Robert Carver is an independent futures trader and author. He is also a former portfolio manager of a systematic CTA fund at Man AHL.
Since the dawn of finance there have been big traders and small traders, and small traders have fantasized about predicting the intentions of big traders. If you can guess what a George Soros type is about to do (without using inside information) then front-running their trades is a route to easy money.
The problem is that big traders aren’t always as predictable as the rest of us would like. Most of the good ones subscribe to Keynes’ motto: “When the facts change, I change my mind.” Potential front-runners would prefer the minds of Soros and other whales to remain firmly unchanged. — it would be easier to predict what they might do next.
But there is a class of big traders whose movements might be a little easier to predict: the Commodity Trading Advisor (CTA), also confusingly known as managed futures. CTAs generally run systematically trend-following investment strategies. And systematic traders that stick to the same strategy should be easier to anticipate.
Total hedge fund assets devoted to systematic trend-following are probably around $300bn. More than enough to spark the interest of potential front runners. As a former CTA guy myself my interest was therefore naturally piqued by this recent ZeroHedge article.
The article is apparently “SO GOOD, IT’S FOR PREMIUM MEMBERS ONLY”. But it mostly quotes from various pieces of Goldman Sachs research which model the likely positioning and flow of the CTA industry.
I should say in advance that I have nothing but the greatest respect for the good burgers of Goldman Sachs, and I am sure that their research is of the highest quality. But I find ZeroHedge’s reporting to be . , , somewhat hyperbolic.
Let us take a step back, and consider how we can model the behavior of a hedge fund, or group of such funds.
It’s possible to obtain lagged data about positions from public filings like the 13-F form in the US, or from investor reports. Banks like Goldman Sachs also have access to more timely flows data from their various roles as algo providers, intermediaries or prime brokers. Using simple statistical techniques (or gimmicky machine learning as per your preference), you can model how historical flows reacted to changes in price. Then it’s possible to infer likely trading patterns if a given change in prices occurs.
There are a few wrinkles to consider when we try this strategy in the managed futures world.
First the bad news: it’s much harder to get position data in futures markets, which is the primary tool of many CTAs. You can use the CFTC’s weekly commitment of traders report, but that lumps in CTA exposure with a whole variety of other market participants. Apart from a few UCITS funds that have to publicly report exposures, you are mostly relying on lagged and incomplete private information. Accurate flow data is also difficult to find. Because futures trading venues are centralized, most CTAs bypass intermediaries to trade directly on exchanges, usually with their own internal algos.
But there is some good news. We know that these funds are mostly looking for trends, and it’s relatively easy to create your own trend-following strategy with a spreadsheet or a few lines of code. Pump in some data for historical prices, plus some price expectations, and you too can model CTA behavior. Even fairly simple trend strategies can do a reasonably good job of forecasting flow. Plus if you happen to be working at Goldman Sachs, you can augment a basic trend model with private data on positioning and trades.
A significant caveat: the strategies used by most CTAs are not the sort of technical analysis used by your typical retail punter, where positions are closed immediately once some mysterious ‘pivot level’ is reached. A large fund which closes their entire position in a single day will move the market significantly and suffer massive slippage. This would be a bonanza for potential front runners, but the vast majority of CTA managers are not that stupid. Instead, they continuously evaluate their positions depending on the current trend strength. Any selling due to the market trend turning negative will be slow and gradual.
The Goldman graph in the ZeroHedge article estimated that CTAs are currently long almost $100bn, and could eventually go short by over $100bn if the market sells off enough (apols for blurriness):
Visually, it looks like they are going from their maximum long to (potentially) a maximum short. This happens over a month, which is a plausible horizon given the typical trading speed of most CTAs. Nevertheless, is it really realistic to expect CTAs to dump $200bn of S&P 500 futures in a month?
Firstly, a total US equity position of $100bn when the entire industry manages about $300bn is a somewhat heroic assumption. Okay, CTAs use leverage. But even if we assume a very generous ratio of exposure to AUM of four, that would imply that they have just over 8 percent of their risk capital allocated solely to the S&P 500. Given most CTAs try and allocate across dozens or hundreds of instruments. , representing multiple asset classes, it’s pretty unlikely that the industry as a whole has nearly 10 percent in just one market — even if that market is US equities.
In fact although the ZeroHedge article talks at length about the S&P 500, the headline $220bn in the original GS research actually refers to the entire global equity market. If you peer carefully at a blurry spreadsheet in the article, you can see that the S&P 500 flow in the worst scenario is a mere $44bn. That would imply an industry risk weighing on the S&P 500 of around 2 percent — much more plausible.
Secondly, most CTAs adjust their position according to volatility. To have the same magnitude of short positions on as they currently have long would require the market to sell off significantly, without any increase in risk estimates. But equities are famous for becoming more volatile when they are going down. It’s more likely that the position on the short side will be smaller than the current long, resulting in a smaller adjustment trade.
Finally, it seems pretty unlikely that CTAs are currently at a maximum long position. Even without a fancy trend-following model, you can see from any chart that the market is hardly in the sort of prolonged upturn that would produce a strong long signal, even if it has been a remarkably sedate bear market.
But let’s be generous and assume the $44bn figure is roughly correct. Is that degree of selling likely to cause a massive downward spiral in stock prices? Whilst $44bn sounds like a lot, it’s actually less than one per cent of the monthly volume in S&P 500 futures. Even ZeroHedge’s $200bn figure is about 3 per cent. Any trading by managed funds will be drowned out by a cacophony of orders from the rest of the market.
The truth is that CTAs are mere minnows in most markets they trade. Even if you could predict their trading flows perfectly, it’s never going to be a sure-fire route to front-running profits. Instead, perhaps Soros can be persuaded out of retirement?