility
  • Score 0.2-0.45: Steady State
  • Score < 0.2 with positive trend: Recovery

Adaptive Weight Profiles

Each regime applies a distinct weight vector to the five PG-MIM sub-scores:

Sub-ScoreCrisisHigh VolSteadyRecovery
Event Impact (EI)0.400.300.200.20
Exposure (EX)0.100.150.250.30
Sector Sentiment (SS)0.100.150.150.25
Macro Compatibility (MA)0.150.200.300.15
Uncertainty (U)0.250.200.100.10

Dashboard Integration

The current detected regime is displayed as a badge on the PulseGrid Dashboard with confidence level. Hovering over the badge reveals the regime description and the active weight profile. The regime badge also appears on individual instrument pages.

Limitations

  • Regime transitions are not instantaneous; there can be a 1-3 day lag in detection
  • The VIX is a US-centric indicator and may not fully capture non-US market stress
  • Regime classification uses simplified thresholds rather than Hidden Markov Models (HMM), which could provide smoother transitions
  • Historical regime detection in replay mode uses reconstructed indicators that may differ from real-time values