Bob Elliott

Bob Elliott spent over a decade at Bridgewater Associates, the world's largest hedge fund, where he served on the investment committee and helped manage the firm's flagship All Weather and Pure Alpha strategies. At Bridgewater, he worked directly with Ray Dalio's team on macro research and portfolio construction across economic environments. He left to co-found Unlimited Funds, and their flagship product HFND aims to replicate the aggregate return of the hedge fund industry in an ETF format at a fraction of the traditional fee.

On this episode, Bob talks with Brad about why most investors' hedge fund exposure underperforms the industry aggregate, how HFND reverse-engineers hedge fund positioning using machine learning, and why he believes this approach solves the biggest structural problems with hedge fund investing.

The Hedge Fund Access Problem

Elliott frames the problem simply: the hedge fund industry as an aggregate generates meaningful alpha over time, but almost nobody gets the aggregate. Instead, investors pick a handful of funds, pay 2-and-20 fees, lock up their capital, and face massive dispersion in outcomes. Even institutional investors with teams of analysts dedicated to manager selection struggle to consistently pick the winners. The data shows that the average hedge fund investor underperforms the industry average because of fee drag, selection bias, and the tendency to invest in funds after they've already had their best performance.

HFND's thesis is that you don't need to pick individual hedge funds if you can replicate what the industry is doing in aggregate. The fund uses machine learning models to analyze publicly available data, including 13F filings, COT reports, and other disclosures, to determine the consensus positioning of the hedge fund industry across equities, fixed income, commodities, and currencies. Then it builds a portfolio of liquid instruments (ETFs, futures, and swaps) that matches that positioning. The result is aggregate hedge fund exposure at an ETF fee rather than 2-and-20.

How the Replication Engine Works

The machine learning system processes regulatory filings from thousands of hedge funds to identify net positioning across asset classes, sectors, geographies, and styles. It looks at both the direction of bets (long vs. short) and the magnitude. The model updates continuously as new filings become available, typically on a quarterly lag for 13F data but with higher-frequency inputs from futures positioning reports.

Elliott acknowledges the lag issue directly. 13F filings are 45 days delayed, and hedge funds can change positions rapidly. His response: at the aggregate level, positioning shifts much more slowly than at the individual fund level. The hedge fund industry as a whole doesn't flip from net long to net short overnight. Aggregate tilts toward value vs. growth, US vs. international, or long vs. short duration tend to persist for quarters or even years. The model captures these slow-moving structural bets effectively, even if it misses the fast-moving tactical trades of any individual fund.

The portfolio typically holds 30-50 positions across asset classes, implemented primarily through liquid ETFs and futures contracts. Turnover is moderate because the underlying hedge fund aggregate changes gradually. Elliott notes that the correlation between HFND and the HFRI Fund Weighted Composite Index (the standard hedge fund benchmark) has been consistently high since launch, validating that the replication engine is working as designed.

Why This Matters for Advisors

Elliott's pitch to advisors is practical: most of your clients can't access hedge funds at all, and those who can are paying enormous fees for uncertain outcomes. HFND gives any advisor the ability to add a hedge fund return stream to a portfolio in a daily-liquid, tax-efficient, low-cost vehicle. He positions it as a replacement for the "alternatives" allocation that many advisory firms struggle to implement. No accreditation requirements, no lock-ups, no capital calls, no K-1s.

He also makes the case that the hedge fund aggregate is a genuinely different return stream from traditional stocks and bonds, which justifies the allocation. Over long periods, the industry has generated returns with significantly lower volatility than equities and with low correlation to both stocks and bonds. For a portfolio construction-focused advisor, that combination is valuable regardless of whether any individual hedge fund is worth its fees.

Key Takeaways

• Elliott spent over a decade at Bridgewater Associates on the investment committee, working on All Weather and Pure Alpha before co-founding Unlimited Funds.
• HFND uses machine learning to reverse-engineer aggregate hedge fund positioning from 13F filings, COT reports, and other public disclosures, then replicates it with liquid instruments.
• At the aggregate level, hedge fund positioning shifts slowly (over quarters, not days), which mitigates the 45-day lag inherent in 13F filing data.
• The fund eliminates the traditional barriers to hedge fund access: no accreditation requirements, no lock-ups, no capital calls, no K-1s, and no 2-and-20 fee structure.
• Correlation between HFND and the HFRI Fund Weighted Composite Index has been consistently high since launch, validating the replication methodology.

Listen to the full conversation on Spotify, Apple Podcasts, or YouTube.