Put stochastic vibration analysis to bed

Listen to this article

00:00
00:00

My children asked me to read them a bedtime story. I am a lazy father, and grow bored reading their books. I usually feign a sore throat to avoid this duty.

But when I saw the book they pulled off their shelves, I could not resist it. It was called Hedgie Blasts Off!, by Jan Brett.

I had never heard of the book or the author, and had no idea how they got hold of it.

It begins: “My name is Hedgie and I want to be an astronaut. I’ve never flown on a spaceship, but I take care of the Zeppadoppler rocket for the Professor, the smartest scientist on Earth. He’s in charge of outer space.”

The beginning of the book reminds me of a meeting I had a few weeks ago.

I was breakfasting at the Four Seasons with people representing a family that had built their fortune sometime between the 15th and 17th centuries – and had managed to let it accumulate into tens of billions in random Swiss banks. They were evaluating the latest wave of emerging hedge funds.

“We just saw an amazing hedge fund,” one gentleman said to me. “The guys worked on the space shuttle and spent $2m taking their stochastic vibration analysis techniques and using them to dominate the markets. They have 50 computers running algorithms.”

“Stop,” I said. “Have you read the papers lately about the space shuttle?”

Everyone laughed.

“Seriously, whatever they are telling you about vibration analysis is a sham. Maybe they are legit, I have no idea. But nobody needs to spend $2m developing space shuttle vibration techniques to beat the markets. That’s just marketing. That’s their pitch to get people suckered in.

“It’s like when academics in the late ’80s built ‘neural networks’ to detect space missiles, so that they could get funding for Star Wars research from the Defense Department. There’s no such thing as a neural network. It’s basic statistics with a fancy name on it. It was just marketing.”

Here is an example of a simple system that historically has done well in both bull and bear markets and does not apply any quantum mechanics or chaos theory. It takes advantage of the fact that the so-called dumb money tends to pile into a stock when the trend has played out.

About five years ago, a specialist on the New York Stock Exchange taught me this trick. He said that if a company has bad news, wait three days, then buy.

I tested the system on all the Nasdaq 100 stocks over the past 10 years. I included stocks that had been deleted in order to avoid survivorship bias. I took into account commissions.

This is how I did it. If a stock is down three straight days, buy the next morning at the open. If you own a stock and it is up two days in a row, sell it the next day at the open.

The idea is that trends run out fast. After two consecutive days up, it is reasonable to expect a stock to slow down. After three down days in a row, all of the sellers are done.

This is not always true. Often a stock goes down many more days than three. When it does, in this system, you buy more. But if it goes up two days in a row, you sell.

I tested from 1997 to 2007. I simulated using only 2 per cent of equity per trade.

If you want to do risk management, do not use a stop-loss. Keep position sizes small. Stop-losses will almost always degrade the results of any system.

You have to give your system a chance to play out. But if you are nervous about a stock going down, say, 50 per cent, keep position sizes small and make as many trades as possible.

There were 16,777 occurrences of this trade triggering. 10,706, or 64 per cent, resulted in profitable trades. The average return per trade, including winners and losers, was +1.93 per cent, and the average holding period was six days.

There were no down years. Bull market years like 1999 performed best, with a +200 per cent result in the simulation. But bear market years like 2000 and 2001 returned 40 per cent and 80 per cent respectively, and 2002 returned +3 per cent.

Long-only, mean-reversion systems love volatility, and the bear market had a lot of volatility.

Some of the stocks this system has worked well for recently include BAE Systems (18 successes out of 21 trades since 2004), Electronic Arts (16 out of 20 since 2004), and Genyzme (19 out of 22).

Should you play this system? I do not know. It is at least worth studying. Before anyone tries to sell you on a $2m system, it seems as good as anything else.

Should you sell short using this system? If a stock goes up two days in a row, why just sell? Why not short?

I tested a variation on this. Let us say the QQQQ (the ETF representing the Nasdaq 100) goes up two straight days and the next morning gaps up in what I call the “sweet spot” – between 0.3 per cent and
0.6 per cent.

Let us short the sweet spot. Let us hold until either the close, or (let us not be greedy) take profits if QQQQ goes down 0.5 per cent. Simple. No chaos theory allowed.

There have been 71 occurrences since the QQQQs opened for trading in April 1999: 65 successful occurrences (92 per cent) for an average gain of 0.36 per cent. Bull market or bear market, this approach has generated positive returns every year.

Returning to the bedtime story, it took me about 10 minutes, but I eventually finished reading the book to my children. It concluded with the words: “Three cheers for Hedgie! Hooray!”

james@formulacapital.com

Copyright The Financial Times Limited 2017. All rights reserved. You may share using our article tools. Please don't copy articles from FT.com and redistribute by email or post to the web.