Our first backtest showed that copying new consensus buys earns nothing after the 45-day disclosure lag. This one tests the opposite construction — and it actually works, with an asterisk big enough to be the headline.
The strategy
Instead of chasing what funds just bought, hold what concentrated managers already own in size:
- Universe: "conviction" funds only — managers reporting ≤100 positions.
- Heavyweight: a position ≥2% of that fund's portfolio.
- Basket: the 8 stocks overweighted by the most conviction funds, equal-weighted.
- Entry: quarter-end + 46 days (after the filings are public). Benchmarks measured over identical windows. Every rule is mechanical — no hand-picking.
It beats SPY. Consistently.
Across 37 rolling one-year starting points since 2016:
| Hold 1 year | Basket | Excess vs SPY | Excess vs QQQ |
|---|---|---|---|
| Median | +27.3% | +12.0% | +2.1% |
| Average | +24.8% | +9.4% | +2.9% |
| Worst start | −38.4% | −25.0% | −18.5% |
| Beat rate | — | 73% | 59% |
A quarterly-rebalanced version compounds to +641.7% over 9.8 years (22.8% annualized) vs SPY's +287.6%. The disclosure lag doesn't kill this signal, because it isn't a timing signal — heavyweight positions persist for quarters, so being 46 days late barely matters.
Now the honest part
1. Most of the "alpha" is just growth beta. Consensus heavyweights are, in practice, large-cap growth names. Compare against QQQ instead of SPY and the median one-year edge shrinks from +12pt to +2.1pt; at three-year horizons the basket actually trails QQQ (−5.4pt median, beating it only 41% of the time). If your alternative is "buy QQQ and do nothing," this strategy mostly re-derives that trade with extra steps.
2. You pay for the returns in drawdown. Daily risk metrics, 2016–2026:
| CAGR | Volatility | Sharpe | Max drawdown | Calmar | |
|---|---|---|---|---|---|
| Basket | 22.5% | 26.3% | 0.78 | −49.3% | 0.46 |
| SPY | 14.7% | 17.9% | 0.71 | −33.7% | 0.44 |
| QQQ | 20.0% | 22.2% | 0.81 | −35.1% | 0.57 |
In 2022 the basket lost −42.8% while SPY lost −18.6%. On risk-adjusted terms QQQ wins outright — higher Sharpe, higher Calmar, a third less drawdown. A strategy is only as good as your ability to hold it through its worst year, and this one's worst year is brutal.
3. The fine print all points one way. Equal weight, no transaction costs, and components with missing price history get dropped — every one of those simplifications flatters the basket. The rolling windows also overlap, so "73% of starts beat SPY" overstates the number of independent bets.
What we'd actually take from this
- Concentrated-manager overweights contain real information — unlike consensus new buys, the effect survives honest entry timing.
- But the information is mostly "own quality large-cap growth," which QQQ
sells you for 0.20% a year with half the drawdown.
- Where the basket genuinely differs from QQQ (2023: +71.0% vs +55.9%; 2024:
+52.6% vs +27.7%), it did so by concentrating harder in the same theme — which is also exactly how it produced −42.8% in 2022.
The consensus of concentrated managers is a useful screen. As a mechanical portfolio, it's leveraged QQQ in a trench coat.
Methodology and code: scripts/backtest_consensus_basket.py. Vintages with more than 2 missing-price components are discarded; survivorship effects inflate basket returns. Nothing here is investment advice.