What is trend following? How does it work? Should active or even passive investors have a trend following fund in their portfolio? And if so, which one?

So many questions!

In this post I’ll try to answer a few of them.

Wealth warning: This post discusses some fairly advanced investing concepts. If you’re a sensible regular investor then by all means read it and learn more, but don’t take it as a recommendation to do anything except more research if it piques your interest.


I will use the terms ‘trend followers’, ‘CTAs’ (Commodity Trading Advisors) and ‘managed futures’ synonymously in this article. They are not, strictly, the same thing. But it’ll do for our purposes. 

Broadly we are talking about funds that trade futures and generally have a ‘trend-following strategy’. That is, the funds buy (are long) things that have gone up, and sell (short) things that have gone down. 

For the avoidance of doubt, in finance speak trend following is not really the same as momentum.

When we say momentum we tend to mean a strategy or factor that is long good performers within an asset class (normally equities) and possibly short the poor performers (within the same asset class). 

How do trend-following funds work?

We are going to look at the Winton UCITS Trend Fund to explain how these things work. Specifically we’re going to dissect its January 2024 factsheet

I’ve chosen this fund because:

It’s an exceedingly vanilla trend-following fund, taking its DNA from the veteran fund manager AHL. (David Harding, the proprietor of Winton, was the ‘H’ in AHL.)

You can actually buy it (and I own some)

However, like everything on Monevator – and doubly so the more esoteric or active stuff – this is definitely not a recommendation. And anyway, the fund is a pretty underwhelming offering, as I’ll come to in a bit. 

Here’s what it says on the tin:

This is a very generic description. It would apply to pretty much every mainstream trend-following fund. 

Trend-following secret sauce

So what is the fund’s ‘rules-based investment strategy’?

First, the fund will tidy up the asset price data by turning it into (log) returns and then they’ll apply some sort of volatility normalisation to it.

After that’s done, the rules might look something like this.

Be long when the asset is trading above its (200, 100, 50, 20)-Day Moving Average (pick one for your rule), and short when below.

Be long when the asset is trading above its (200, 100, 50, 20)-Day Moving Average, and short when below, but ignore the last five days.

Be long when the asset is trading above its (200, 100, 50, 20)-day Exponentially Weighted Moving (EWMA), and short when below.

Don’t use the ‘current price’ to measure above / below-ness. Use a short-term EWMA (1, 2, 5) day figure. 

What is the current price anyway? Last, Bid, Ask, Mid? Order Book Weighted mid-price? Ten-minute Moving Average? Of which price? 

Or the fund might not normalise volatility but use some sort of Z-score metric within the return history.

Or any of about a million possible combinations of these rules.

It will end up with something that delivers activity that look a bit like this: 

Now in the real world you don’t use one rule. You might use a handful. That’s because they’ll all give you slightly different results, have correlations slightly below 1, and, since you don’t know what the best parameter choice in the future will be, averaging lots of them is a reasonably conservative position.

Whether to choose what worked best in the past versus averaging lots of parameters/methods that just worked okay is a design decision. 

Refined company

Once we’ve got our signal, we might pass it through a Cumulative Distribution Function (CDF) to give us something like this:

Then I might set my max weight for this instrument to $10m. This means that when I’m max long I have +$10m of exposure, and when I’m max short I have $-10m of exposure. 

Fine. Then someone asks: “Is it really sensible to stay max long when the thing goes parabolic?”

So we might put the signal through some sort of response function, like this…

(Pretend I can draw. )

Which would in turn produced this sort of affect:

But then someone will point out we could actually ‘train’ the shape of the response function for each asset / rule using machine learning…

And so on. This is the sort of thing that quants who spent four years doing a Physics PhD will get up to for the first couple of years after they join the fund. 

Yet even though we don’t know Winton’s secret sauce – and even after it’s done all this clever stuff – we’ll still be able to tell if the fund is likely long or short an asset just by looking at the asset’s price chart.

Back to Winton

Don’t believe me? Let’s consider a few of Winton’s top positions by ‘risk’ and check the charts.

See if you can guess whether Winton will be long or short the following markets?

