Paper: arXiv 2311.01985

Authors: Michael Pinelis, David Ruppert

Abstract

We construct the maximally predictable portfolio (MPP) of stocks using machine learning. Solving for the optimal constrained weights in the multi-asset MPP gives portfolios with a high monthly coefficient of determination, given the sample covariance matrix of predicted return errors from a machine learning model. Various models for the covariance matrix are tested. The MPPs of S&P 500 index constituents with estimated returns from Elastic Net, Random Forest, and Support Vector Regression models can outperform or underperform the index depending on the time period. Portfolios that take advantage of the high predictability of the MPP’s returns and employ a Kelly criterion style strategy consistently outperform the benchmark.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 6.0/10 (capped from 8.0: transaction costs not modelled; see penalty flags)
  • Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 8.0)

Why this score: The paper employs advanced matrix algebra, convex optimization, and rigorous derivations (e.g., eigenvectors, KKT conditions) but is grounded in a detailed, multi-stage empirical implementation with a strict holdout set, out-of-sample validation, and specific ML models for S&P 500 stocks.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Maximize Portfolio Predictability"] --> B["Input: S&P 500 Constituents & Historical Data"]
  B --> C["Compute Predicted Returns<br>Elastic Net | Random Forest | SVR"]
  C --> D["Estimate Covariance Matrix<br>of Prediction Errors"]
  D --> E["Optimization<br>Construct Maximally Predictable Portfolio MPP"]
  E --> F["Strategy: MPP + Kelly Criterion"]
  F --> G["Outcome<br>Consistent Outperformance vs Benchmark"]
  G --> H["Key Finding<br>High R<sup>2</sup> enables predictive strategy"]
  
  style A fill:#e1f5fe
  style H fill:#e8f5e8

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 constructing a maximally predictable portfolio (MPP) of stocks using machine learning forecasts and applying a Kelly‑style timing strategy to achieve outperformance.

Source sentence
We construct the maximally predictable portfolio (MPP) of stocks using machine learning.

Assumes 6

  • Rt is a jointly stationary and ergodic stochastic process with finite expectation. statistical
    Source sentence
    Assumption 1 Rt is a jointly stationary and ergodic stochastic process with finite expectation E[Rt] = µ = [µ1, µ2, ..., µn]⊤and autocovariance matrices E[(Rt−k−µ)(Rt−µ)⊤] = Σk, where k ≥0 without loss of generality.
  • The conditional forecast ˆZt equals the conditional expectation E[Zt|Ft−1]. model
    Source sentence
    the conditional expectation of the asset returns with the information set at the previous time.
  • Forecast errors ϵt are independent of the forecasts ˆZt. statistical
    Source sentence
    where E[ϵt|Ft−1] = 0 ∈IRn, var[ϵt|Ft−1] = Γ ∈IRn×n and importantly for the following results ˆZt and ϵt are assumed independent.
  • Equation 3 holds only if forecast errors are independent from the forecasts. model
    Source sentence
    Equation 3 only holds if assuming the forecast errors are independent from the forecasts, which is often not realistic.
Show 2 more
  • Sample covariance matrices Σ0 and Γ are estimated each month with rolling‑window historical data. data
    Source sentence
    In the empirical implementation, each expression above can be estimated with historical data Σ0 and Γ are estimated each month with rolling window sample data as sample covariance matrices.
  • The information set Ft−1 includes observable economic variables such as interest‑rate spreads and stock‑specific factors. other
    Source sentence
    In the information set Ft−1 are observable leading economic variables like interest-rate spreads or stock-specific factors like momentum, fundamental company performance measures, liquidity, and volatility.

Data

  • CRSP database · CRSP market-prices public
    Source sentence
    Stock prices, volume, and shares outstanding are from the CRSP database.
  • Compustat database · Compustat fundamentals public
    Source sentence
    Fundamental company data is from Compustat.
  • Kenneth French data library · Kenneth French benchmark public
    Source sentence
    Kenneth French’s website contains the historical risk‑free
  • CRSP/Compustat merged database · CRSP/Compustat market-prices public
    S&P 500 constituents
    Source sentence
    The CRSP/Compustat merged database gives stock returns and the set of stocks that are in the S&P 500 index each month.

Recognised sources: Compustat CRSP

Costs

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

Finds 5

  • The MPPs achieve high monthly coefficients of determination (R²), with OLS attaining the highest average R² among the models. predictive-accuracy
    Source sentence
    The predictability for portfolios formed with the different models is similar, but OLS actually attains the highest coefficient of determination over the three different weight maxes.
  • Machine‑learning‑based MPPs generate higher annual returns and Sharpe ratios than the S&P 500 benchmark. performance
    Source sentence
    The machine learning MPP has higher annual returns than the market index.
  • Timing the MPP using its conditional excess‑return and volatility expectations improves risk‑adjusted returns relative to both the index and a reward‑risk‑timed index. performance
    Source sentence
    When an investor times the machine learning MPP with its conditional excess return and volatility expectations, he or she can earn substantial improvements in risk‑adjusted returns over not just the index but the portfolio which reward‑risk times the index.
  • The Kelly‑style weighting scheme, where weights are proportional to the probability of a successful bet, is proposed to enhance out‑of‑sample performance. theoretical
    Source sentence
    …a Kelly strategy (Kelly, 1956) where the MPP weights are proportional to the probability of successful bet.
  • Out‑of‑sample R² for individual stock return forecasts is reported, indicating the predictive power of the ML models. predictive-accuracy
    Source sentence
    To assess the out‑of‑sample predictive performance for individual stock return forecasts, we calculate the out‑of‑sample R2 as R2 oos = 1 −

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 28 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 1 paper in the archive (1 by independent authors)

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.