Every time I’ve opened up the BBC news app this week, I’ve seen this:

Flash Trader Arrested

Here’s a close-up:

Thanks BBC.com
Thanks BBC.com

The basic summary:

  1. Back in May 2010, US equity markets experienced a now-infamous flash crash (pictured above), during which nearly a trillion dollars of value was lost, and then just as quickly, recovered.
  2. Apparently, this is all the fault of a 36 year old who was trading out of his bedroom in Hounslow, London.
  3. And now he’s been charged with one count of spoofing.

Now I don’t want to sound like a skeptic – but I am one, so I’ll have to sound like one.

One guy? Almost a trillion dollars in losses? From his bedroom in Hounslow?


  1. The market is ridiculously easy to manipulate, and any guy in his Hounslow bedroom can do it. Or:
  2. Scapegoat.

If anyone is interested, they should go and read “Flash Boys” by Michael Lewis immediately. According to him, Wall Street and the High Frequency Trading community are obsessed with building their servers almost on top of the market exchange because nanoseconds of trade information are vital to their…trade.

But somehow, Navinder Singh Sarao managed to interfere using his Hounslow broadband connection? Okay – what I actually mean is, via an automated program that was slightly customised for his orders via his broker?

Seems unlikely.

But that aside, let’s talk about spoofing. Because it’s a fun thing.

How To Spoof The Market, Assuming Good Network Connection

When people talk about the price of something, they are mostly talking about the supply and demand for it. But often, expectations of the future supply and demand also play a role.

Two examples:

  1. When Standard Bank was building its new headquarters in Rosebank, Johannesburg, that might have been a good time to buy real estate in the general vicinity. Why? Because demand for real estate in the area was going to go up once the hundreds of bank employees arrived for work and decided that they’d prefer to stay nearby in order to spend less time in traffic.
  2. If you hear that there might be a highway built through your suburb, you might decide to sell at that point, before the news gets out. Why? Because once the highway is built, your real estate investment will lose value due to all the noise and unsightliness and general security risks nearby.

In both those scenarios, you’re “speculating”. That is: you’re making your buy/sell decision based on your expectation of future demand and supply. And that future expectation impacts the current market pricing.

Now pretend that you could influence that future expectation. Let’s take that second scenario:

  1. You start a rumour that a highway project has been approved.
  2. You get some experts to declare that prices in the area will fall.
  3. The price falls, because everyone is speculating the same way.
  4. You buy up property on the cheap.
  5. You then get a formal announcement made to say that there is no such highway project.
  6. Property prices recover.
  7. You sell.
  8. Profit, minted.

Of course, it’s quite risky to pull that off. People are sometimes hard to manipulate en mass for prolonged periods of time – and buying and selling property is so time consuming (to say nothing of the transaction costs).

Now imagine that you could do it to:

  1. Robots,
  2. Instantly*
    *at least, virtually instantly

This is “spoofing”. And that is what Navinder was involving himself in, allegedly.

Here’s the technical description from the Department of Justice:

Sarao allegedly employed a “dynamic layering” scheme to affect the price of E-Minis.  By allegedly placing multiple, simultaneous, large-volume sell orders at different price points—a technique known as “layering”—Sarao created the appearance of substantial supply in the market.  As part of the scheme, Sarao allegedly modified these orders frequently so that they remained close to the market price, and typically canceled the orders without executing them.  When prices fell as a result of this activity, Sarao allegedly sold futures contracts only to buy them back at a lower price.  Conversely, when the market moved back upward as the market activity ceased, Sarao allegedly bought contracts only to sell them at a higher price.

That is:

  1. Saroa told the market that he was about to sell large amounts of E-Mini S&P 500 index futures by flooding the market with orders.
  2. The algorithms of the market went “OMG – so much upcoming supply – sell sell sell!”
  3. As they sold their contracts, the price of the contracts would drop, and Sarao (through his algorithm) would buy them at the lower price.
  4. Sarao would then cancel all his sell orders.
  5. The algorithms of the market would go “OMG – NO upcoming supply of the contracts – buy them back, quick!”
  6. As they bought up the contracts, the price would go back up, and Sarao (through his algorithm) would then sell them at the higher price.
  7. Sarao would then tell the market (again) that he was about to sell large amounts of futures contracts by flooding the market with orders…
  8. And so the cycle continued.
  9. And because he was dealing with robots, not people, he could do this repeatedly. Which he did. Allegedly.

So let’s talk about the alleged events of May 6, 2010.

At about quarter past 11 that morning, Sarao turned on his automatic trading program, and placed 6 sell orders (worth about, oh, $4 million)*. Over the next 2 hours or so, he modified or replaced those orders about 19,000 times* (about twice every second), traded contracts worth about $3.5 billion*, made $879,018 in profits*, and then called it a day at 1:40pm*.

For my money, trading $3.5 billion worth of contracts for less than $1 million in profit seems like a horribly risky thing to hand over to an algorithm. But I guess it means you can close your eyes and pray until it’s over. Then wake up two hours later, richer.

The thing is, by 1:40pm when he cancelled his orders, those 3,600 sell orders were equal to almost the entire buy-side of the order book.

So the argument (I gather) is that once he pulled out, the market just collapsed.

Thanks Matt Levine and Bloomberg
Thanks Matt Levine and Bloomberg

Which is naughty, because it left all the nice High Frequency Traders with less money.

Although I’m not sure why we’re being so protective of High Frequency Traders all of a sudden.

The Sum Conclusion

  1. To be honest, I’m still not convinced that a final cancellation of 3,600 orders worth $4 million collectively could have caused almost $1 trillion of equity value to be temporarily wiped off the market cap. I mean, that’s a multiplier effect of 25,000,000%.
  2. Even if you take the $3.5 billion in notional exposure (which seems bizarre to do – as that wasn’t what was cancelled right before the crash), that’s still a multiplier impact of 28,600%.
  3. Also, how can one guy with almost no money* cause a $1 trillion flash crash?
    *after all – he only needed to post margin on the trades that actually executed. But if he was continually buying and selling – he wouldn’t have needed very much margin money at all.
  4. If he could do it, from his bedroom in Hounslow, then why isn’t everyone doing it?
  5. But even if he did do it (maybe he’s a mad genius with wizardly fast internet connection speeds?), then I see no real problem with High Frequency Traders losing some money…

For more, two Matt Levine articles:

Rolling Alpha posts opinions on finance, economics, and the corporate life in general. Follow me on Twitter @RollingAlpha, and on Facebook at www.facebook.com/rollingalpha.