Hedge Fund Clone Portfolios: Sizing, Timing, Caveats

What a hedge fund clone portfolio is built from

A clone portfolio has exactly one legal input: Form 13F, the quarterly holdings report that institutional investment managers must file once they cross $100 million in covered US equities. The filing is a snapshot of long positions on the last day of the quarter, due 45 days later. The SEC's 13F FAQ defines the scope precisely, and Investor.gov's summary gives the plain-English version: long US-listed equities plus certain options and convertibles. No short positions, no cash levels, no futures or swaps, almost no foreign listings.

The corpus is big enough to clone nearly anyone: for Q1 2026, Arkolith's dataset covers 1,824 institutional filers reporting 1.87 million long positions worth $53.7 trillion. A clone strategy uses the data twice: the current filing supplies target weights, and the diff against the prior quarter shows what the manager actually did, since 13Fs report holdings rather than trades. If the format is new to you, how to read a 13F filing walks through it line by line.

One cleanup step matters before any weighting math. 13Fs report put and call option legs as separate line items alongside common stock, and a put against a ticker is not a long conviction bet. Strip the option legs, or bucket them separately as hedging context, and weight only the long equity that remains.

Pick managers whose edge survives 45 days

The first design decision is not the weighting scheme, it is the manager. The 45-day lag destroys most of the information in a fast-turnover book and leaves much of it intact in a slow one. If a fund holds positions for years, the portfolio you see in February still resembles the one that exists in March. If it turns the book over monthly, you are copying a photograph of a car that has already left.

Screen for three traits. Low turnover: quarter-over-quarter diffs should show few changes relative to book size. Concentration: a 15 to 40 name book signals conviction you can actually mirror, while a 3,000-line quasi-index gives you an index fund with extra lag. And a long-horizon style: managers whose letters and filing history say they buy businesses, not quarters.

Berkshire Hathaway is the canonical clone target for exactly these reasons, and its Q1 2026 13F shows why: a concentrated book that changes slowly enough for the lag to be survivable. You can inspect the disclosed book on the Warren Buffett fund page and screen the wider filer universe on the investors leaderboard for concentration and turnover before committing to anyone.

The inverse screen matters just as much. Multi-strategy platforms, statistical arbitrage shops, and anything that hedges visible longs with invisible shorts are structurally poor clone targets: the 13F shows one leg of trades whose other legs you cannot see. Cloning half a trade can be actively wrong.

Position sizing: three weighting schemes that get used

Once you have a manager and a cleaned long book, you need a rule for translating their weights into yours. Three schemes cover most real implementations:

Two practical notes. First, the long tail is mostly noise: tiny positions are often residue, hedge stubs, or in-progress entries and exits, and copying them adds transaction costs without signal. Most clones truncate the tail at some fraction of book weight. Second, decide whether you are tracking the manager or the idea list. Pro-rata tracks the manager: if their largest position is 40% of the book, yours is too. Top-N equal weight tracks the ideas while imposing your own risk preferences. Neither is wrong, but switching schemes mid-stream turns your track record into an unreadable blend.

Whatever you pick, write it down before the first rebalance and apply it mechanically. The premise of cloning is that the manager's judgment is the alpha source; override individual weights by feel and you are running your own fund with extra steps and a 45-day handicap.

Rebalance timing: trade the filing calendar, not the quarter

A clone rebalances when information arrives, and 13F information arrives on a statutory schedule: 45 days after each quarter ends. The 2026 deadlines are February 17, May 15, August 14, and November 16 (the full calendar, with the holiday adjustments explained, is in the 2026 deadline guide). Most large managers file at or near the deadline, so the practical rebalance window opens the day after each one, four times a year.

Two refinements are worth the effort. First, do not trade the instant a filing drops. Amendments are routine: filers correct errors, restate line items, and add previously confidential positions, so a clone built on the first version of a filing can be a clone of a draft. Waiting a few sessions and re-pulling the data before executing costs little; the error modes are catalogued in how accurate is 13F data.

