Following the 'smart money' — 13F cloning

A note on the managers — read this first. Buffett, Hohn, Ackman and Akre are among the finest investors of their generation; their own long-term records are exceptional and nothing here questions that. This study is about the cloning mechanism — whether a delayed, public, long-only copy of their 13F filings can capture that edge — not the managers' skill. Where a clone lags, it is a limitation of the clone: proof that their genius lives in the things a 13F can't show (position sizing, entry/exit timing, non-US holdings, shorts, and — for Buffett — wholly-owned private businesses). A clone underperforming is never a comment on the investor.

The short version

Can you get rich copying famous investors' filed holdings? Copying them broadly loses — the funds built to do exactly this trailed the market, and most shut down. But copy the right kind of manager — one who holds a handful of stocks for years (like Chris Hohn) — and it genuinely beat the market for over a decade, even copying 45 days late. The hard part isn't the copying; it's picking the right manager.

Abstract

"A famous investor is buying it" feels like a signal. We tested the most direct version — copying the disclosed holdings of institutional managers from their SEC 13F filings — on our own data, using the real-world products built to do exactly this: the 13F-cloning ETFs. The verdict is clear. The broad clones underperformed: Global X Guru (GURU) trailed the S&P by ~3.2pp/yr over 14 years, and AlphaClone (ALFA) trailed by ~4.2pp/yr over a decade before being liquidated in 2022. The one cloning product that beat the index — Goldman's Hedge Industry VIP (GVIP), +1.4pp/yr — did so by concentrating in the same mega-cap growth names the hedge-fund crowd already crowds into, which is a factor tilt, not demonstrated skill. Layered on top are three structural problems that make naive following worse than it looks: a 45-day disclosure lag, the fact that 13F shows only the long US equity sleeve (no shorts, options, timing or cash), and that the documented edge lives only in managers' top, high-conviction positions, not the long tail. But selectivity by manager profile changes the picture, and the standout result is genuinely impressive: copying Chris Hohn's (TCI) lagged 13F top holdings beat the S&P by ~+2.7pp/yr for over a decade (§3b). Hohn's own record is exceptional, and the striking part is that you could follow him through public filings — 45-day lag and all — and still outperform the market. That works because his book is the right profile (ultra-concentrated, very-low-turnover, genuinely long-only); the hedge-fund-style filers we also tested (Buffett, Ackman) miss most of their engine in a 13F, and their clones lagged. Conclusion: naive, broad 13F cloning is not a signal — but cloning the right manager profile (concentrated, low-turnover, long-only) really can beat the market. The honest limit is attribution: even Hohn's clone trailed a mega-cap-growth index (QQQ), so over one growth decade we can't fully separate skill from a quality/growth tilt — treat it as a promising, lag-acknowledged positioning input, not yet a standalone thesis.

1. The claim (and why it matters to us)

13F filings are public: every large US manager must disclose long equity holdings each quarter. If a great investor's picks predict returns, copying them is a free, legal edge — and a whole industry (hedgefollow.com, WhaleWisdom, several ETFs) is built on the premise. We need to know whether it survives contact with our own data and our own costs.

2. What the literature claims — and our scepticism

• The optimistic case: "best-ideas" research (and cloning vendors' backtests) shows portfolios built from the high-conviction top positions of skilled managers beat the market even after the filing lag — selectivity is everything (e.g. "Outperforming the Market: Portfolio Strategy Cloning from SEC 13F Filings"; "Systematic 13F Hedge Fund Alpha", Angelini et al.; Quantpedia "Alpha Cloning").
• Our scepticism (confirmed below): the headline edge is for carefully selected managers and concentrated positions, often measured pre-publication and pre-crowding. Vendor backtests are self-serving. CXO Advisory's review of 13F clone performance was mixed-to-underwhelming for naive clones. The real-world products are the honest test — and most of them failed or closed.

3. Our own analysis — the cloning ETFs ← the heart of the report

Data: yfinance total-return (dividend-adjusted) daily closes for the purpose-built 13F-cloning ETFs, each measured over its own trading life against the S&P 500 (SPY) over the identical window (so the comparison is apples-to-apples, not cherry-picked dates).

(Total returns over each life: GURU +406%, ALFA +149%, GVIP +358%.)

