Skip to main content

Market Microstructure · Machine Learning

When Signals Stop Working: A Dynamic Predictive Regime-Switching Architecture for Cryptocurrency Contagion Alpha

From the documented failure of lagging retail indicators to a fee-aware, microstructure-driven execution architecture built around contagion, calibration, beta dynamics, and live Level 2 order-book states.

Working Paper · Forthcoming
Full article ↗Repository ↗

The research begins with a direct empirical question: when do conventional technical signals such as RSI and Bollinger Bands stop working? Rather than assuming an edge and tuning parameters around it, the paper first subjects classical indicators to purged chronological validation. Their failure becomes the starting point for a different architecture that monitors the institutional plumbing of cryptocurrency markets, separates the Bitcoin trigger from the altcoin target, and changes execution according to the observed market regime.

Research question

Can a cryptocurrency strategy replace lagging price-derived indicators with directly measurable contagion and microstructure variables, then route execution dynamically enough to produce cost-survivable alpha across materially different market regimes?

Method

The empirical programme uses 12,141 SOL/USDT four-hour observations from 2021–2026 with five chronological purged folds, then extends the information set with Bitcoin perpetual-futures funding rates, quote-asset volume velocity, cross-asset beta, Gradient Boosting magnitude forecasts, isotonic probability calibration, and live Level 2 bid-ask imbalance. The final DPRS architecture treats Bitcoin as the contagion trigger and routes capital according to target-asset beta and order-book state rather than imposing one static strategy on every shock.

Main findings

Traditional signals

≈ 0.700

Mean Log Loss for RSI, Bollinger Bands, and the combined indicator model is approximately 0.700 versus a 0.6931 random baseline. The indicators fail to deliver stable incremental predictive value and deteriorate most sharply in the Stress regime.

Hazard boundary

11.0%

The contagion engine identifies an 11.0% isotonic hazard threshold on the Bitcoin anchor. Above that structural breakpoint, the model switches from passive monitoring to contagion-aware execution and risk controls.

High-beta stress

+17.46%

The reported V8 15-Minute Vacuum Breaker produces a +17.46% retail net ROI on SOL in the January 3, 2024 stress-event test after the stated 0.24% round-trip taker fee.

Medium-beta routing

+7.80%

For ETH, the reported 45-Minute Gravity Hold reaches +7.80% retail net ROI in the same event, illustrating the paper's beta-dependent volatility half-life and the need for regime-specific exit horizons.

Why it matters

The paper's core claim is methodological rather than indicator-specific: during structural panics, price-derived technical signals observe the consequence of stress after the causal machinery is already moving. The DPRS framework therefore shifts attention toward funding stress, cross-asset transmission, transaction-cost constraints, and visible order-book liquidity. Its central engineering idea is that execution should be routed by the physical state of the market rather than by a fixed signal rule.

Full research record

The external scholarly record contains the complete manuscript and citation metadata. This ShockBridge page is the research showroom summary, not a substitute for the full paper.

Open scholarly record ↗