Opinion: Why Most Trend-Following Systems Can Be Improved

By Deb Shaw | on August 16, 2026 9 min read

Table of Contents

Opinion: Why Most Trend-Following Systems Can Be Improved

Trend-following works well across all major asset classes and over two centuries of history. The strategy handily beats buy-and-hold, and especially in protecting downside. But in practice most trend-following funds, looking at the SocGen Trend Index (an equal-weighted institutional benchmark that tracks the daily performance of the 10 largest trend-following funds), generate mediocre returns. Like many others, I found this odd and decided to dig deeper.

Having built hundreds of trading systems over the years, I’ve developed an intuition for why a particular system may be flawed. Looking at the performance of a typical trend-following fund, I had a strong sense that they were over-allocated to commodities and fixed income, and under-allocated to equities and gold.

This makes sense given the history: trend-following is heavily concentrated around futures markets in Chicago. That universe was not designed for long-term investors. It was designed for commodity producers, consumers, and interest-rate hedgers. As a result, the typical Commodity Trading Advisor (CTA) ends up with a book that is structurally heavy in commodities and fixed income, and structurally light in the two assets that have done the most compounding over the last half-century: stocks and gold.

That inheritance explains a lot. It explains why CTAs look wonderful on correlation and skew. It also explains why their standalone returns have been so modest.

What CTAs Are Actually Good At

In a 2025 report, DUNN Capital Management, one of the oldest trend-following firms, made a case that I largely agree with. The value of trend-following is not a pretty Sharpe ratio. It is timing. A useful alternative should be negatively correlated with equities, positively skewed, and profitable in equity crises.

From January 2000 through June 2025, DUNN reported the following for the SocGen Trend Index, which tracks the largest trend-following CTAs, and for DUNN’s own high-volatility WMA program:

SocGen Trend Index

• Annualized return: 5.0%

• Annualized volatility: 13.4%

• Maximum drawdown: –20.7%

• Correlation to equities (annual): –38%

• Skew (annual): +0.6

DUNN WMA

• Annualized return: 7.7%

• Annualized volatility: 29.2%

• Maximum drawdown: –58.4%

• Correlation to equities (annual): –42%

• Skew (annual): +0.8

Over the same window, the S&P 500 compounded at about 7.8%, with a –55% drawdown and negative skew. CTAs earned their keep as a portfolio overlay: they made money in 2008 and 2022, when stocks did not.

The problem is the other column. Five to eight percent a year, with double-digit volatility, is not a compelling standalone engine. High-volatility implementations like WMA improve crisis convexity, but they do it by amplifying the same return stream. You get a bigger hedge. You do not get a better underlying bet.

My view is that this is not because “trend-following doesn’t work.” It is because the typical CTA is following trends in the wrong mix of markets.

A Trading Universe Built for Farmers, Not Compounders

Trend-following as an industry grew up in the futures pits. That is not an accident of branding. Futures exchanges clustered in Chicago because farmers, ranchers, and miners needed a way to hedge the price of wheat, cattle, corn, and metals. Later, the same infrastructure became the natural home for interest-rate futures, because bonds have many tenors and because banks and dealers needed to hedge duration.

If you are a CTA, you trade what is listed, liquid, and scalable. That list is still dominated by:

• agricultural and energy commodities

• government bonds and short-term rates across many maturities

• currencies

• a much smaller set of equity-index futures

Even when a manager equal-risk-weights sectors, the starting universe is still a futures catalogue. Gold is one metals contract among many. Stocks are a handful of indices, not the primary growth engine of the book. The system can go long and short coffee. It cannot, in any concentrated way, do what a long-term investor often needs: sit in equities for a decade, or sit in gold when monetary confidence is fading.

I cannot prove that contract-count is the whole story. Many sophisticated CTAs try not to let soybeans dominate the portfolio. But I do not think you can look at the history of futures exchanges, then look at 25 years of 5% to 8% CTA returns, and conclude that the universe is an accident.

The method, trend-following, is sound. The problem is the catalog of products being traded.

Stocks and Gold Are Not Optional

A trend-following system that cannot allocate aggressively to stocks will miss the most persistent source of real returns in modern markets. A system that cannot allocate aggressively to gold will miss the asset that has repeatedly led when investors start to doubt paper money.

In an earlier post, I showed that even a very crude dual-filter system, rotating monthly among Treasuries, gold, the S&P 500, and cash, already produced higher returns than the SocGen Trend Index over 2000–2024, with far less complexity. It did not match CTAs on negative correlation or positive skew. It did not need to in order to make the point: once stocks and gold are allowed to lead, the return profile changes.