3b. Does selectivity rescue it? Cloning PROVEN managers (v1.1, 2026-06-02)

The fair objection to the ETFs is "those clone mediocre baskets — what if you only clone genuinely great, concentrated value managers?" So we tested exactly that, on real EDGAR 13F filings: four managers whose reputations were established well before our test window — Chris Hohn (TCI), Warren Buffett (Berkshire), Bill Ackman (Pershing Square), Chuck Akre (Akre Capital). For each we aggregated holdings by issuer, took the top positions, and rebalanced into them at the 13F filing date (so the 45-day lag is baked in), quarterly, ~2013/2015→2026. Total-return CAGR of the clone vs three benchmarks:

(Benchmarks differ slightly per row because each is measured over that manager's own filing window.)

The decisive result: not one clone — of any manager, at any weighting — beat QQQ. Every clone returned 10–17%/yr while the plain mega-cap-growth index returned ~20%. Versus the S&P it is a coin-flip: only Hohn's clone clearly beat it (+2.7pp) and Akre's top-5 modestly (+1.1pp), while Buffett's and Ackman's clones lagged the S&P. Versus QUAL (a quality factor) it is roughly a wash.

Hohn is the real exception — and it's the most instructive result here. A clone that beat the S&P by +2.7pp/yr for ~11 years just by copying lagged top-10 holdings is a genuinely strong outcome; most investors would take it without blinking, and we should say so plainly. It is also the only one of the four that clearly worked, and the reason is his profile: TCI runs an ultra-concentrated (~10 names), very-low-turnover, long-only-in-substance book of large-cap quality compounders held for years. That is exactly the profile where the 45-day lag barely matters (he isn't trading around it) and where the 13F long book ≈ the actual strategy (no shorts, no private companies, little non-US). Ackman (activism, hedges, a listed vehicle) and Buffett (huge private businesses, insurance float) are the opposite — their 13F is a fraction of what they do, so the clone misses most of the engine. The sharpened, usable rule: cloning is most likely to work for low-turnover, concentrated, genuinely long-only managers whose 13F is their whole book.

The honest counterweight: cloning even hand-picked greats mostly delivers their factor exposure, and we cannot tell from one manager over one growth-friendly decade whether Hohn's clone captured skill or a well-timed mega-cap quality/growth tilt — measured against that tilt (QQQ) it still trailed, exactly as GVIP did. The managers' own records remain far better than any of these clones, because the part a 13F can't show — lag, sizing, non-US, shorts, timing, private holdings — is where the skill lives. Coverage caveat (we ate our own cooking): the handful of unpriceable holdings (Yahoo/Altaba, Allergan, Time Warner Cable, a Liberty tracker) were takeovers cashed at a premium — names we missed, so the true clones are if anything slightly higher, nowhere near enough to close a 3–10pp gap to QQQ. This is an existence check, not a strategy: we picked these managers because they won, which no one could do in advance — the finding is the stronger for it, since even with perfect hindsight on the manager the lag-respecting long-only clone does not beat an honest benchmark.

4. Findings

• Naive/broad cloning underperformed, in the real world, net of fees. GURU (−3.2pp/yr, 14 years) and ALFA (−4.2pp/yr, then liquidated) are the verdict on copying a manager's whole disclosed book.
• The one winner is a beta story, not a skill story. GVIP's +1.4pp/yr comes from concentrating in the mega-cap growth names hedge funds already cluster in — a factor tilt that happened to pay in 2016–26. Unadjusted, that is not evidence of cloning skill; it is the skill-vs-beta trap (you'd need a factor-matched benchmark before crediting any alpha).
• The survivors are thin and the failures closed. A signal whose flagship products keep liquidating is telling you something.
• Selectivity by profile is what matters (§3b) — and it can genuinely beat the market. Of four hand-picked greats, Hohn's clone beat the S&P by +2.7pp/yr over ~11 years — impressive, because following a public filing with a 45-day lag still outperformed. His concentrated, low-turnover, long-only book is the profile a lagged clone can actually capture; for hedge-fund-style filers (Ackman, Buffett) a 13F can't show shorts/private/timing, so the copy lags — not the managers, whose own records are stellar. The edge therefore lives in manager selection (its own discipline), not the copying. None beat QQQ, so skill-vs-tilt attribution stays open — but beating the market by copying public filings is a real, usable result.

