Your firm’s strategies are based heavily on algos. How do you develop them? Do you ever source anything externally or is everything in house?
AS: We do all the research and development internally, and we have several independent teams that are responsible for different parts of the process so that idea generation is semi-autonomous. We also have a proprietary method for evaluating the robustness of a system.
Are there any particular issues that come up in developing everything in house?
AS: One constraint is just the development. Obviously we could do more and faster in terms of just coding if we outsourced more. However, we like to control that process.
Different firms often tend to have different overall preferences when it comes to strategies and systems. What kind of strategies do you tend to favor?
AS: We have an open view and really go where the research leads us. We like to blend. Not only do we not have only one research or strategic approach, but we like to blend non-correlated strategies. In practice, we have short-term trend and momentum based systems. That’s one bucket. The next bucket is mean-reversion or intraday reversal systems. The third bucket I would categorize it loosely as pattern recognition. And then, unlike most CTAs or systematic or quant firms, we have a discretionary macro strategy implemented by the trading team. In addition, the allocation between these four buckets is a discretionary process.
I like to describe what we do as “managed systematic”. That means that our head of trading actually has discretionary trading ideas from time to time that are not generated automatically. Though the discretion is mainly risk-off, we see this activity as further diversification. At certain periods, we can apply common sense to a trading situation that you don’t see with the pure systematic firms.
As a case in point, most trend following systematic companies had massive returns in the first quarter of the year. And rather than just locking in profits, many had significant drawdowns in the last week of April and in the beginning of May. Trend followers were particularly hard hit. It’s just an example where a little bit of common sense could be applied to smooth out the curve.
I know what I’m saying is considered total heresy in the quant space, but I don’t mind swimming upstream!
Speaking of discretion, your firm is based in Switzerland, which is where a massive shock for the FX market occurred this year. Were you affected by the SNB news?
AS: That did not affect us because we did not have anything in the euro-franc pair. We had looked at the euro-franc pair, but the risk management committee rejected the pair outright so we had zero exposure there. That was a discretionary decision which obviously in hindsight was a good one. Overall, Pecora’s Quantitative Diversified Program returned 0.44% in January, so we emerged unscathed.
In opting to have both systematic and discretionary overlays, how do you break it down? How much is automated at the moment and how much discretionary?
AS: It’s 80% systematic and our natural preference is to be systematic. That’s our background, that’s what we prefer. We just feel that we can have further diversification with the discretionary trading strategy on top. Since we employ trending systems, mean reversion systems and pattern recognition systems, we have defined a method of allocating amongst those systems that accounts for the prevailing market environment.
How long have you been trading currencies and how has the algo scene in FX evolved in that time?
AS: We’ve always traded currencies. The core trading system incepted in 2008 with exposure to the five largest currencies. In addition to that, we’ve evolved into a global macro strategy which includes 40 of the most liquid futures markets.
During that time, have there been important changes in market structure, conditions or the way business is done that affect you?
AS: FX as an asset class has been in a liquidity vacuum due to overall lack of interest rate differential globally. So it’s been a difficult space up to last year. If you look at probably April last year, that was really maybe the low in FX volatility. And a lot of small players left the business. Of course at the bottom, that’s always the best place to get in, right? But finding the bottom is tricky. Since then we’ve seen a surge in FX volatility with the uptrend of the US dollar. We had been predicting the surge of the US dollar as an asset overall for some time now. I had called that on CNBC on March 27, 2014 when the dollar index was around 80, where basically we had been expecting dollar strength last year. At that time, I recommended to sell Euro at 1.37. However, our strategies are quite versatile and, therefore, in May we profited from reversals in US dollar, where we had longs in EUR and Cable. In terms of the structure of the market, it’s becoming more efficient. Trading platforms and matching liquidity and electronic execution is improving. So, from an IT perspective, it’s improved dramatically over the last five years. On the contrary, from a regulatory standpoint, FX as an asset class has not improved at all. You have large banks getting fines and all this destroys trust in the minds of smaller traders. That hurts liquidity.
So from a technical perspective the market has improved, but from a regulatory, or just an overall confidence perspective, I would say that we might be stuck in neutral.
The journey from idea to live trading
In terms of backtesting and risk management, how do you approach those issues?
AS: I have a general overview and review the work product of our research department. I tend to be very skeptical; all backtests are guilty until proven innocent by real trading. One has to be rigorous in accounting for transaction costs and slippage. It all starts with data recording and storage. Then we focus on backtesting, simulation and live trading before a system can be added to the Quantitative Diversified Program. We test all our systems with proprietary capital before graduation to client portfolios.
So you’re looking to punch holes in something until it can hold up?
