Bob Elliott

Bob Elliott spent the majority of his career at Bridgewater Associates, developing proprietary investment strategies across a wide range of asset classes. Between him and his co-founder, they have 50 years of building hedge fund strategies and making billions of dollars for clients. Now he's running Unlimited Funds, a firm he describes as "Vanguard for the 2-and-20 side of your portfolio," using proprietary machine learning technology to replicate hedge fund returns in low-cost ETFs.

On this episode of Behind the Ticker, Bob returns to talk about his new 2x target return products, why hedge fund replication is now in its third generation, and how boutique ETF issuers can actually compete without a marketing budget.

The Two Things Nobody Would Say Out Loud

Elliott's thesis for Unlimited comes from two observations he made at Bridgewater that everyone in the industry recognized but wouldn't publicly admit. First: once you reach institutional quality, no single hedge fund manager is meaningfully better than peers over time. Any given year one might outperform, but over long periods they converge. Second: they were all getting paid way too much. The vast majority of hedge fund alpha was being consumed by fees, leaving investors not much better off than they'd be on their own.

The solution: diversify across managers to reduce idiosyncrasy, and slash fees. That's the same playbook that transformed stock and bond investing through indexing, applied to hedge funds. But you can't just invest in the funds directly because then you're stacking fees. So Unlimited built technology that looks over the shoulder of how managers are positioned in real time, translates that into long and short positions in liquid securities, and packages it into ETFs.

Third-Generation Replication

Elliott loves the history of replication, and the evolution matters for understanding what Unlimited actually does. The concept started 30 years ago with Sharpe, who studied active equity mutual funds to understand which factor exposures described their alpha. Andrew Lo picked it up in the early 2000s with rolling regression: regress hedge fund returns against stocks, bonds, the dollar, gold, and infer positioning.

That rolling regression approach worked reasonably well for managed futures, where there's high-frequency performance data and constrained positioning (if the yen is falling, managed futures managers are short the yen). But it failed for strategies like global macro, where managers might hold 30 positions that could be long or short in any combination. You'd need at least 30 months of data to run the regression, ideally 60 or more, and hedge funds change positions far more often than every five years.

Unlimited's third-generation approach uses a proprietary Bayesian machine learning model. The key insight is that managers can't flip positions instantaneously. If you're running any reasonable amount of money, you can't be long stocks one day and short the next. Positioning is path-dependent. Today's portfolio has to be adjacent to yesterday's, which was adjacent to the day before. This drastically narrows the set of plausible portfolios that explain observed returns. The model solves for today's positions in the context of those previous positions, picking up tactical alpha without averaging months of data together. Elliott notes this kind of compute-intensive approach wasn't commercially viable even 5 or 10 years ago.

The Product Suite: HFQ, HFMF, and HFGM

HFQ replicates equity long-short managers, who take net exposures across stock sectors, factors, sizes, and geographies. When you aggregate the "wisdom of the crowd" (some managers long Tesla, some short Tesla), you get quality alpha, especially at 95 basis points for a 2x target return. Elliott shared a comparison on the Unlimited blog: equity long-short managers outperform the vast majority of the top 100 actively managed equity ETFs across sharp ratio, information ratio, and straight returns on a matched time frame.

HFMF targets managed futures, a trend-following approach that goes long when prices rise and short when prices fall. The 2x target return matters here because managed futures is episodic. You get long stretches of modest performance, then extraordinary diversifying periods (like 2022, when stocks and bonds both dropped). If it's a big cash line item at 1x volatility, advisors have to defend the boring periods. At 2x, it's a smaller allocation that delivers the same diversification punch without dominating portfolio conversations.

HFGM is the global macro strategy, which Elliott calls "all-weather alpha." It can go anywhere globally, long or short, across different markets. For advisors balancing diversification against tracking error, global macro sits in a sweet spot: it can keep up in strong markets and protect in downturns. It's their most popular product, crossing $100 million in about nine months.

Strongness Over Optimization

One of the best exchanges in the episode was about the difference between academic strategy development and real money management. Elliott's approach to opportunity sets is instructive: rather than optimizing which combination of 200 possible assets produces the best backtest (which is just noise), he writes down the top 20 assets that matter on a blank sheet of paper. Five equities, five currencies, five fixed income, five commodities. No empirical optimization. It looks worse in backtests but produces far better real-world results. He calls it "comprehensive but parsimonious," and says no academic would build a strategy this way because it fails traditional academic tests, even though it's a much better way to manage money.

Both Brad and Bob agreed: 95% of explorations to improve a strong systematic process fail, and that's actually a good sign. It means you haven't over-optimized. The worst thing you can see in someone's backtest is something that looks too good.

Key Takeaways

• Unlimited's Bayesian machine learning model exploits the fact that hedge fund positioning is path-dependent, solving for today's portfolio in context of recent portfolios rather than averaging months of data through rolling regressions.
• The 2x target return products (HFQ, HFMF, HFGM) run at equity-index-level volatility for 95 basis points, making hedge fund strategies cash-efficient enough that advisors don't have to defend large portfolio line items during quiet periods.
• HFGM (global macro) has been the breakout product, crossing $100 million in nine months, because it offers "all-weather alpha" with diversification potential and less tracking error than pure defensive strategies.
• Equity long-short hedge fund managers outperform the vast majority of the top 100 actively managed equity ETFs across multiple performance metrics on matched timeframes.
• Elliott's advice for boutique issuers: stay small, stay focused, don't blow through millions on sales and distribution. Use live RIA holdings data to target advisors who are most likely interested instead of spamming everyone.

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