Paper: arXiv 2406.19105

Authors: Matthew J. Schneider, Rufus Rankin, Prabir Burman, Alexander Aue

Abstract

The M6 Competition assessed the performance of competitors using a ranked probability score and an information ratio (IR). While these metrics do well at picking the winners in the competition, crucial questions remain for investors with longer-term incentives. To address these questions, we compare the competitors’ performance to a number of conventional (long-only) and alternative indices using standard industry metrics. We apply factor models to measure the competitors’ value-adds above industry-standard benchmarks and find that competitors with more extreme performance are less dependent on the benchmarks. We also uncover that most competitors could not generate significant out-performance compared to randomly selected long-only and long-short portfolios but did generate out-performance compared to short-only portfolios. We further introduce two new strategies by picking the competitors with the best (Superstars) and worst (Superlosers) recent performance and show that it is challenging to identify skill amongst investment managers. We also discuss the incentives of winning the competition compared to professional investors, where investors wish to maximize fees over an extended period of time.

Complexity vs Empirical Score

  • Math Complexity: 6.0/10
  • Empirical Rigor: 6.0/10 (capped from 8.5: no out-of-sample test; see penalty flags)
  • Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 8.5)

Why this score: The paper employs advanced statistical methods (factor models, hypothesis testing on performance metrics) and shows rigorous empirical analysis using competition data, benchmark comparisons, and detailed risk metrics.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Assess M6 Competitors' Investment Skill<br>Beyond Competition Metrics"] --> B["Data & Inputs<br>M6 Competitor Predictions<br>Industry Benchmarks<br>Factor Model Data"]
  B --> C["Methodology<br>1. Calculate Standard Metrics (IR, etc.)<br>2. Apply Factor Models<br>3. Compare to Random Portfolios<br>4. Create 'Superstar'/'Superloser' Strategies"]
  C --> D["Computational Analysis<br>Benchmark-adjusted returns<br>Statistical significance testing"]
  D --> E{"Key Findings & Outcomes"}
  E --> F["Winners lacked<br>significant alpha vs. benchmarks<br>Skill identification is challenging"]
  E --> G["Extreme performers<br>less benchmark-dependent"]
  E --> H["Incentive misalignment:<br>Competition vs. Long-term investor goals"]

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 documents M6 competitors' performance with industry metrics, applies factor models to assess value‑add, and introduces Superstar/Superloser strategies.

Source sentence
... Second, we apply factor models to competitors’ returns and find that competitors who performed in the extremes tended to be less dependent on the benchmark indices.

Assumes 5

  • All benchmark indices and competitors traded on the same 238 trading days for a fair comparison. data
    Source sentence
    we removed the Crypto returns that occurred on the weekends (non-trading days) and assumed that all benchmark indices and competitors traded on the same 238 trading days for a fair comparison.
  • Returns across the competition days are independent and identically distributed (i.i.d.) for MDD analysis. statistical
    Source sentence
    trading days in the competition are independent and identically distributed (i.i.d.)
  • A low fee structure of 1% management and 10% performance fees is applied daily. execution
    Source sentence
    we assume a low fee structure (1%/10%) and deduct performance fees and prorated management fees daily.
  • Crypto returns on weekends are excluded from the analysis. data
    Source sentence
    For the analysis in this paper, we removed the Crypto returns that occurred on the weekends (non-trading days)
Show 1 more
  • Raw returns are used instead of log returns, approximating log returns via exp transformation. model
    Source sentence
    which differ from log returns used by Makridakis et al. (2023)

Data

  • M6 competition competitor NAVs benchmark unknown
    238 trading days · individual large‑cap equities or ETFs
    Source sentence
    The majority of assets in the universe for the competition are individual large-cap equities or ETFs that track some set of large-cap equities.
  • Major benchmark indices (including equal‑weighted benchmark) benchmark unknown
    238 trading days · large‑cap equity market
    Source sentence
    Table 1 presents the financial metrics for the major benchmark indices, including the equal-weighted benchmark in the M6 Competition.

Costs

Modeled Costs are part of the reported headline results.

A low fee structure of 1% annual management fee and 10% performance fee is deducted daily from NAVs.

Source sentence
we assume a low fee structure (1%/10%) and deduct performance fees and prorated management fees daily.

Finds 5

  • Competitors with extreme performance are less dependent on benchmark indices, showing low R‑square values. performance
    Source sentence
    We find that those competitors with extreme alphas (and high values of SR) all had R-square values below 50% (and frequently below 20%).
  • Only 12 of 163 alphas in the seven‑factor model are statistically significant at 5%, and 11 of those are negative. statistical
    Source sentence
    Of the 163 alphas in the seven factor model, 12 alphas were statistically significant at the 0.05 level, including the largest (annualized) alpha of 30.8%. However, 11 of the 12 statistically significant alphas were negative.
  • GRS tests reject the null that all alphas equal zero for both one‑factor and seven‑factor models (p < 0.0001). statistical
    Source sentence
    For the one factor model, the test statistic is 15.3 with a p-value < 0.0001, rejecting the null hypothesis that all alphas are equal to 0. For the seven factor model, the test statistic is 15.1 with a p-value < 0.0001, rejecting the null hypothesis that all alphas are equal to 0.
  • Management and performance fees of 1%/10% slightly worsen median competitor performance (SR from -0.09 to -0.17, IR from -1.3 to -2.6). performance
    Source sentence
    For the median competitor, the 1%/10% fees decreased the SR from -0.09 to -0.17, the IR decreased from -1.3 to -2.6, and the MDD increased from 15.5% to 16.0%.
  • Most competitors did not add value over an equal‑weighted portfolio, but a majority expressed smaller losses than long‑only benchmarks. performance
    Source sentence
    while most competitors did not add value over the equal-weighted portfolio of universe constituents, a majority did manage to express smaller losses than widely used long-only benchmarks.

Checked by others

Supported by

  • Measuring the risk or reducing it, that is the question: is risk measurement necessary for risk reduction? (2026)
    On the question: Do active portfolio strategies add value relative to an equal‑weighted (1/N) benchmark?
    Compare the findings

    This paper: Most competitors did not add value over an equal‑weighted portfolio, but a majority expressed smaller losses than long‑only benchmarks.

    That paper: Active portfolio strategies generally achieve lower Sharpe ratios than the 1/N benchmark due to transaction‑cost‑induced return erosion.

    Both find that active strategies generally do not outperform the 1/N benchmark: most competitors add no value and show smaller losses, and active portfolios have lower Sharpe ratios due to transaction‑cost erosion.

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.