TiltFolio Adaptive goes further. It can be 100% in stocks, bonds, gold, commodities, long volatility, or cash, depending on which regime is actually in force. It is still trend-following. It is not trying to trade 70 futures markets. It is trying to own the one major asset class that is both trending and aligned with the volatility environment.

The Same Window, A Different Outcome

Using the same January 2000 to June 2025 window as DUNN, backtested results for TiltFolio Adaptive were as follows. Correlation and skew are measured on calendar-year returns, matching DUNN’s annual columns.

TiltFolio Adaptive (backtested)

• Annualized return: 15.5%

• Annualized volatility: 14.8%

• Maximum drawdown: –23.4%

• Correlation to equities (annual): –31%

• Skew (annual): +1.2

Adaptive compounded at roughly three times the SocGen Trend Index and twice DUNN WMA. Its volatility sat with SocGen, not with high-vol WMA. Its maximum drawdown was in SocGen’s neighborhood, and far better than WMA’s –58%. Annual skew was higher than both. Annual equity correlation remained clearly negative.

In other words, you did not have to surrender the thing CTAs are famous for in order to fix the thing they are bad at. The crisis-timing properties survived. The weak standalone return did not.

This is the heart of the opinion: most trend-following systems can be improved by changing the universe, not by adding more moving averages.

There is an apparent contradiction worth resolving. On the performance page, Adaptive’s correlation with the S&P 500 over the full history is modestly positive, about 0.14. That is the day-to-day figure from 1992 through 2026. A few paragraphs ago I said annual correlation was –31%. Both are true. They answer different questions.

The +0.14 is the unconditional number. Adaptive is allowed to own stocks. When stocks are the trend, it often does, so of course it moves with the S&P on ordinary days. A permanently negative correlation would mean refusing to participate in the asset class that does most of the compounding.

The more useful question is what happens when the S&P is struggling. On the same daily series:

• When the S&P is above its 200-day moving average, Adaptive’s correlation is about +0.47.

• When it is below that average, the correlation flips to about –0.23.

• In calendar years the S&P finished negative, Adaptive’s correlation was –0.87. In those years Adaptive averaged about +20% while the S&P averaged about –17%.

That is not a permanent anti-stock overlay. CTAs tend to stay negatively correlated even when equities are compounding, which is why they make good hedges and modest standalone engines. Adaptive is willing to look like stocks when stocks are working, and unlike stocks when they are not.

A few caveats, because they matter:

First, SocGen and WMA are live, net-of-fees return streams. Adaptive’s figures here are backtested. That is a real distinction, and I will not pretend otherwise. A 8 to 10 percentage-point gap in annualized return is still large enough to be the story, not a rounding error.

Second, CTAs can go short. Adaptive does not. Short equity-index futures are part of why managed futures can profit in certain crashes. Adaptive’s 2008 was still strongly positive, because it could rotate into the assets that actually worked. The mechanism is different. The outcome, in that episode, was not worse.

Third, I am not arguing that every CTA is naive, or that commodities and bonds should be excluded. Both belong in a regime-aware system. 2022 made that obvious. I am arguing that they should not be the default center of gravity just because that is what Chicago listed first.

What “Improved” Actually Means

If you evaluate trend-following only as a 10% sleeve meant to hedge a 60/40 portfolio, the CTA design makes sense. Negative correlation and positive skew are the product. Weak standalone returns are tolerated because the sleeve is small.

If you evaluate it as a way to compound capital, the same design is harder to defend. You are using a powerful idea, follow price, on a market list that under-represents the assets that do the compounding.

That is why I built Adaptive the way I did. A long-term investor should be allowed to own stocks when stocks are the trend, and gold when gold is the trend, without first diluting that decision across dozens of grain, energy, and rate contracts.

The Takeaway

Most trend-following systems are not failing because the rules are too simple. They are underperforming because their trading universe still looks like a commodities-and-rates exchange with some equity indices attached.

That universe produces the statistics allocators like: negative correlation, positive skew, crisis profits. It also produces the statistic everyone quietly apologizes for: mid-single-digit to high-single-digit returns.

Give the same philosophy a universe that treats stocks and gold as first-class states, and the historical record looks very different. The diversification does not disappear. The compounding finally shows up. TiltFolio Adaptive has only been live for a year at this point, but I’m confident it will continue to outperform over time.

How TiltFolio Works Series

This post is part of the “How TiltFolio Works” series. Explore all posts in the series: