Navigating a market correction with DataSwarm predictive analytics

Correcton Recovery

(Graphic – DataSwarm fund performance vs S&P500 and NASDAQ over September)

On the 1st September the DataSwarm Fund rose to more than 30% above its pre-Covid peak in February, due to a few of the stocks in its portfolio rising due to good earnings. On the 2nd of September, the US markets reached an all time high for 2020.

On the 3rd of September, the market tanked. This was the start of the September “correction” when the entire market dropped by over 10% at its deepest.

So what does a crash look like from the point of view of a predictive analytics system? This is how the DataSwarm system saw it.

Late August – conflicting signals

In the last third of August the system’s indicators showed contradictory signals, the overall market Zeitgeist was still rising, but the Zeitgeist for our portfolio was dropping vs the market indices. Also there was quite a lot of volatility on daily sell suggestion numbers. So despite rising market optimism, our forward view was getting worse.

But these down signals were too weak to justify changing our overall strategy, and besides more people lose money by selling early than selling late, so we carried on buying into the rise. We did start “profit taking” however, given the sell signals were rising.

Sep 1st -2nd   – market hits peak

Initially it looked like we made the right call – we ended August on an all time high, and as mentioned above we shot up even further on Sep 1st  and the markets hit new heights on the 2nd

But what goes up (so fast)….

Sep 3rd – 8th – market drops 10%

The markets dropped by 5% on Sep 3rd, but we dropped by 7%, and the next day saw another 1% drop in the markets but we went down 2 %. Those stocks of ours that had gone up so high dropped like stones in the panic selling on the day! (You can see the spike in the chart above)

On that first day the system was again recalibrating but by the second day of the drop (4th Sep) it was suggesting some stocks to exit, and we started to sell off. But it also wanted to buy. Some stocks were going up despite the crash, and it wanted in. Also, some had dropped hugely on the 3rd and it wanted in on those too. So we were buying as well.

Over the September long weekend the system signalled the market would drop again. It did, the system said sell in volume, and we dumped the stocks. The market went down 4% on the day, but by selling off on the 4th and early on the 8th our rate of fall slowed markedly (when we sell we go into cash, and cash doesn’t fall in the short term).

Sep 8th – 18th Stabilise

Over this week the markets continued to drop but as you can see from the graph we had stabilised and (despite the bumps) kept level.

However it was not all plain sailing, some of the stocks that we had bought the previous week (because they had fallen and appeared good value) just carried on falling.

This is an important point to bear in mind with predictive systems – they (nearly) all work on probabilities, and will get some things wrong. It’s thus necessary to calibrate the risk of a probability going wrong, and sometimes those can go badly awry, and some mistakes can cost more than you expect. For example we got badly caught by TSLA – bought in after it had fallen by 16% against a market fall of 6%, but  it still fell another 20% (predictions can be wrong). It’s not a cheap stock, so we really felt that (high risk if wrong)!

Sep 11th – 18  Recycle

As noted above, the system was buying new rising stocks from the beginning. In this period it started to buy back far more, as well as continuing to sell poorly performing stocks.

This recycling meant increasingly more of “our” stocks were gainers, and the fallers were reducing, so our overall position started to rise even as the market continued falling.

Sep 18 onwards – and upwards

We were rising from the 18th. By the 24th the markets had hit their nadir, and slowly started to rise.

Note that, while cash gives bouyancy in a falling market as it doesn’t drop in value, the reverse occurs in a rising market as cash does not rise in value. The market indices are always 100% invested in stocks. (ie if you are 30% in cash, your stocks’ descent is slowed by 30% in a falling market – but your stocks need to perform 30% better in a rising market just to keep up with it).

But our system only signals “buy” for stocks it believes are very good value, so it takes time to build up its stock position again. We have to trust it to pick stocks that outperform the market to mitigate the cash % we were holding after selling off.  At any rate – as of 2nd October, a month after our Zenith spike – the system has outperformed the market, gaining 4% extra so far. And every stock it buys reduces our cash position, and should accelerate the gain.

It has been an “interesting” ride to say the least. Not quite at the level of the February – March crash  but there have again been some very useful lessons (and some a bit painful). But overall, more evidence that the system works.