Data on our data: Unless otherwise stated all price charts in this article are from Koyfin. This up-and-coming data provider is offering Monevator readers a special sign-up offer via our affiliate link.

And here are Winton’s long/short positions – a.k.a. the answers:

Well done, full marks. Not that complicated, is it?

Portfolio construction

So far we’ve only worried about the signal for a single asset. What about portfolio construction?

We’ve already identified that these funds make no attempt to be ‘market neutral’. We can see that clearly if we look at Winton’s sector exposure: 

Winton is long bonds and stocks, neutral in currencies and metals, and short in softs. 

Sounds simple. However under the surface there’ll be quite a bit of clever portfolio construction going on – especially with respect to trying to balance out volatilities between assets so that the fund is taking similar risks in each asset.

For instance, if you want your full signal in asset A to mean the same thing as a full signal in asset B but asset B has twice the volatility, then you’ll only invest half the $ amount in asset B as in A, to get the same ‘risk contribution’.

For bonus points you might even use implied volatility from the options markets to size your positions, given that’s forward looking. 

The choice of which markets to trade is also highly relevant. One decision required is whether to only trade markets that have ‘worked’ (i.e. trended) in the past. Alternatively, you might take the ideological approach that all markets trend, and you’ve just not observed it in the data yet.

You can usually come up with some rationale for whatever you want to decide the data is telling you!

If you take the view that all markets trend, then the more uncorrelated markets you add, the better your performance will be. Your Sharpe ratio will go up with about the square-root of the number of zero correlation assets you add – but good luck finding them.

It might appear in the backtest (before anyone could actually trade them) that ‘Mongolian horse cheese non-deliverable forwards’ are completely uncorrelated with the rest of your portfolio. But that tends to end the day you add them to the real portfolio. At that point it turns out MHC forwards are pretty much just a really difficult and expensive way to trade the Spooz1.

Things are always uncorrelated until your bonus depends on them staying that way. 

Regardless, any correlation less than one is worth adding to the mix – provided that its market is reasonably well-behaved and cheap to trade. 

Rough trade

We haven’t talked about the actual trading bit yet – there’s quite a bit of that going on whenever your signal changes.

First you’ve got to decide how much of a hurry you’re in (i.e. what’s your ‘alpha decay’ profile).

Then you’ll hand it over to a whole other room of quants who do short-term signals to work out whether to trade now or trade later.

Then, once you’ve actually decided to do your trade, you’ll give it to a machine to schedule.

And that machine will give it to an algo, which will give it to a smart-order router, which will finally send it to an exchange for execution.

Suffice to say the likes of Winton know how to do this well (or at least pay a broker to do it).

Are trend following funds any use though?

You’d think after all that clever stuff we just walked through, these funds would shoot the lights out, right?

Honestly, not really. 

Source: Driving with the Rear-View Mirror 

Is a Sharpe ratio of 0.45 good or bad? I guess it depends what you compare it to. Of course nothing beats the Spooz: 4.7% for trend following vs nearly 12% for the S&P is pretty unexciting. Trend even underperforms the Global 60/40 portfolio.

What’s the point of it?

Well here’s the thing: returns are not what you buy trend following for. 

No, what you buy trend following for is this:

Source: Winton (Note: Its Sharpe is overstated here, because this fund hasn’t been around for long.)

Yes: the fund is giving you negative correlation with both stocks and bonds. 

Where does this negative correlation come from?

Well, because these funds can go short, and markets – including equities – trend, then once stocks start going down, trend funds short them. They therefore make money when stocks lose money. They also tend to go long bonds and risk-off currencies when bad things happen.

This is not a guarantee – they can’t see the future. Sudden shocks could leave them long when the stock market is going down.

In theory the perfect trend-following fund to add as a hedge to an equity-heavy portfolio would not trade the ‘long’ signals in equities, only the short. This would make it a better hedge. But it would reduce returns, since stocks mostly go up, which is why in practice you don’t see such funds.

Anyway if you add something with even fairly ‘meh’ returns, to, say, a 60/40 portfolio, that is actually negatively correlated with it, then you will improve its Sharpe ratio – although not its returns.