5. Conclusion — naive cloning fails; selectivity helps only for the right manager profile

Copying disclosed 13F books is not a broad signal — the ETF clones lost to a plain index fund, and three of four hand-picked great managers' clones failed to beat the S&P. But the exception is real and worth respecting: Hohn's clone beat the S&P by ~2.7pp/yr for over a decade, and the reason is a profile we can state in advance — low-turnover, ultra-concentrated, genuinely long-only managers whose 13F is essentially their whole book (so the lag and the disclosure blind-spots barely bite). For hedge-fund-style filers (Ackman, Buffett) — superb investors both — a 13F simply cannot show most of what they actually do (shorts, private businesses, intra-quarter timing), so the copy lags: a limit of the clone's blind spots, not the manager. And note what the question really reduces to: where a clone works, the edge is not in the copying — that is mechanical and free — it is in identifying the right manager in advance. That is its own science and art (allocator craft, fund due-diligence), every bit as hard as stock selection. We picked Hohn with hindsight; doing it prospectively — telling the next Hohn apart from a manager merely riding a tilt — is the actual challenge. The honest statistical limit echoes it: even Hohn's clone trailed a mega-cap-growth index (QQQ), so over one growth decade we can't fully separate captured skill from a well-timed factor tilt. Practical stance: a proven low-turnover concentrated manager's top holdings can be useful positioning context with the 45-day lag acknowledged — promising for the right profile, but the real work is the manager selection, and we'd not make it load-bearing yet (see ../../concepts/signals-are-overlays.md). This refines our Congressional-tracker verdict: the naive "follow the famous buyer" trade doesn't work; a tightly-conditioned version, resting on genuine manager selection, can.

6. Limitations / the three structural traps

• The 45-day lag. 13F is filed up to 45 days after quarter-end, so a follower acts on positions 45–135 days stale — the manager may already have sold.
• Only the long US-equity sleeve is disclosed. No shorts, no options, no non-US, no cash, no intra-quarter timing. For a hedge fund, the disclosed longs are a distorted fraction of the strategy.
• Concentration. The documented edge is in the top positions only; cloning the long tail dilutes it to nothing.
• ETF wrapper noise. Fees, rebalancing rules and an overlay (ALFA's hedge) add tracking error beyond pure cloning — directional evidence, not a clean alpha estimate.

7. How we would extend this

1. Test the "Hohn profile" prospectively: assemble a small set of low-turnover, concentrated, long-only filers (TCI-like) and clone them out-of-sample from today, to see if the §3b exception repeats when we can't pick the winner with hindsight — runnable now.
2. Factor-regress the Hohn clone on market + quality + growth factors to measure whether any alpha survives once the mega-cap-growth tilt is removed (QQQ trailing is the crude version).
3. Top-position event study: abnormal return after a qualifying manager initiates a top-5 position, entered at a realistic post-filing date (respecting the lag).

Sources

• Pulled ourselves: yfinance total-return histories for GURU, ALFA, IBLN, GVIP vs SPY; EDGAR 13F filings for TCI/Hohn, Berkshire, Pershing Square, Akre — clones vs SPY/QQQ/QUAL (§3b).
• Cited: "Outperforming the Market: Portfolio Strategy Cloning from SEC 13F Filings" (SSRN 5399672); "Systematic 13F Hedge Fund Alpha" (Angelini/Iqbal/Jivraj); Quantpedia "Alpha Cloning: Following 13F Filings"; CXO Advisory "Form 13F Clone Portfolio Performance".

Changelog

• v1.1 (2026-06-02) — §3b: cloned four PROVEN managers (Hohn/Buffett/Ackman/Akre) from EDGAR 13F vs SPY/QQQ/QUAL, lag baked in. None beat QQQ; only Hohn (+2.7pp/yr) and Akre's top-5 beat the S&P. Hohn is the instructive exception — his low-turnover/concentrated/long-only profile is where cloning works; hedge-fund-style filers' clones miss the engine. Selectivity helps only for the right profile.
• v1.0 (2026-06-02) — our own pull of the cloning ETFs (GURU/ALFA/GVIP vs SPY over matched windows): broad clones underperformed (−3.2 / −4.2pp/yr; ALFA liquidated), the lone winner (GVIP +1.4pp) is a mega-cap beta tilt. Verdict CONDITIONAL — naive cloning rejected; conditioned form is context-only.