Using Prediction Analytics Through the Market Crash – Some 20/20 Hindsights

DataSwarm Trading 6 Months Prediction Problems

Chart shows progress of fund from start in December 2019 until end May 2020, tracking market performance from portfolio buildup into the trading stage, and then navigating the market crash and rebuild

We talked in more detail about what we have done trading using the Zeitgeist signal over here. There were three “interesting” phases from the point of view of using a predictive analytics approach that we have been asked about quite a bit, shown in the chart of our performance to date (we are the blue line). Here are some thoughts:

1. Did we see the Crash coming?

The answer is yes and no:

  • Yes, the market seemed to be at an all time high and many people were thinking “this can’t last” and we were picking that up, but timing, as they say, is everything – and we didn’t know what might trigger it.
  • No, in that although we did see Coronavirus rising in the East, the market only crashed when it hit Italy (hard) and the crash was almost immediate – it was the fastest drop ever.

Also, if you do the probability maths, it is usually better to be all-in in a market rise and exit on the drop, but be aware the Black Swan can drop stuff on your head at any time.

2. How do Prediction Analytics work in Turbulent times?

  • The answer is “Interesting”. In our case it was clear that something was happening on Friday 22nd Feb (the crash continued over the next 3 weeks), and the system was keen on selling quite a few of the stocks straight away – but it wasn’t yet clear how serious it would turn out to be.
  • During the crash the signals of the stocks we held were all doom-filled, and when prices were bouncing around on the bottom most signals were confused.  We have since had to do quite a lot of work since to both decrease system sensitivity (crash time) and increase it (bobbing on the bottom time).
  • However, what we did see fairly early was that system signals were pointing us at stocks that were worth buying, and these signals suggested a strong move into new stocks such as Covid drug companies, so we effected that shift and that’s where you see us pulling out and climbing again on the chart.

3. Did you expect the bounceback to be so rapid?

  • The answer is again, yes and no.
  • During March/April we did some system dynamic modelling of the Virus so had some understanding of how it works and could predict likely outcomes (or at least have some view of how it may play out). We actually gave a higher % probability of a worse case outcome than has occurred, so we were quite conservative in April positioning.
  • As the bounceback occurred the predictive analytics started to firm up, pinpointing sectors and companies within those sectors to buy into and ones to exit – thus shifting our portfolio.

 

The Outcome

The DataSwarm Virtual Fund has managed a 12.6% gross return over 6 months – compared to an S&P loss of -2.3% (see chart at top of page).

 

An Update for Yesterday’s drop

After we wrote this, and before publishing it here, the market had a major drop yesterday – did we see it coming? Well, I’m afraid the answer is pretty much the same as at the top of this article, with 2 exceptions…

  • Our reaction to that one was mild, we have changed the algorithms a bit to sharpen it up.
  • We give a higher % probability of another big crash today, so have held more in reserve.

(Turns out it’s on its way up today so far. We shall see……)

 

Would you like to use the DataSwarm system?

 To understand the DataSwarm system better, test it, and see it in action, why not sign up for a free or paid monthly subscription account?  Our system is live and allows you to subscribe to our basic service.  Sign up to explore the financial Zeitgeist.

For further insights, a system demonstration of our or more advanced trading services there is more information here or contact us at info@dataswarm.tech