Whether it is the Sharpe ratio or returns that matter more to you depends on what sort of investor you are.

But before we dig into that, we need to expose trend following’s dirty little secret.

Trend’s little secret: cash 

The futures contracts and other synthetic instruments that trend followers trade are highly capital efficient. They are all, essentially, just a bet on the direction of a thing, not the purchase of the thing.

This means you only need to post a tiny fraction of the notional as ‘margin’.

For example, for the S&P500 ‘E-minis’ futures, which has a per-contract value of $50 per lot, margin is $12,650 per lot.

So with the S&P at 5,000, you’d need to post $12,650 of margin to get $250,000 worth of exposure ($50*5,000), which is about 5%. 

We can see from the Winton factsheet that the ‘UCITS commitment leverage’ is about 640%: 

Naturally, I love this

The 640% figure is a gross sum of all notional exposure (which is an insane way to measure leverage for rates trades, but, this is UCITS so whatever).

Assuming that the margin requirements across all Winton’s instruments are the same as for the S&P500 the fund would need to post:

640% * 5% = 32% margin

Most margin requirements, measured against notionals, are much, much lower than this.

Generally, in a moderately diverse trend-following portfolio, the margin requirements are about 20% per 10% volatility of the fund. And since Winton is actually targeting 10% volatility for this fund, their margin requirements are about 20% of the investors’ cash.

So what happens to the other 80%?

What do you think? It sits in the bank earning interest. 

Now, there are no free lunches in Finance. So that’s not free money for Winton. The financing cost of a position is obviously reflected in the price of the futures’ basis. (It has to be, otherwise you could make free money with the ‘cash-and-carry’ trade).

However, half the time trend followers are short, and hereby earning, not paying, this carry.

And anyway this structure just reflects the reality that the cash you invest in the fund will pay you the risk-free rate plus any ‘alpha’. 

It all means that the headline returns on trend-following funds are higher when interest rates are positive. Because they are mostly just cash!

Of course, none of this makes any difference to the Sharpe Ratio, where we subtract Rf…

…but psychologically it makes a huge difference.

Let’s say I’m buying my trend following fund as insurance for a mostly equities but some bonds long-only portfolio:

If Rf is zero and that insurance costs me a percent in negative returns, then that’s expensive insurance!

But if Rf is 5%, and so the insurance actually earns me 4% p.a. net of fees, what’s not to like?

Of course, this is just  money illusion. Assuming inflation was 0% in the first scenario and 5% in the second, then there’s no difference. In fact, the second case is worse, because I’m paying the fund manager fees on what is just inflation.

The other key observation is that – at a 10% volatility – the fund’s margin utilisation is so low that it could run at much higher volatility than this without a problem. 

Fund VolatilityMargin Utilisation10%20%20%40%30%60%40%80%?

Now, there’s a few operational reasons why you probably wouldn’t want to run 80% margin utilisation. But you could certainly run say 50% – giving your fund a volatility of 25%. 

Why doesn’t Winton? Well, it does, for institutional investors. They can basically do a ‘dial-your-own-volatility’ version of the fund (called a ‘managed account’).

But generally fees scale with volatility. And the higher the volatility the less you need to invest.

The UCITS fund is low volatility because it’s aimed at a somewhat-retail audience that doesn’t really understand this stuff and would be scared by high volatility. And Winton has anyway generally reduced the volatility of its funds as it has removed the performance fee.

In doing so the fund shop is just responding to incentives. As a manager, if you have performance fees you want high volatility, in order to maximise the potential return and hence your take. Whereas if you don’t have performance fees, you want more assets and lower volatility – because people have to invest more for the same return. 

So this is one of my criticisms of the Winton UCITS fund – its volatility is far too low.

The higher the volatility, the less of it I need to add to my 60/40 portfolio to have the hedging impact I’m after.

Adding trend to the 60/40 portfolio

If I’m going to add trend following to my 60/40 portfolio to improve its Sharpe ratio, I’ll clearly need to reduce my allocation to something else to make room. 