Second, use the faster filings between rebalances. Form 4 insider transactions arrive within two business days of the trade, Schedule 13D activist stakes within five business days of crossing the threshold, Form 3 within ten days of someone becoming an insider, and Form 5 within 45 days of fiscal year end. None of these rebuilds a fund's book for you, but they can flag that something changed at a company you are holding on 45-day-old conviction. The differences are laid out in 13F vs 13D vs 13G vs Form 4; Arkolith tracks 51,000+ insider transactions alongside the 13F corpus for exactly this cross-check.

Building the clone programmatically

The mechanics (resolve a manager, pull holdings, diff quarters, compute weights) are retrieval work that a script or an AI agent handles well, provided the data is grounded. Every Arkolith datapoint carries provenance back to its SEC EDGAR accession number, so any weight in your clone traces to the literal filing it came from rather than a model's recollection of one.

# Resolve a manager to a CIK
curl -H "Authorization: Bearer YOUR_KEY" "https://arkolith.com/api/v1/search?q=berkshire"

# Pull the disclosed long book for weighting
curl -H "Authorization: Bearer YOUR_KEY" "https://arkolith.com/api/v1/funds/1067983/holdings"

# Enumerate the filer universe to screen clone candidates
curl -H "Authorization: Bearer YOUR_KEY" "https://arkolith.com/api/v1/funds"

From there the pipeline is short: filter option legs, truncate the tail, apply the weighting scheme, diff against current positions, and emit orders. The quickstart covers minting a key and connecting the same data over MCP if you would rather have Claude or another agent run the loop conversationally.

Keep one invariant regardless of how you build it: the clone's state should always record the accession number and report date it was constructed from. When performance questions arrive later (they will), "which filing was I holding against in September" should be answerable from your own logs rather than from memory.

The performance caveats that belong on the label

Be honest about what the structure costs, because the costs are structural rather than fixable. The lag is permanent: you always hold the manager's quarter-end book from at least 45 days ago, and your entry prices are not the prices they paid. In fast-moving names that gap can swallow the idea entirely. Academic work on 13F cloning has generally found that clones of patient, concentrated managers retain a meaningful share of long-book performance, that the retained share shrinks as turnover rises, and that cloning does not beat the cloned manager.

The blind spots are total, not partial. A 13F shows no shorts, no cash, no leverage, no swaps. A manager sitting half in cash reads identically to one fully invested, so your clone can be far more exposed than the fund it copies. Quarter-end snapshots can also be managed for appearance, one more reason to weight a filing history over any single filing.

Crowding is the quiet caveat. Every clone of a famous book rebalances in the same post-deadline window into the same names, so the most-copied positions tend to get the worst post-filing entry prices. Add the mundane drags (transaction costs, short-term tax treatment on copied exits) and the survivorship bias in guru lists, which by construction show only managers who lasted. None of this makes cloning useless. It makes cloning a sourcing and structuring discipline, not a free ride on someone else's alpha.

Frequently asked questions about hedge fund clone portfolios

How do you build a hedge fund clone portfolio?

Pull the manager's latest 13F holdings, remove option legs and tail positions, apply a written weighting scheme (pro-rata, top-N equal weight, or capped conviction), and rebalance after each quarterly filing deadline. The diff against the prior quarter shows what changed. Everything after that is execution discipline.

How often should a clone portfolio be rebalanced?

Four times a year, in the window after each 13F deadline: February 17, May 15, August 14, and November 16 in 2026. Waiting a few sessions after the deadline lets amendments settle before you trade. Rebalancing more often adds cost without adding information, because 13Fs are the only full-book disclosure.

Do hedge fund clone portfolios actually work?

Academic work has generally found that cloning patient, low-turnover, concentrated managers preserves a meaningful share of their long-book results, while cloning high-turnover funds mostly copies stale noise. The 45-day lag and the missing shorts and cash mean a clone is never the fund itself. Manager selection does most of the work.

Can you clone a hedge fund's short positions?

No. Form 13F discloses long positions in covered US securities only; shorts, cash, futures, swaps, and most foreign listings never appear. Put option legs do show up and can hint at hedges, but the full short book is invisible, which is why long-short and multi-strategy funds make poor clone targets.

This article explains public filings and data concepts. It is not investment advice.