The Somatic Marker of Markets (The Coherence Meter)

Horizon 0 of the Algoplexity Research Program: Falsifying Statistical Complexity.

This repository contains the code and computational proofs for the paper: "The Somatic Marker of Markets: Falsifying Statistical Complexity in Structural Break Detection"

---

Abstract

Financial markets are often modeled using high-dimensional multivariate statistics. This paper presents a rigorous falsification of such approaches for the specific task of structural break detection. Building on the algorithmic ontology established in [Mak, 2023], we test whether a "Microscope" (Vector Autoregression/Covariance) can detect regime shifts better than a "Stethoscope" (Univariate Aggregation).

Using the Minimum Description Length (MDL) principle, we demonstrate that high-resolution statistical models are overwhelmed by parameter complexity (Negative MDL Cost), effectively blinding them to the signal. In contrast, a simple Predictive Error Proxy (The Coherence Meter) detects the Q4 2018 crash with 23.9 bits of evidence—double the baseline.

To test for generalizability, we conducted a systematic validation on the Algoplexity Structural Break Benchmark (hosted on Hugging Face). This dual-phase experiment revealed a profound capability: the univariate diagnostic functions as a powerful early-warning system, detecting instability with a mean lead time of 31.21% in-sample and 39.37% on unseen, out-of-sample data.

This confirms the "Somatic Marker Hypothesis": that systemic instability is best detected not by modeling the state of every asset, but by measuring the agony (predictive failure) of the collective system. These findings necessitate a shift from statistical curve-fitting to Algorithmic Information Dynamics, setting the stage for future topological investigations (Horizon 1).

---

📊 Datasets & Experimental Design

To ensure robust validation, this research employs a dual-track data strategy, combining real-world historical case studies with large-scale systematic benchmarks.

1. Historical Case Study: The Q4 2018 U.S. Equity Downturn

• Purpose: Qualitative validation, falsification of "Microscope" (multivariate) approaches, and "mechanism" demonstration.
• Source: Yahoo Finance (Public API).
• Assets: Daily closing prices for a representative basket of US Blue Chips (AAPL, MSFT, BA, CAT, DIS, GE, IBM, TSLA).
• Usage: Used in computational_narrative.ipynb to demonstrate the Coherence Meter's detection logic on a known, real-world crisis.

2. Systematic Benchmark: The Algoplexity Structural Break Dataset

• Purpose: Quantitative validation of Generalization Capability. We measure Early-Warning Lead Time on both In-Sample (-31.21%) and Out-of-Sample (-39.37%) data.
• Source: Hugging Face (Immutable Scientific Artifact).
• Assets:
  • X_train.parquet: 1,000+ continuous financial time series (In-Sample Training).
  • y_train.parquet: Ground-truth labels for precise structural break timing.
  • X_test.reduced.parquet: Unseen series from the Falcon competition (Out-of-Sample Generalization).
• Usage: Used in systematic_validation.ipynb.
• Access: Hosted permanently at algoplexity/computational-phase-transitions-data.

---

Foundational Work

The research in this paper is a direct extension of the foundational work completed in the author's Master's capstone thesis. The thesis establishes the core principle that financial markets possess discoverable algorithmic structure, providing the theoretical and empirical bedrock for the "Coherence Meter." The full thesis and its accompanying code are publicly available:

• 📄 Read the full foundational thesis on ResearchGate
• 💻 Access the thesis source code on GitHub

---

Reproducing Our Findings

All key findings of the paper can be reproduced by running the three Google Colab notebooks provided in this repository.

1. Main Narrative: The Q4 2018 Case Study (Section 3 & 4)

Click the badge above to run the main computational_narrative.ipynb. This notebook tells the primary story of the paper, from the "Stethoscope vs. Microscope" showdown to the definitive "Final Showdown" victory of the Coherence Meter.

• Data Source: Yahoo Finance (Live/Public API).
• Execution Time: ~2-3 minutes.

2. Definitive Proof: The Early-Warning System (Section 5)

Click the badge above to run systematic_validation.ipynb. This notebook performs the large-scale validation on the CrunchDAO Structural Break Benchmark, proving the early-warning capability.

• Data Source: Hugging Face (Cloud Artifact).
• Experiments: In-Sample Validation (-31% Lag) & Out-of-Sample Generalization (-39% Lag).
• Execution Time: ~5-10 minutes.

3. Foundational Experiments & Full Audit (Section 2 & 3)

Click the badge above to run supplementary_evidence.ipynb. This notebook contains the "full audit trail" for our foundational experiments, including the code to generate the conceptual figures ("Rule 37 Delusion," "Cost of Complexity") and the complete, computationally intensive "Microscope" scans.

---

🔗 Citation

If you find this work useful in your research, please cite the permanent DOI:

@misc{algoplexity2025somatic,
  title   = {The Somatic Marker of Markets: Falsifying Statistical Complexity in Structural Break Detection},
  author  = {Mak, Yeu Wen},
  year    = {2025},
  publisher = {ResearchGate},
  doi     = {10.13140/RG.2.2.35243.09767},
  url     = {https://doi.org/10.13140/RG.2.2.35243.09767}
}

@misc{algoplexity2025code,
  author = {Mak, Yeu Wen},
  title  = {The Coherence Meter: Code and Computational Appendices},
  year   = {2025},
  url    = {https://github.com/algoplexity/Coherence-Meter}
}