Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Do Whitepaper Claims Predict Market Behavior? Evidence from Cryptocurrency Factor Analysis

Murad Farzulla

TL;DR

This study asks whether whitepaper claims predict cryptocurrency market behavior by linking textual narratives to market structure through a three-space representation: (1) a claims matrix $\oldsymbol{\mathbf{C}} \\in \\mathbb{R}^{N \times K}$ generated from zero-shot NLP across $K=10$ semantic categories, (2) a market statistics matrix $\\boldsymbol{\mathbf{S}} \\in \\mathbb{R}^{M \times 7}$ across $M=49$ assets, and (3) latent factors $\\boldsymbol{\mathbf{F}} \\in \\mathbb{R}^{M \times R}$ derived from CP decomposition with rank $R=2$ explaining $92.45\%$ of variance. Alignment across spaces is tested with orthogonal Procrustes rotation and Tucker's congruence coefficient, revealing weak narrative alignment: $\\phi=0.341$, $p=0.332$ for claims--statistics and $\\phi=0.077$, $p=0.747$ for claims--factors, while statistics--factors show $\\phi=0.197$, $p<0.001$. Sensitivity analyses (rank variation, decomposition method, rolling windows, and Bitcoin-exclusion) confirm the robustness of the null narrative finding, with a few assets (e.g., NEAR, MKR, ATOM) contributing modestly to alignment. The results suggest a structural separation between narrative content and market-factor structure in crypto, carrying implications for narrative economics, investor due diligence, and regulatory disclosures. The paper contributes a reproducible cross-modal pipeline and a nuanced view on the limits of whitepaper claims for explaining market dynamics, while acknowledging power limitations due to a small common-entity sample.

Abstract

Cryptocurrency projects articulate value propositions through whitepapers, making claims about functionality and technical capabilities. This study investigates whether these narratives align with observed market behavior. We construct a pipeline combining zero-shot NLP classification (BART-MNLI) with CP tensor decomposition to compare three spaces: (1) a claims matrix from 24 whitepapers across 10 semantic categories, (2) market statistics for 49 assets over two years of hourly data, and (3) latent factors from tensor decomposition (rank 2, 92.45% variance explained). Using Procrustes rotation and Tucker's congruence coefficient, we test alignment across 23 common entities. Results show weak alignment: claims-statistics (phi=0.341, p=0.332), claims-factors (phi=0.077, p=0.747), and statistics-factors (phi=0.197, p<0.001). The statistics-factors significance validates our methodology, confirming the pipeline detects relationships when present. Inter-model validation with DeBERTa-v3 yields 32% exact agreement but 67% top-3 agreement. Cross-sectional analysis reveals heterogeneous contributions: NEAR, MKR, ATOM show positive alignment while ENS, UNI, Bitcoin diverge most. Excluding Bitcoin confirms results are not driven by market dominance. We interpret findings as weak alignment between whitepaper narratives and market factor structure. Limited power (n=23) precludes distinguishing weak from no alignment, but strong alignment (phi>=0.70) can be confidently rejected. Implications for narrative economics and investment analysis are discussed.

Do Whitepaper Claims Predict Market Behavior? Evidence from Cryptocurrency Factor Analysis

TL;DR

This study asks whether whitepaper claims predict cryptocurrency market behavior by linking textual narratives to market structure through a three-space representation: (1) a claims matrix generated from zero-shot NLP across semantic categories, (2) a market statistics matrix across assets, and (3) latent factors derived from CP decomposition with rank explaining of variance. Alignment across spaces is tested with orthogonal Procrustes rotation and Tucker's congruence coefficient, revealing weak narrative alignment: , for claims--statistics and , for claims--factors, while statistics--factors show , . Sensitivity analyses (rank variation, decomposition method, rolling windows, and Bitcoin-exclusion) confirm the robustness of the null narrative finding, with a few assets (e.g., NEAR, MKR, ATOM) contributing modestly to alignment. The results suggest a structural separation between narrative content and market-factor structure in crypto, carrying implications for narrative economics, investor due diligence, and regulatory disclosures. The paper contributes a reproducible cross-modal pipeline and a nuanced view on the limits of whitepaper claims for explaining market dynamics, while acknowledging power limitations due to a small common-entity sample.

Abstract

Cryptocurrency projects articulate value propositions through whitepapers, making claims about functionality and technical capabilities. This study investigates whether these narratives align with observed market behavior. We construct a pipeline combining zero-shot NLP classification (BART-MNLI) with CP tensor decomposition to compare three spaces: (1) a claims matrix from 24 whitepapers across 10 semantic categories, (2) market statistics for 49 assets over two years of hourly data, and (3) latent factors from tensor decomposition (rank 2, 92.45% variance explained). Using Procrustes rotation and Tucker's congruence coefficient, we test alignment across 23 common entities. Results show weak alignment: claims-statistics (phi=0.341, p=0.332), claims-factors (phi=0.077, p=0.747), and statistics-factors (phi=0.197, p<0.001). The statistics-factors significance validates our methodology, confirming the pipeline detects relationships when present. Inter-model validation with DeBERTa-v3 yields 32% exact agreement but 67% top-3 agreement. Cross-sectional analysis reveals heterogeneous contributions: NEAR, MKR, ATOM show positive alignment while ENS, UNI, Bitcoin diverge most. Excluding Bitcoin confirms results are not driven by market dominance. We interpret findings as weak alignment between whitepaper narratives and market factor structure. Limited power (n=23) precludes distinguishing weak from no alignment, but strong alignment (phi>=0.70) can be confidently rejected. Implications for narrative economics and investment analysis are discussed.
Paper Structure (66 sections, 3 theorems, 13 equations, 8 figures, 10 tables)

This paper contains 66 sections, 3 theorems, 13 equations, 8 figures, 10 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Cryptocurrency Narratives and Sentiment
  4. Factor Models in Cryptocurrency
  5. Tensor Methods in Finance
  6. Factor Comparison Methods
  7. Data
  8. Market Data
  9. Whitepaper Corpus
  10. Methodology
  11. Tensor Construction
  12. Tensor Decomposition
  13. CP Decomposition
  14. Alternating Least Squares
  15. Rank Selection
  16. ...and 51 more sections

Key Result

Theorem 4.1

The optimal rotation is $\mathbf{Q}^* = \mathbf{V}\mathbf{U}^\top$ where $\mathbf{U}\boldsymbol{\Sigma}\mathbf{V}^\top = \text{SVD}(\mathbf{A}^\top\mathbf{B})$.

Figures (8)

  • Figure 1: Market tensor slice (asset $\times$ feature) at mid-sample timestamp. Values are z-normalized. Structure reveals asset clusters and feature correlations.
  • Figure 2: Assets in CP factor space (rank 2). BTC, GALA, and SC are statistical outliers ($>2\sigma$). Colors indicate clusters from cross-sectional analysis.
  • Figure 3: Claims matrix: Zero-shot classification scores across selected assets and 10 functional categories. Full corpus includes 24 assets; subset shown for readability.
  • Figure 4: Rank sensitivity: Explained variance and alignment $\phi$ vs CP rank. Variance jumps at rank 2; alignment improves gradually.
  • Figure 5: Temporal evolution of alignment coefficient across 6-month rolling windows (3-month stride).
  • ...and 3 more figures

Theorems & Definitions (9)

  • Definition 4.1: Market Tensor
  • Definition 4.2: CP Decomposition
  • Definition 4.3: Orthogonal Procrustes Problem
  • Theorem 4.1: schonemann1966generalized
  • proof
  • Definition 4.4: Tucker's $\phi$
  • Theorem A.1
  • proof
  • Proposition B.1