Our earlier backtests showed that timing 13F signals doesn't work — consensus new-buys earn nothing after the disclosure lag, and the consensus heavyweight basket is mostly leveraged QQQ. But those tested strategies. This one tests the managers themselves: if you had cloned each fund's entire disclosed book, who would you actually have wanted to follow?
The method
For all 59 funds with 3+ years of history, we build a daily clone NAV:
- Holdings: each fund's top-20 reported positions.
- Rebalance: only on filing dates — you hold what was last disclosed, so there's no lookahead and the 45-day lag is baked in.
- Alpha: clone CAGR minus SPY CAGR over the same window.
This is the fairest test of "is this manager worth copying" — it doesn't depend on catching any single trade, just on the quality of the book.
The result: a coin flip at the median
Across all 59 funds, the median clone alpha was −0.1 points a year, and only 29 of 59 (49%) beat SPY. Cloning a hedge fund picked at random from our tracked set was, over the last three years, indistinguishable from a coin flip against the index.
That's the headline every "follow the smart money" pitch leaves out. The average institutional book, replicated honestly, does not beat a cheap index fund.
But the tails are enormous — and real
The spread from best to worst is 73 points a year:
| Rank | Fund | Clone CAGR | Alpha vs SPY | Max drawdown |
|---|---|---|---|---|
| 1 | Slate Path Capital | 64.1% | +43.8pt | −31% |
| 2 | Greenoaks Capital | 62.7% | +42.5pt | −37% |
| 3 | CAS Investment Partners | 59.7% | +39.4pt | −38% |
| 4 | Duquesne (Druckenmiller) | 41.2% | +21.0pt | −24% |
| 5 | Whale Rock Capital | 39.5% | +19.2pt | −34% |
| … | ||||
| 57 | Abdiel Capital | −0.9% | −21.2pt | −40% |
| 58 | ShawSpring Partners | −5.9% | −26.2pt | −52% |
| 59 | Icahn Enterprises | −9.1% | −29.3pt | −57% |
The top funds didn't beat SPY by a little — they doubled or tripled it, for years. This is why "13F cloning doesn't work" is as wrong as "13F cloning works." Both are true for different funds, and the difference is the whole game.
The catch: you can't pick the winners in advance
Three things keep this from being a free lunch:
- Survivorship and hindsight. We can see that Slate Path's clone returned 64% a year because it already happened. Three years ago, ranking these 59 funds by future alpha would have been guesswork — past clone alpha is a weak predictor of next year's.
- You pay for the top returns in drawdown. Every fund above +19pt alpha also drew down 30–38%. The concentrated books that produced the returns are the same ones that will hurt most in a bad year — and holding through that is the part backtests never capture.
- Even the ranking is a moving target. Duquesne (Druckenmiller) at #4 is a macro book that turns over heavily; its clone alpha next year could be anywhere. Concentrated single-theme winners (Whale Rock, Coatue — both tech) rise and fall with one trade.
What we'd actually do with this
- Use the clone rankings as a research filter, not a buy list: the funds at the top are worth understanding — what do Slate Path and Greenoaks see that the median manager doesn't?
- Weight the drawdown column as heavily as the alpha column. A +20pt alpha you can't sit through is a −20pt alpha in practice.
- Treat any single fund's clone as one concentrated bet, not a diversified strategy. The median result is the honest base rate; the tails are stories, not guarantees.
Cloning the right manager would have been spectacular. Cloning the average one would have matched the index. Knowing which is which ahead of time is the only question that matters — and it's the one the data can't answer for you.
Methodology and code: scripts/build-fund-performance.ts, from daily clone NAV series with no lookahead. 59 funds with ≥3 years of history; as of the latest price close. Survivorship effects flatter the sample. Nothing here is investment advice.