Bonds are the obvious candidate. Bonds generally underperform both stocks and trend, and are somewhat there to insure against bad equity markets – which is also what I’m (hoping) the trend-following fund is going to do.

Which bonds should I dial back? Well, the shorter-term ones since, as we’ve already identified, trend is mostly just cash anyway. And what’s the difference between cash and short-term bonds, really?

Let’s consider three portfolios:

Portfolio 1: 60 equities / 40 bonds

Portfolio 2: 54/36/10 AQR managed futures

Portfolio 3 60/30/10 AQR managed futures

Source: Portfolio Visualizer

We can see in the graph that if we replace part of our bond allocation with trend, we get marginally better returns with slightly lower volatility. Also a better Sharpe ratio and much reduced drawdowns.

However I do acknowledge this period I’ve illustrated is too short and too recent. If we’d taken this snapshot in late 2021 we’d have seen a different result.

Do not try this at home

Can we do better? (Those of you who’ve been following me can guess what’s coming here)

That 10% allocation to trend in the above example – we’ve already identified that it’s, like, 80% cash.

So what I’ve really got is a 60/30/2/8 stocks/bond/trend (at 50% volatility) / cash portfolio.

What if – and hear me out here – I took the cash that was inside the trend following fund and… used it to buy stocks!

Then I could have a higher Sharpe ratio and higher returns. 

Now, in the case of the Winton Fund, that’s easier said than done. I could buy Corey Hoffstein’s Return Stacked US Stocks and Managed Futures ETF – which just buys S&P 500 futures with the cash collateral. But most Monevator readers couldn’t, because it’s US-listed. 

Considering the Winton Fund specifically, can I sort of synthetically achieve the same thing?

David Harding presumably doesn’t leave that investor cash laying around in a vault in Hammersmith somewhere. No, he pays it into a bank, where it earns interest.

There’s absolutely nothing stopping me going to that bank and borrowing the money to leverage up the rest of my portfolio – is there?

In fact, it doesn’t even have to be the same bank or the same money. I can simply borrow the same amount of cash as is inside my share of the Winton fund, and net, I’ve not borrowed any money at all.

In fact, my bank could be the futures market – or indirectly the futures market by buying a leveraged ETF.

Of course borrowing costs money – interest – but that’s offset (at least somewhat) by the interest I’m earning on the cash inside the trend fund.

What does this look like then? 

Portfolio 1: 60/40, Portfolio 2: 50/40/10 AQR Managed Futures / -10 Cash

Source: Portfolio Visualizer

Higher returns, lower volatility and lower drawdown (though not much better than simply replacing some bonds with trend to be honest).

Which trend-following fund should I buy?

I’m not going to make a recommendation. There are not many available anyway. And I only invest in funds where I know the principal – so my list is pretty short.

I can however share the ones I own:

Winton Trend Fund (UCITS) – GBP I shares

Fees are a bit high (1.06%) for what it is

Not a very diverse set of instruments

Volatility is a bit low for my taste

Return Stacked US Stocks and Trend ETF (RSST)

Most UK investors can’t buy this because of MiFiD

The trend bit targets 13% volatility

Fee is a more reasonable 1.04%  – so per unit of vol this is the equivalent of 0.8% compared to the Winton Fund

And you get 100% US stocks thrown in for ‘free’

Trend construction is not as sophisticated as Winton or AQR

AQR (I’m in the process of trying to buy these)

AQR Mgd Futures UCITS F GBP (K and C)

Cheap: 59bps 

Mixed reports as to whether you can actually buy this. (I’ve failed once)

Cliff Asness was kind enough to respond to me when I complained about this

AQR Alternative Trends IAG1 GBP Acc

Expensive: 1.8%

Trades all sorts of crazy markets (+)

Very good recent performance (which means nothing)

I would love to hear any other ideas in the comments.

If you enjoyed this, you can follow Finumus on Twitter or read his other articles for Monevator.

The S&P 500 forwards contract.

The post Trend following: is the trend your friend?  appeared first on Monevator.

Receive the latest news in your email
Table of content
Related articles