Does 12-1 momentum actually beat the market?
Grade B — reproducible and robustness-validated; in-sample, survivorship-limited, retrospective.
Status — PROVED — live (/momentum) · score v1
The claim
Buy the stocks that have gone up the most over the past year — specifically the top-decile performers on 12-month return, skipping the most recent month ("12-1"), equal-weighted, rebalanced monthly — and you beat the market. This is the single most-documented anomaly in academic finance (Jegadeesh & Titman, 1993 onward), and it's the backbone of countless "buy strength" strategies sold to retail traders. We put it on trial the way we test everything: does it beat a plain baseline out of the box, and does it survive when reality is added back in?
The verdict, in one line
Yes — a real, statistically significant edge, but a smaller one than the backtest advertises. The honest advantage is the risk-adjusted tilt (Sharpe ≈ 1.24 vs the market's 1.02) carrying ~8.8%/yr of genuine alpha (t ≈ 2.9), not the eye-catching +11-point CAGR gap, most of which is survivorship bias and a historic bull market.
How we tested it
- Universe: S&P 500 (a 503-name snapshot), monthly closes, 2010–2026 (185 months).
- Signal: 12-1 momentum, long the top decile (~50 names), equal-weighted, monthly rebalance.
- Baseline: the S&P 500 bought and held over the identical window (≈0 turnover).
- A note on method: this was a retrospective test — we did not write the hypothesis and pass/fail bar down in advance. We compensate with an unusually heavy robustness battery, below, and our score reflects that limitation rather than hiding it.
What we found
The decile sort works, and shows something interesting. Sorting next-month returns by momentum decile, the top decile returns 2.06%/mo vs a ~1.3% baseline — a clear win. But the bottom decile (extreme losers) is also elevated at 1.61% — the classic short-term reversal, or dead-cat bounce. So momentum is a "smile," not a clean ramp: winners keep winning and extreme losers bounce back. That detail matters later.
Head-to-head, long-only top-decile momentum beats the baseline on return, risk, and drawdown:
| Strategy | CAGR | Ann vol | Sharpe | Max DD |
|---|---|---|---|---|
| Market, buy-and-hold | 14.3% | 14.2% | 1.02 | −23.9% |
| Equal-weight all names (baseline) | 17.0% | 15.2% | 1.12 | −23.9% |
| Momentum, top decile, long-only | 25.5% | 19.3% | 1.28 | −21.0% |
It's real alpha, not just extra market exposure. A CAPM regression puts momentum's beta at 1.16 — yes, it takes more market risk — but it still carries 8.8%/yr of alpha with a t-stat of 2.94, significant after accounting for that beta. The edge is genuine skill, not merely "more beta dressed up as skill."
The overlays that failed (this is the important part)
We tried four popular "improvements" on the plain base. All four failed to beat it — and that is the strongest evidence that the base isn't a fragile, over-fitted curve:
| Enhancement | Result | Why it failed |
|---|---|---|
| Market-regime switch (only hold when the market is above its 200-day average) | Sharpe 1.28 → 1.17, drawdown barely moved | The filter whipsawed and missed rebounds |
| Long/short (long winners, short losers) | Sharpe collapsed to 0.30, DD −47.7% | The losers bounced (that smile) — shorting them bled money |
| Beta-neutral momentum | Lower Sharpe (1.23 vs 1.31) | Gave up more alpha than the risk it removed |
| Volatility targeting | Lower Sharpe at every setting (1.25 → 1.15) | Long-only large-cap has no "momentum crash" to cushion; delevering just missed rebounds |
The lesson: on a single history, "adding things" mostly adds overfitting risk. A clean base that resists improvement is a healthy base. So we ship the plain top-decile 12-1, monthly.
Does the edge survive reality?
- Costs: yes, comfortably. At 60%/month turnover, even 35 bps/side (high for liquid large-caps) leaves momentum ahead — 22.8% CAGR / 1.16 Sharpe vs the market's 14.3% / 1.02.
- Survivorship (partial de-bias): re-running on only names that were already trading in Jan 2010 (removing later-promoted winners) cost ~3.8%/yr CAGR — meaningful but not fatal. Momentum still cleared its own baseline and the market. The edge is not just an artifact of picking names we already knew got promoted.
- Parameter grid: every variant of {3, 6, 9, 12-month lookback} × {top 5/10/20%} beat the market. No lucky-cell dependence; the base sits mid-grid, not perched on an optimized peak.
- Calendar: beat the market in 13 of 16 years, and defended in 2022 (−3.4% vs the market's −18.2%) by rotating out of falling growth. The three down-years were mild underperformance, never disasters.
- Sub-periods: no decay — the relative edge was actually larger in the harder, more volatile 2018–2026 half.
Honest caveats (do not skip)
- 1.The magnitude is inflated. Using today's S&P 500 means only survivors — names that died or delisted are absent, which lifts the returns. Expect live CAGR below 25%. The robust number is the relative, risk-adjusted edge, not the headline.
- 2.Sharpe 1.28 is too high to be real. Published momentum Sharpes are ~0.5–0.8. The gap is survivorship + a bull-dominated sample + no market-impact costs.
- 3.One ~16-year US large-cap sample. Past ≠ future; 2022's defense is one episode, not a law.
- 4.Turnover is high (~720%/yr). It survives costs, but it is not a low-maintenance hold.
How the score breaks down
| Dimension | Score | Why |
|---|---|---|
| Edge | 22 / 25 | Statistically significant alpha (t ≈ 2.9); beats the baseline on return, Sharpe, and drawdown. Docked because the honest magnitude is the modest Sharpe tilt, not the CAGR gap. |
| Robustness | 22 / 25 | Survives costs to 35 bps, a partial de-bias, the full parameter grid, 13/16 calendar years, and both sub-periods; resisted four overlays. Short of full marks only because dead-names survivorship can't be fixed on free data. |
| Practicality | 15 / 25 | Runnable in a retail account (~50 names, fractional), but ~60%/month turnover and monthly rebalancing make it hands-on rather than set-and-forget. |
| Evidence | 17 / 25 | Reproducible, robustness-validated, honest caveats. Capped below 20 by the retrospective (not pre-registered) test and the survivorship data limit. |
| Total | 76 / 100 | Qualified edge — a real, well-evidenced edge whose live magnitude will be lower than the backtest, and whose turnover makes it work, not passive income. |
Evidence Grade B: we trust the direction strongly; we discount the magnitude by design.
What it means for you
Momentum is a genuine, literature-backed edge you can run in a normal account — but treat the backtest CAGR as a ceiling, not a forecast, and expect to do monthly work. It defends in some bear markets (2022) but not all — it can whipsaw in sharp V-shaped reversals. Of everything we've put on trial, this is the one we'd call a true edge rather than merely a way to lose less.
References
- Jegadeesh & Titman (1993), "Returns to Buying Winners and Selling Losers." Journal of Finance.
- The four-dimension scoring rubric behind the number above: The Tapelab Score.
Scored under score v1, the fixed public rubric summarised on the research index. Report last updated 2026-07-03; published from the research repository's canonical write-up, unedited. Nothing here is investment advice.