Is trend following dead?
I get asked this query at the least once per week. As those of you which have met me IRL ('in real life') will recognize I even have restricted endurance and I'm without difficulty bored. I'm definitely bored of answering this query. This submit is the ultimate time I'll answer it.
There are extensively two approaches to answer this query:
- Looking at fundamental reasons why trend following is less likely to work
- Some kind of statistical analysis
Are there fundamental motives why trend following might not paintings any greater?
Some humans spend their whole lives opining about why this or that method no longer makes sense (google 'is value/ momentum/xxx useless' and spot what number of results come back). Personally I discover that a very unnecessary occupation.
I've usually felt it is very difficult to forecast the future, and your first-class bet is to hold an publicity to a varied source of different return factors. You'd be mad to have a hundred% of your capital exposed to fashion following (it is about 15% throughout my entire portfolio). However you'd be similarly mad to have 0% of your capital exposed to it due to the fact you believe you studied it become dead. In this oft quoted current interview with David Harding, he became reducing the exposure of his fund within the method in half, to twenty-five%.
When I have occasionally checked to see if exogenous conditions can be used to predict trading strategy returns I've found very weak effects if anything at all, as in this post on the effect of QE on CTA returns. Also; trend following returns tend to have negative auto correlation for annual returns (alluded to in this blog post). So bad years tend to be followed by better years.
Before my patience is tested to it's limit let me quickly discuss just two of the reasons why people think trend following is dead:
- Strategy is overcrowded; possibly but trend following is mostly a self reinforcing strategy, unlike say relative value strategies where profits get squeezed out when investors rush in, having more trend followers causes trends to last longer. Having said that overcrowding is potentially problematic when trends end and numerous investors rush to the exits, especially as there are other players like risk parity investors whose behaviour will closely resemble trend followers (see February 2018). It's worth reading the excellent work of Dr. Robert Hillman on this subject.
- World is unpredictable (see Trump, also Brexit): perhaps this is true, but this unpredictability also affects discretionary human traders - I doubt any human can predict what Trump is going to do next (that probably includes Trump). Also trend following as a strategy has been around a long time, and on average it's worked despite the fact that there have always been unpredictable factors in the world. I'd be more concerned about a strategy that worked really well in the Obama presidency, but hadn't been tested before that (especially as the Obama presidency was a strong bull market in stocks).
Can statistical evaluation tell us if trend following is lifeless
Statistical evaluation is first rate at telling us approximately the beyond. Less beneficial in telling us about the destiny. But possibly it could tell us that trend following has genuinely stopped operating? I'm keener in this technique than thinking about fundamentals - as it's a beneficial workout in knowledge uncertainty. Let's find out.
First we need a few records. I'm going to apply (a) the SG Trend index, and (b) a again-test of fashion following method returns. The gain of (a) is that it represents actual returns by using fashion fans, at the same time as (b) is going lower back longer.
Heres the SG trend index (month-to-month values, cumulated % returns equivalent to a log scale):
It in reality looks like things get as a substitute unpleasant after the center of 2009.
Heres a backtest:
 |
| Backtest of 3 similarly weighted EWMAC policies over 37 futures |
And for comparision a zoom of the backtest considering 2000 to match the SG index:
 |
| Backtest of 3 similarly weighted EWMAC policies over 37 futures, since 2000 only |
The backtest via the manner is simply an same weight of 3 EWMAC guidelines, similarly weighted across 37 abnormal futures units in my dataset generated the usage of pysystemtrade with the followingconfiguration.
In phrases of answering the query we can reframe it as:
Is there statistical evidence that the overall performance of the strategy is bad inside the last X years?
Two open questions then are (a) how do we measure overall performance, and (b) what is X?
I typically degree overall performance the usage of Sharpe Ratio, but I assume it is more appropriate to apply return here. One function of trend following is that the vol isn't very stable; it tends to be low when dropping and excessive when earning profits. This effects in very awful rolling Sharpe Ratios in down periods, and now not so proper rolling Sharpes in the top instances. So just this as soon as I'm going to use go back as my performance measure.
In phrases of the term we've the usual competing consequences; short periods may not deliver us statistical significance, longer periods won't assist take a look at the hypothesis of whether or not fashion following is currently useless. The recent length of terrible overall performance began either in 2009 or 2015 depending on how a ways again you need to move. Let's use a 2 yr, three yr, 5 year and 10 12 months window.
What I'm going to be chickening out is the rolling T-statistic, trying out towards the null hypothesis that returns have been 0 or bad. A high fine T-statistic suggests it is probable the returns are notably wonderful (yeah!). A low bad T-statistic shows that it's in all likelihood that returns are appreciably poor (boo!). A middling T-statistic method we do not without a doubt know.
Here's the python:
from scipy.Stats import ttest_1samp # dull wrapper feature as pandas apply functions must go back a floatdef ttest_series(xseries): go back ttest_1samp(xseries, zero.0).Statistic # given some account curve of monthly returns this will return the rolling 10 year
# collection of t-statisticsacc.Rolling(a hundred and twenty).Observe(ttest_series)
Incidentally I also tried bootstrapping the T-statistic, and it didn't affect the effects very an awful lot.
Here are the consequences. Firstly for my returned test, 2 12 months rolling window:
Let's study three years:
Surely 5 years have to display considerable outcomes:
Okay, allow's switch to the SG CTA index. Starting with two years:
Some can also accuse me of straw-manning here; "pay attention Rob we're not pronouncing fashion following is so damaged it'll lose cash; simply that it hasn't and might not do in addition to in the past". Well looking once more at those rolling plots I see no evidence of that eithier.
Looking at the SG index there has perhaps been a moderate degradation in overall performance after 2009, however taking the long time view over the backtest I'd say that during the last 30 years at the least performance has been very similar and the current length of negative returns is in no way as horrific as matters have were given within the beyond earlier than improving.
One greater VERY IMPORTANT POINT: It's arguably silly to observe the overall performance of any trading approach in isolation; like I stated above handiest a moron might have one hundred% of their money in trend following. One of the arguments for fashion following is that it gives 'crisis alpha' or to be greater particular it has a poor correlation to other assets in a undergo marketplace. Unfortunately it is definitely impossible to say whether trend following still retains that assets, considering that (he stated wistfully) there hasn't been a decent crisis for 10 years.
You ought to be satisfied to spend money on crisis alpha although it has a anticipated go back of 0 over all history - arguably you should also be glad to pay for it is coverage homes, and put up with a slightly terrible return. Since the 2009 fashion following has added some modestly advantageous overall performance; arguably higher than we have a right to count on. We may not realize for certain if trend following can nonetheless deliver until the following big disaster comes alongside.
Summary
"Is trend following dead?" I don't know. Probably not. Now leave me alone and let us never speak of this again.The subsequent man or woman who asks me this question will get a deep sigh in reaction. The one after that, a complete eye roll. And with the third man or woman I will have to hotel to physical violence.