AS: That’s right. When I see a backtest, my rules of thumb is, cut the average return in half and double the volatility that you see, and that would generally, in my experience, be most realistic. What happens, especially in FX, is it’s very difficult for backtests to properly account for transaction costs. That seems like it would be straight forward, but it is not because of the overnight interest and also the fact that prices in FX are really more theoretical. Just because you have data doesn’t mean that you would have gotten that price if you had done that trade. It all depends on who you’re trading through and what venue you have and all these things.
The reliability of backtesting, specifically in FX, is I would say very, very low. What I do is, first we start with a backtest and we use our proprietary method for determining the robustness and to make sure that the strategy is NOT curve fit to the data set. That’s number one. Secondly, we look and we can see if it can be applied generally or if it only works on a specific pair and it doesn’t work on other currencies. That for me is a red flag that shows you may have some over-optimization.
After we look at ideas and consider best in class trading systems, the first thing that we will do is start paper trading in real time. That is in order really to just get some experience and feel for the strategy. After that we’ll allocate partner money and trade the strategy live, with real assets. All of our systems, they must be live, traded in real assets, for a significant period of time, certainly more than a year, before we would ever incorporate a given strategy into the overall portfolio.
What sort of returns are you able to generate with this approach?
AS: The Quantitative Diversified Program, has average returns of 16.99% since 2008, with standard deviation of 9.5% and maximum drawdown of 12%. It’s never had a down year.
If I can bring in William, could you talk about your role and the approach you take?
WA: I manage the entire portfolio and lead the ongoing research, development and evaluation of existing models and new strategies. Aaron has explained how we have achieved broad, but viable diversification. FX is the mainstay and we have introduced other asset classes into the model. Now that the onboarding is more or less complete, we are beginning to realise some positive results.
“The primary challenge or limitation that most traders face is gaining a proper view of the market when it comes to execution and pricing.”
Twin priorities: capital protection and alpha
You’ve spoken a lot about signal generation and strategies. How about on the execution side? How do make sure you get best execution and minimise market impact?
WA: That’s our secret sauce. In other words, it’s proprietary. Aaron and I both matriculated from institutional grade environments. The primary challenge or limitation that most traders face is gaining a proper view of the market when it comes to execution and pricing. We really don’t have that challenge for a number of reasons. The first is that we’re not a high frequency shop. Some other managers may tend to be higher frequency so their edge generally comes from having a low execution fee while being able to execute trades within a tight range.
That’s not our strategy. When you take that approach, order placement and access to market can limit your efforts, especially with trading size. Occasionally, there will be some skewness on price and there will be a lot of slippage. Our execution method is more defined and even though we have several overlays that allow us to take a specific view on the market, we’re in a position to trade spot or in various parts of the calendar or curve tactically.
We basically have an orientation that prioritises capital preservation. Institutional concerns are really accentuated by two things. First, protection of the capital. So when you trade in an environment that doesn’t have a big appetite for high daily volatility, one has to be very efficient with order placement. Consequently, we bias the system towards risk management strategies and capital preservation. That’s the first priority.
The second priority is alpha. Since we predicate our approach based on the initial premise of protection of capital, that means we’re able to take exposure in different ways in the market that allow us to participate in the move but also to a greater extent we get to filter out the noise because we don’t have a directional bias in any regard for any market. So, we can be long and short on a trade at the same time and take advantage of that potential as well. The approach depends on the prevailing market, current volatility, and the underlying signal.
Circling back to execution, how do you think it differs in FX from other markets?
WA: Our approach is mainly reliant on price. That’s the most important number for us. That’s why I explained the risk element. Every trade has its own risk budget. We also segregate those markets amongst several baskets. We have core basket markets and we have satellite basket markets. The allocations for those markets vary. Then each one of those markets within those baskets has a specific trading band.
Just for sake of example we could say in the core basket it could go anywhere from 50 basis points of risk to up to three percent. I’m not going to get into the specifics of how we do it but just to give you an understanding. Sometimes we stagger positions so we’re not taking a full allocation on anything important at any one point in time. The market activity generally determines how much exposure we take to a specific market.
Since we rely so heavily on price as our prime indicator and use price action to develop various metrics for measuring volatility, we work with limit orders. If the market never gets to that price, we’re not going to execute. So in this sense, we don’t have any challenges entering and exiting markets.
From a technological point of view – looking at hardware, software, platforms etc – how have things evolved for you? What would say have been some of the most important developments for hedge funds?
WA: We have redundancy – it’s probably n+4 on the trading side. We have the heavy algo and number crunching happening on separate PC’s.
Currently, I can manage approximately 85% of the trading process on an iPad. I just finished a big tour in China and we had no issues at all. That was a great test. That’s the level of redundancy we have. We’re now able to keep a full view of the portfolio even when we are on the move.