When – If Ever – To Meddle With Your Trading System?

Over the years I have been in many debates with colleagues about if, and when, one should interfere with a trading system. Mostly, I have been opposed to intervention as it generally defeats the purpose of adopting a systematic approach. It never ceases to astonish me how ‘systematic’ traders begin to fiddle with their systems when volatility rises or a trading system enters an inevitable drawdown. In most cases, I have found that such interventions achieve little more than medicating trader anxiety about trading losses to the detriment of long-term returns.

Nonetheless, such debate has made me ponder the conditions under which it may be justified to fiddle with a systematic trading strategy. I have distilled this deliberation into seven reasons, of which I believe only about half are justified. Interestingly, most of the interventions I have witnessed fall into the “unjustified” bucket.

So what reasons do systematic traders put forward to meddle with their systems? Let’s dive right in.

1.      Traders Medicating Their Anxiety

The first reason is that humans are simply just emotional animals. No matter how scientific we try to make our approach to systematic trading, there will always be an inevitable loss of sorts. During these episodes, often the most scientific traders among us will revert to their monkey brains. It is so important during these times to take a step back and look at any drawdown in the context of a strategy’s size and historical profile. For some reason, this seems to come to me naturally but seems foreign to many.

This guy obvious has some dandruff and an itchy scalp. Or he is stressed out. Or both.

Many traders fear the potential for losses at times when volatility rises and their perception of uncertainty is heightened. This line of thinking is flawed, however, as the future is always uncertain. Fiddling with trading systems during such times amounts to little more than traders medicating their anxiety about not knowing the future, as opposed to making meaningful improvements. More often than not, such fiddling yields poorer results than just letting the system run.

This line of thinking also invalidates much of the research that went into designing the trading system in the first place. Systematic traders put tremendous amounts of effort and research into designing trading systems, not into whether they have any “edge” in their discretion (not least of all when they’re panicked and under pressure due to trading losses).

2.      Traders Justifying Their Salaries

The second reason traders intervene with systematic strategies is due to the nature of such strategies. They rely heavily on technology and automation. In this environment, many traders feel the need to justify their expensive salaries – and existence – and are somewhat incentivized “to do something” or “be seen to be doing something” by management, like tweak parameters here and there. Generally, this does not generate meaningful improvements. Rather, it creates additional noise. 

I want extra fries with that Big Mac. Please.

3.      Event Risk

The third reason for system interference – event risk – is nuanced. For regular events, such as non-farm payrolls, FOMC meetings, CPI announcements, etc, arbitrarily reducing system risk is counter to using a systematic approach, as such adjustments could have been systematized in the first place. Hence, if reducing risk or changing parameters for such events was profitable, it should be systematically incorporated into the strategy. If not, the strategy should be left to run, as there are plenty of such volatility-inducing events in any good data sample used to backtest a trading system. 

Some may push back on this approach in the case of one-off or irregular events, such as Brexit or presidential elections. This is where the nuance comes in; however, I would still urge against intervention and instead suggest that if your trading positions are a function of realized volatility, incorporating forward-looking measures such as option-implied volatility can help systematically address such risks. 

I believe the aforementioned three types of intervention account for most scenarios where systematic traders meddle with their strategies. I also believe that they are all unjustified. So, when could it be justified to meddle with a system?

4.      Liquidity

There are legitimate reasons to fiddle with a system. Generally, a systematic approach relies on many assumptions. One key assumption is liquidity and the ability of a system to execute in a given market at a given price with some predictable level of slippage. When this assumption is invalidated, the integrity of a backtest becomes questionable. Generally, if liquidity in an asset dries up, this asset may need to be removed from a strategy, which will require some necessary fiddling.

5.      Data Quality

Another justifiable reason for intervention is when the quality of the underlying data becomes questionable. If the data becomes sketchy, it may be better to turn off the system until appropriate data filters are applied or a better, more reliable data source is found. 

6.      Obvious Fundamental Reasons

I’ve always believed it is better to have a strong fundamental driver underlying the reasoning behind a signal. This is what former AHL manager and independent trader Robert Carver calls taking an “ideas first” approach to research, as opposed to a “data first” approach.

Why is this important? Without a fundamental reason as to why a signal should work, it becomes difficult to stick with the signal during periods of underperformance. For example, if a researcher cobbles together a random match of variables and parameters into a signal that appears to be profitable, it is much harder to decide when to turn that signal off compared to a clear, unambiguous change to some fundamentally driven signal. For example, you may have had some currency-driven signal but then a central bank announces it is going to peg that currency. Or, you may have had a signal that relies on regulatory arbitrage, but then the regulations change. Under such circumstances, the reason to turn off a signal are quite obvious.

7.      System Refinements

There is no doubting that trading financial markets is a fiercely competitive endeavour. Not only do you need to act counter to basic human nature to be successful, but you are also pitting yourself against some of the best and brightest minds of the world. While there appear to be some longer-term investment strategies that continue to work through time, mostly by virtue of the fact that they are low Sharpe strategies that are difficult to stick with (such as trend following and value investing), at the cutting edge of the spectrum strategies can degrade quite quickly as technology improves and markets change. In this context, it pays to continually focus on research to offset the steady decay in alphas.

In this respect, implementing system improvements based on thorough research is a form of intervention that is warranted in my opinion. Having said that, I would still stress that the bar for including any new signals or system refinements should be incredibly high, as the lure of profits can make it very easy to fool yourself in this game. In the past I have witnessed sloppy, ad-hoc and inconsistent approaches to research for some researchers, let alone researchers in the same team, and this has been to the detriment of subsequent strategy returns. I would therefore highly recommend developing your own rigorous research protocols to ensure that any system adjustments are genuinely beneficial, statistically valid and are of a high standard.

It Literally Pays to Stay Humble

This guy is probably quite humble.

A final point worth mentioning here is that it literally pays to stay humble in trading. Humans are prone to overestimating their ability and the bar to incorporating new signals into your system should be high. Always remember that when you produce a backtest, it relies on assumptions that may not hold in reality, let alone the future, and that it is essential to have a healthy dose of scepticism about the validity of any results. 

All in all, while I believe most interventions in systematic trading systems are unwarranted and driven by human emotion and the need to justify salaries, there are some instances where they are justified. These include changes in liquidity, data quality or obvious fundamental drivers. In sharing these principles, I hope to stir some thoughts and help traders avoid making unnecessary and potentially harmful interventions.