Paper: arXiv 2512.11913

Authors: Chorok Lee

Abstract

We derive a specific functional form for factor alpha decay – hyperbolic decay alpha(t) = K/(1+lambda*t) – from a game-theoretic equilibrium model, and test it against linear and exponential alternatives. Using eight Fama-French factors (1963–2024), we find: (1) Hyperbolic decay fits mechanical factors. Momentum exhibits clear hyperbolic decay (R^2 = 0.65), outperforming linear (0.51) and exponential (0.61) baselines – validating the equilibrium foundation. (2) Not all factors crowd equally. Mechanical factors (momentum, reversal) fit the model; judgment-based factors (value, quality) do not – consistent with a signal-ambiguity taxonomy paralleling Hua and Sun’s “barriers to entry.” (3) Crowding accelerated post-2015. Out-of-sample, the model over-predicts remaining alpha (0.30 vs. 0.15), correlating with factor ETF growth (rho = -0.63). (4) Average returns are efficiently priced. Crowding-based factor selection fails to generate alpha (Sharpe: 0.22 vs. 0.39 factor momentum benchmark). (5) Crowding predicts tail risk. Out-of-sample (2001–2024), crowded reversal factors show 1.7–1.8x higher crash probability (bottom decile returns), while crowded momentum shows lower crash risk (0.38x, p = 0.006). Our findings extend equilibrium crowding models (DeMiguel et al.) to temporal dynamics and show that crowding predicts crashes, not means – useful for risk management, not alpha generation.

Complexity vs Empirical Score

  • Math Complexity: 6.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper derives a specific functional form for alpha decay from a game-theoretic equilibrium model, indicating advanced mathematical formulation, and rigorously tests it against multiple alternatives using extensive historical Fama-French data with statistical metrics like R-squared and out-of-sample forecasts.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Model & Validate Factor Alpha Decay"] --> B["Data Input<br>Fama-French Factors 1963-2024"]

  B --> C["Methodology<br>Game-Theoretic Equilibrium Model"]
  
  C --> D["Computational Process<br>Fit Hyperbolic vs Linear/Exponential Models"]
  
  D --> E{"Key Findings"}
  
  E --> F["<b>(1) Model Validation</b><br>Hyperbolic Decay fits Mechanical Factors<br>(Momentum R² = 0.65)"]
  E --> G["<b>(2) Crowding Divergence</b><br>Mechanical vs Judgment-based Factors"]
  E --> H["<b>(3) Acceleration</b><br>Crowding increased post-2015<br>(ETF growth ρ = -0.63)"]
  E --> I["<b>(4) Return Efficiency</b><br>Crowding selects failed strategies<br>(Sharpe 0.22 vs 0.39)"]
  E --> J["<b>(5) Tail Risk Prediction</b><br>Crowding predicts crashes<br>(Reversal 1.7-1.8x crash risk)"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a hyperbolic functional form for factor alpha decay derived from a game‑theoretic equilibrium model.

Source sentence
We derive a specific functional form for factor alpha decay -- hyperbolic decay alpha(t) = K/(1+lambda*t) -- from a game-theoretic equilibrium model

Assumes 6

  • The stylized one‑period quantity‑competition model excludes inventory risk, funding constraints, fees, and heterogeneous information. model
    Source sentence
    This stylized one-period model excludes inventory risk, funding constraints, fees, and heterogeneous information.
  • Traders are risk‑neutral and face a linear price impact with coefficient γ > 0. model
    Source sentence
    There are N ≥1 risk-neutral traders. Each chooses quantity qi ≥0 and faces the same expected pre-impact edge a > 0. Aggregate quantity is Q =
  • The realized one‑period profit includes a zero‑mean noise term ϵ with constant variance σ². statistical
    Source sentence
    Suppose the realized one-period profit of an active trader is Πi = q*i (e*+ ϵ), conditional on fixed a, N, γ, where E[ϵ] = 0 and Var(ϵ) = σ2 > 0.
  • The edge decays inversely with participation: e(t) = a/(N(t)+1). model
    Source sentence
    The economic motivation is inverse-linear residual edge e(t) = a/(N(t) + 1) under a stylized quantity game, not inverse-linear trader profit.
Show 2 more
  • Rolling‑window Sharpe statistics are descriptive and not structurally identified estimates of participation. statistical
    Source sentence
    The fitted rolling‑Sharpe hyperbola is a descriptive feature construction; it is not a structurally identified estimate of participation or a derived tail‑risk model.
  • Data are taken as monthly returns without further adjustments (values divided by 100, RF not subtracted again). data
    Source sentence
    Values are divided by 100 once; RF is not subtracted again from Mkt‑RF or from long‑short spreads.

Data

Backtest sample: US equities (Fama-French factors) · equities

  • Official monthly five‑factor, momentum, short‑term reversal, and long‑term reversal archiv · French Data Library market-prices public
    July 1963–December 2024 · Eight Fama‑French factors (Mkt‑RF, SMB, HML, RMW, CMA, Momentum, Short‑term reversal, Long
    Source sentence
    We downloaded the official monthly five-factor, momentum, short-term reversal, and long-term reversal archives on 26 September 2026. All four headers identify the August 2026 CRSP database. The analysis uses July 1963–December 2024, a complete 738-month calendar for all eight factors.

Recognised sources: Kenneth French Data Library

Costs

Not modeled A trading or execution result is reported without costs.

Source sentence
our single-signal model excludes that mechanism.

Finds 5

Reported performance (hyperbolic decay factor model): Sharpe 0.22

  • Hyperbolic decay fits mechanical factors best; momentum R²=0.65, outperforming linear (0.51) and exponential (0.61) baselines. statistical
    Source sentence
    Momentum exhibits clear hyperbolic decay (R^2 = 0.65), outperforming linear (0.51) and exponential (0.61) baselines
  • Crowding accelerated post‑2015; the model over‑predicts remaining alpha (0.30 vs. 0.15) and this over‑prediction correlates with factor‑ETF growth (ρ = -0.63). statistical
    Source sentence
    Crowding accelerated post-2015. Out-of-sample, the model over-predicts remaining alpha (0.30 vs. 0.15), correlating with factor ETF growth (rho = -0.63).
  • Crowding‑based factor selection fails to generate alpha; Sharpe 0.22 versus 0.39 for the factor momentum benchmark. performance
    Source sentence
    Crowding-based factor selection fails to generate alpha (Sharpe: 0.22 vs. 0.39 factor momentum benchmark).
  • Crowded reversal factors exhibit 1.7–1.8× higher crash probability, while crowded momentum shows lower crash risk (0.38×, p = 0.006). other
    Source sentence
    crowded reversal factors show 1.7--1.8x higher crash probability (bottom decile returns), while crowded momentum shows lower crash risk (0.38x, p = 0.006).
  • Average returns are efficiently priced; crowding predicts tail risk rather than mean returns. theoretical
    Source sentence
    Average returns are efficiently priced. Crowding

Checked by others

Supported by

  • Model-Free Market Risk Hedging Using Crowding Networks (2023)
    On the question: Does crowding predict tail risk and can it be used for hedging?
    Compare the findings

    This paper: Average returns are efficiently priced; crowding predicts tail risk rather than mean returns.

    That paper: Adding the crowding‑based long‑short hedge to a baseline 60/40 portfolio offers an alternative way to hedge tail risk without costly option‑based strategies.

    Both findings agree that crowding predicts tail risk rather than mean returns; A adds a crowding‑based hedge, B confirms the predictive power.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.