ASRI: An Aggregated Systemic Risk Index for Cryptocurrency Markets - An Interpretable Crypto-Native Stress Composite for Retrospective Systemic-Risk Discrimination

By Murad Farzulla, Andrew Maksakov

Rating

1438
Battle Count: 95

Relevance

5/10
ASRI is explicitly framed as a retrospective monitoring framework, not a validated early-warning system or trading signal. Its primary value is interpretive (channel attribution, lead-time, regime structure) rather than discriminative. For quantitative trading, it could serve as one input to a risk overlay for institutional crypto allocators (trimming exposure when systemic readings elevated), but the authors stress this is illustrative given the four-event limitation. The index does not provide point forecasts (negative R² for continuous prediction) and its discrimination is matched by a standalone VIX series. Procyclicality concerns are acknowledged. More relevant for risk management and macroprudential surveillance than for alpha generation or algorithmic trading strategy development.

Implementation Complexity

7/10
Moderate-to-high complexity. Requires multiple data sources (DeFi Llama, FRED, CoinGecko, Token Terminal, RWA.xyz, GDELT), mixed-frequency data handling with interpolation/forward-fill protocols, four sub-index calculations with piecewise mappings and normalisation, HMM estimation with EM algorithm, VAR estimation for connectedness, HAC inference, moving-block bootstrap, and walk-forward validation. However, complete implementation is open-source (Python 3.11, pandas, statsmodels, scipy, scikit-learn, hmmlearn), well-documented with API endpoints, and includes a four-layer data pipeline (ingestion, normalisation, computation, publication). The main complexity lies in data quality management, proxy validation, and the statistical inference framework rather than in the index computation itself.

Reproducibility

4/5
Code available on GitHub (MIT license), frozen dataset on Zenodo (10.5281/zenodo.17918239), detailed replication scripts mapped to tables/figures via REPRODUCIBILITY.md, DATA_PROVENANCE.md hash-freezes the series. However, the published daily series is NOT bit-reproducible from the live ingestion pipeline (original point-in-time API inputs and protocols/bridges universe were not retained). The archived frozen series is the reproducibility artefact. Pseudo-real-time replication module included. Random seeds set to 42 for all stochastic procedures.

About this paper

Methodology: Aggregated Systemic Risk Index (ASRI) - Four-Channel Weighted Composite. Problem types: Risk Management, Anomaly Detection, Classification, Time Series Forecasting, Dimensionality Reduction.

The interactive Everscope explorer (charts, battles, favorites) loads below.