Algorithm trading is no oil painting

If you have been paying close attention to the stock market recently, you may have seen a strange phenomenon that you couldn’t quite put your finger on. The reason that you can’t put your finger on it, is because the phenomenon that I talk of has become so commonplace that it actually no longer feels like a phenomenon; it feels like any other day in the market.

What I’m talking about is unjustified or unusual volatility in the market. In other words, it’s when the pricing of the index quickly increases or decreases in value for no apparent reason. It happened just a couple of weeks ago when the FTSE100 index, unprompted and with no obvious justification, corrected by several hundred points in the space of less than a week.

Rewind 20 years, and you may recall that such a move would have been regarded as highly suspect. Indeed, it is likely to have been met with considerable panic amongst investors and would probably have even featured on the 10 o’clock news. Now it barely gets a mention in the local gazette. So, it got me thinking, why is this?

Well, let me tell you what it’s not.

It’s not because companies have suddenly become inherently riskier. Contrary to popular opinion, this is absolutely not the case. After all, I don’t see any evidence to support the argument that a company manufacturing cars in the 1990s is more volatile than one manufacturing cars today. One could even sensibly argue that with better regulation and government assisted controls, many industries have become less volatile. So, if the markets are more volatile but the component parts of those markets are not, then what is going on?

Increased volatility is a result of how the market is being traded

I believe that the increased volatility actually has nothing to do with the companies in the marketplace at all but rather it’s how the marketplace itself is being traded.

What I mean is that the way in which companies can be traded, which includes highly leveraged derivative trading, gives rise to the significant bouts of volatility which we not only witness but have now become accustomed to on a daily basis. Spread-betting, CFDs, options, futures and a whole host of other investment vehicles mean that the underlying asset can largely be ignored whilst traders speculate on the derived asset. And this is what causes the volatility.

Bizarrely enough, we now find ourselves in the odd paradoxical situation where the derivative actually leads the underlying, not the other way around.

But it’s not just derivatives.

The other big change that has been responsible for the increased volatility is ‘algorithm trading’. This is a software process that identifies market inefficiencies using highly sophisticated models of data and then tries to profit by either buying or selling to take advantage of those inefficiencies. In other words, pre-programmed computers are replacing humans when it comes to picking stocks, which means that more trades are executed more quickly.

In fact, quantitative hedge funds are now responsible for huge swathes of trading activity in the marketplace and their highly complex mathematical models which automatically execute trades are now proving more popular than ever before. In other words, rather than an individual trader or hedge fund manager spending time looking at and researching a company before buying, a computer-based algorithm will do exactly the same thing in a fraction of a second and cost.

Not only that. Algorithms have the benefit of looking at multiple execution channels including traditional trading venues like the stock exchange, and non-exchange venues that are not open to most investors. This means that the volume being pushed through is massive.

Algorithms will look at a pre-determined set of variables before deciding to either buy or sell. That may sound like a great way to trade because it improves efficiency but it can also cause a problem which is that because other algorithms are also doing the same thing, it means that when certain criteria are reached there can be excessive trades concentrated over a short period of time.

Imagine that a computer algorithm identifies a sell signal on Vodafone and there are scores of other algorithm programs which simultaneously identify the same signal. The net result is potentially thousands of large sell orders in Vodafone being triggered at the same time.

Some analysts argue that algorithm trading actually helps to reduce volatility but I am yet to be convinced by this argument. For a start, when other traders and investors start seeing the price of a share fall, the first thing that they will do is to follow suit and sell. Logic and reason go out the window when panic sets in and I believe that the algorithm creates this panic. But apart from my own theory there are also many well-documented examples of ‘flash crashes’ which have resulted from algorithm trading.

For example, in 2010 the US equity market dropped by almost 8% in just a few minutes, wiping out billions of shareholder value before recovering. This had a profound effect on the whole marketplace and it turns out after closer inspection that a mutual fund sold $4 billion of futures contracts on the back of – yep, you guessed it, an algorithm trading system.

As you can probably guess, I am not a fan of algorithm trading and that’s simply because I don’t think it works. I’m all in favour of fast execution but these programs are taking speed to a whole new (and very unhealthy, in my opinion) level. I mean we are talking not seconds but millionths of a second! Call me old-fashioned but I think that’s really unnecessary.

At London Stone we have a policy of trying to not let the phone ring more than three times before answering and I have to say sometimes we struggle even with that! I can’t imagine what my team might say if I said that next year’s target is to start executing trades to the millionth of a second! A mutiny would be on the cards for sure.

Of course, with certain types of trading and specifically arbitrage trading where you are looking to exploit small differentials, then a millionth of a second can actually make a difference. But this goes against everything that makes investing not only fun but profitable. Investing is about analysing, researching and searching for opportunities. It’s about human emotions of fear and greed and it’s about expecting the unexpected and seeing things before they happen. No algorithm can show you how to do that.

It’s an art as much as a science and I can’t help but think that the more that we automate these processes, we diminish the genuine skill-set behind investing.

The risks of algorithm trading are large – and growing

I also know that the risks with algorithm trades are massive. In order to make any type of financial return, the trades being placed are generally massive bets with the tiniest of margins which therefore means significant risks for relatively small gains. So, when it goes wrong it can go badly wrong.

In August 2007 several quantitative hedge funds lost between 20% and 30% in less than a week whilst the market index barely moved. Quite bizarrely the very thing that is usually responsible for creating volatility lost millions for its investors when there was no volatility at all. That stinks. Telling your client that he or she has lost 10% of their portfolio because the market has crashed is bad enough but how do you tell them that they have lost 30% in less than a week and the market hasn’t even gone down?!

You may also remember a chap by the name of Narvinder Singh Sarao, later known as the ‘Hound of Hounslow’, after his algorithm trading caused major flash crashes in the S&P 500 index. His story is an interesting one because not only did it involve algorithm trading but he coupled that with massive price manipulation through ‘spoof trades’.

This is where he entered large and entirely fictitious buy and sell orders in the market which had the impact of creating panic and moving the price in his favour. He then cancelled those orders before the price reached his levels. And through powerful algorithms he was able to do this automatically hundreds of times in a day and by doing so he moved the market.

Whilst this may all appear very controversial – and certainly if you watched the BBC news you might think he would be partly responsible for World War III if it ever broke out – I don’t actually believe that Narvinder did anything wrong. From what I can see there are no rules against spoof trading and I don’t think that what he was doing actually constitutes ‘market manipulation’. If he had done that as a private investor nothing would have been made of this story.

However, he seems to have been a victim of his own success by virtue of the fact that he was particularly good at what he was doing and made a lot of money whilst making a lot of professionals look very silly, not least the American authorities.

Obviously, the US didn’t see it that way and he was ordered to pay nearly $40 million and is currently awaiting sentencing which could be as much as 30 years or more. That sounds harsh to me.

For me this doesn’t make any sense. If algorithm trading is so dangerous that it allows a lone and largely inexperienced trader like Narvinder to bring panic to the global stock markets simply by running a few algorithms from the bedroom of his Mum and Dad’s 3-bedroomed semi in Hounslow, then surely there is something wrong with the system.

In fact, Narvinder should probably be applauded for highlighting to the investment world how fragile the markets have become because of algorithm trading. No system, no matter how amazing it is purported to be, should ever have the opportunity to wield the power that algorithm trading now has – and the sooner we realise that, the better.

This problem hasn’t gone away. Mark my words on that. Algorithm trading is a problem waiting to explode and next time it won’t be Mr Singh playing on his home PC from Hounslow to blame – it will be a much bigger organisation and the consequences will be significantly worse.