The Gateway to Algorithmic and Automated Trading

Shut up!

Published in Automated Trader Magazine Issue 24 Q1 2012

Key feature of today's electronic markets is noise, says Shayla Walmsley. But does that matter? Should we care?

Cast your mind back to pre-history, 2008 to be exact. It was not exactly a time of unparalleled naïveté, unless perhaps you had just talked your way into a first job at a remote and out of touch Savings & Loan, structuring these new-fangled derivative products around 110% mortgage instruments coming out of the Midwest, but it was nevertheless still a time when it was possible to be surprised by the wrong kind of surprises. We all knew by then - didn't we just? - that banks could fail. We all knew that liquidity risk was a sure-fire winner as a conference topic.

We had views on structured products. Many of us had updated our plans for getting out of the city and buying a big small ranch on an enormous mortgage signed off against our self-certified future bonus potential with our own money. We weren't enjoying the experience, but we were pretty sure we'd stumbled across the mother of all harsh lessons, the harsh lesson to end all harsh lessons until the cycle brought us round to another kinda predictable-ish crash in, oh, a generation.

Yeah, but then we got the news that the parent company of United Airlines had filed for bankruptcy. Or rather, we didn't. Our machines got the news and reacted memorably - down, up, money made, money lost. Then we got the news and reacted appropriately - hey, this sounds familiar. And it did: it was a six-year-old newspaper story recycling through the internet in one of those little e-glitches that make the online life so bracing. Wrongly taken by some to be a new filing, the story caused the firm's share price to crash within minutes. Rightly taken by others to be a re-release of old news, it brought some useful volatility to the share prices not only of UA, but also of every other constituent of the airline sector.

With knock-on effects across the whole travel industry - but not in any particularly meaningful way across the information industry, oddly enough. The whole UA thing has been eclipsed by the subsequent 'flash crash', but it's a convenient marker for the beginning of the end to come. Consider: UA, flash, just about every other unexpected recent bump; this is sentiment writ large. Markets, prices have always moved on rumour, but a move big enough to prompt systemic anxieties, followed by another one, followed by - you know, I can't wait to read the histories of these times. The six-year lag between that original UA filing and the stock's big trading day may represent a fine example of high-latency trading, but the enduring discussion point is, surely, the departure from reality.

We have come to tolerate a 100 per cent disconnect between cause and effect. Nothing happened; the story was rubbish. But the share price still tanked. The flash crash - we don't really know; we could just say that it caused itself. This is a cultural as well as a market phenomenon. You will forgive me, I hope, if I somehow can't raise the energy even to Google the details of this example, but at around the same time as nothing actually happened at UA, or maybe it was a bit earlier, the British Prime Minister of the time, Tony Blair, went on television - 'serious' news TV - to express his concern for the well-being of a soap-opera character.

There's a potentially interesting conversation to be had about the 'Skynet factor', which is the possibility that algorithms run away with themselves. Those of us who have never quite worked out how to erase our youthful Facebook pages will know how that one feels. But you don't have to have trading algorithms becoming self-aware (as in the recent popular novel, The Fear Index , by Robert Harris) and turning against their creators - by bankrupting them rather than, say, launching missiles at them. You just have to ask yourself: how long has it been since I was last able to use the term 'efficient market' without sniggering?

It's noise. We've designed machines capable of listening to, and reacting to market noise, without teaching them how to distinguish between noise and, as it were, music. Wasn't there a paper a while back - heck, there have been hundreds of papers - no, this one was from the Fed - heck, there have been - anyway, paper(s) on the persistent effects of false new shocks? Efficient markets are the ones where prices react to fundamental signals, not noise - and not even noisy news. There's probably a paper to be written on the prospect that 'frontier', relatively low-tech, new and exciting markets might turn out to be less 'noisy' than the big ones.

But that's for another day. Take another, more recent, example: US security products firm Diebold's crash and almost equally speedy recovery last June after an overreaction to the firm's settlement with the SEC over accounting practices. Then-Senator Ted Kaufmann, a long-time advocate of "urgent regulatory action" over algos, blamed them for aberrant reactive triggers to breaking news, and then, ah, 'blamed' the stock's speedy recovery on the intervention of "human judgement". Well, maybe, Senator, we'd like to think so, and we love your implicit affirmation that algorithms are susceptible to human intervention, correction, control.

Steffen Gemünden

Steffen Gemünden

But if we're going to be talking about real-life electronic trading, out here in the wide open e-spaces where the algos roam free, we should be thinking around the sometimes rather slight differences, in terms of the noise they potentially generate, between a real fundamental, a fake fundamental, a tasteful musical evening with the Bath Baroque ensemble, a new track to download from the heavy metal band Metallica, a Mozart piano concerto, and what happens to domestic bliss when the family teenager discovers percussion. Their impact is potentially similar, and potentially deceptive. If it quacks like a duck it's probably one of those decoys you can buy at a gun shop, and if it walks like a duck, it's probably the duck-hunter who bought it. Wearing waders that don't fit.

There's a view that noise is a fundamental, at least in any meaningful definition of that f-word. There's also a view that all data is noise until you find the code that turns it into music. But more immediately to the point, there's a view that Kaufmann was wrong to advocate the specific measures he did - mandatory speed limits and resting periods as a panacea for fleeting liquidity. Rhodri Preece, director of capital markets policy at the investment professional-accrediting CFA Institute, says: "The minimal resting time wouldn't help because the person submitting the order would be put at risk by algos and counterparties using the information in the order against them and to their own advantage." Somebody else would jump in and take the biscuit.

Market activity is itself noise. In the typical price-impacting non-event, orders submitted were rapidly cancelled. That made a noise, as did the process of identifying that it was noise. An expensive one. Preece comments: "Analysing volumes of data to understand what happened is an expensive business, especially when you're talking about the volumes of data you needed to analyse to understand what happened in an incident like the flash crash." Yeah, right, true enough, but the other side of this is that we're all getting used to competing in a noisy environment. "Before, when someone put a bid in, it meant they wanted to buy it," says Eric Scott Hunsader, founder of data analysis firm Nanex. "These days, bids and offers are just meant to trick other market participants into acting."

So where does all this get us? The theme here is noise, but the key point, really, is simply that times have changed, and markets with them. It's a truism to say that electronic trading is much more than just recreating a human activity in a more efficient, streamlined, rapid form. Preece suggests that the focus needs to be directed towards two other risk management mechanisms: first, pre-trade screening to ensure that what hits the market isn't rogue or erroneous; and, secondly, a system of circuit breakers - but not just any such system. To be effective in a fragmented equity market, circuit breaker technology would need to run across platforms: in other words, you would need a harmonised system.

Harmony is the big ask, and we've said enough in recent months about a certain well-known EU Directive beginning with M- and ending with -iFID [Not quite enough. See page 26. Ed.] . A dimension of this debate is the contention that effective risk management must be multilayered, involving the clearer, the exchange, traders and regulators. Steffen Gemünden, CEO of RTS Realtime Systems Group, says: "The talk is about algos going haywire, but that can be prevented by pre-execution limits. But there is also a need to look at limits such as profit and loss." Gemünden goes on to argue that traders "have a natural interest in getting risk under control. They're not necessarily the bad boys. These are our customers, and they take this seriously."

It's almost enough to make you wonder whether somebody, somewhere, should be building a regulatory algorithm, to roam the electronic mean streets in search of e-wrongdoers.

Unless that's happening already, in some secret laboratory somewhere?