Academic audit

Failedequity XS alpha

CAPM Alpha Ranking Strategy on Dow 30 Companies

The three-gate gauntlet · genuine only if it clears all three and survives adversarial refutation
Eliminated here
Gate 1
Survivorship-free
biased
survivor-only universe
Gate 2
Placebo ≥ P95
not evaluated
Gate 3
Cost-aware net
not evaluated
Failed
Rejected at the survivorship gate — a survivor-only universe, so any edge can't be confirmed.

The strategy ranks Dow-30 stocks by their CAPM (Jensen 1968) alpha — the intercept from regressing each stock's return on the market — and holds the highest-alpha names, rebalancing periodically. It is a cross-sectional stock-ranking rule, not a standalone system.

What we found

Within the tested universe the ranking does carry real skill: the signal separates from random baskets at the 96.9th placebo percentile. But the universe is the problem. The test runs on the CURRENT Dow-30 members — a hardcoded set of survivors in which removed names such as GE, Citigroup, GM, AIG, and Kodak are simply absent. That survivorship bias inflates the headline result, and the effect cannot be confirmed as genuine without point-in-time Dow membership. It also has a losing worst year (worst-year RF -0.85), so even the biased version is not a dependable standalone rule. Verdict: failed on survivorship bias.

How we tested it
2005–2026 test windowmodelled liquidity-aware costssurvivorship biased
  • The current Dow-30 constituents only — a survivorship-biased 30-name universe (removed/delisted names like GE, Citigroup, GM, AIG, and Kodak are absent), daily, 2005-2026. Realistic modelled costs.
  • Placebo / robustness test: real result vs random baskets or shuffled signals (real vs the 95th percentile of random)
Source: Jensen (1968) alpha ranking on Dow-30 (QuantConnect implementation)
Read the paper ↗
← The Academic Audit — all 54 studies

Research, not investment advice. “Validated” factor-legs are market-neutral diversifying building blocks with a losing worst year — none is a standalone tradeable strategy. Metrics are cost-aware and modelled (not live fills); the 2005–2026 test window is out-of-sample versus the source paper. Dollar figures are not returns and are omitted by design.