The Flash Crash on India’s exchange highlights the perils of instituting catchy mechanisms like circuit breakers without really analyzing what causes problems in the first place.
"You just need to look at the news to see why voice trading is still needed," said a sales person, formerly of various large-name banks, to me over dinner in New York a few weeks ago.
"Er," I mumbled through a mouthful of linguine, pausing to swallow. "What do you mean?"
"How many fat finger errors do you see where some idiot forgets a decimal place, or puts an extra zero on the end, and crashes something as a result?" They continued. "You can get as electronic and as high-tech as you like, and fine, it creates efficiency," The glass of wine in their hand waved dangerously with the emphasis of their gesticulation. I nodded as the liquid swirled to the event horizon of the glass lip, before dipping back and avoiding spilling out over me. I exhaled. "Sometimes you need someone on the other end to go, ‘Hey, do you really want to do that?'"
"Doesn't pre-trade risk kind of do that already?" I murmured. The glass sloshed upwards.
"That's just it, you'd think so," they said, becoming hazardously animated again. "But you'll have circuit breakers and speed checks and the like, but a lot of the time, the pre-trade controls just aren't there to avoid something that's so simple as a manual input," the glass went down. "Especially when you're dealing with really, really big orders. It's easy to make mistakes when you have someone hammering you constantly."
The problems on the Nifty will probably incite regulators to clamp down more stringently on algorithmic trading as a result. But it's the enforcement of pre-trade controls and regulation that needs a more forensic eye here, rather than the algo itself
The conversation, seemingly sprung out of nowhere, turned out to be relatively prophetic. Not too long afterwards, trading halted on the National Stock Exchange of India (NSE) due to an erroneous barrage of sell orders. Details are still being picked over, but vaguely, it's being blamed on a human error from Emkay Global. Apparently, a hapless trader mistook the price of the shares to be sold for the volume, costing the broker $10 million along with a suspension. The market's circuit breakers kicked in when the Nifty index fell 10 percent, although it actually fell to 16 percent after it stopped accepting new orders due to existing transactions being completed, and temporarily wiped out $60 billion from the value.
It was, for all intents and purposes, India's 6 May 2010 on a smaller scale (the Dow Jones momentarily lost $850 billion on that day).
The questions that should be asked shouldn't center on why it happened, though. That's relatively understandable, and although Emkay is paying the price, it was a mistake by most accounts. The real one to ponder is why it was allowed to happen in the first place.
Shouldn't pre-trade risk controls, not only on the part of the broker, but also on the exchange's behalf, pick up what was clearly such an erroneous order and block it?
You'd be forgiven for thinking so. It just goes to show that while circuit breakers, mandated in US exchanges after the Flash Crash and in place for a while in Europe, are effective, they shouldn't be the first and last line of defense against aberrant price movement. When I put in an order on Amazon, I have a number of screens asking me if I'm sure that I want to place it, and my details are correct. I get that in the high-speed world of modern equities trading that's not always possible, but there should be some sort of confirmation process and some sort of system to flag up these things before they hit the exchange and ruin everyone's day.
The problems on the Nifty will probably incite regulators to clamp down more stringently on algorithmic trading as a result. The Securities and Exchange Board of India has already proven itself to be wary, at best, of the practice. But it's the enforcement of pre-trade controls and regulation that needs a more forensic eye here, rather than the algo itself. Arguably, it did its job rather well, stock market crashes notwithstanding.
Waters Wavelength Podcast Episode 75: An Update on the Julia Programming Language; AI & Alternative Data; Digital Currencies
Julia Computing's Viral Shah talks about the programming language he helped create and what's ahead for it. Then James and Anthony talk about the pairing of AI & alternative data, digital currencies, and Game of Thrones.Subscribe to Weekly Wrap emails