'Flash crash' trader loses US extradition battle
bbc.co.uk
bbc.co.uk
How could a small trader like that possibly move enough volume to even budge the market as a whole, never mind crash it?
They claim he was spoofing trades... That's more the domain of HFTs, where you can automate the process, not a lone trader, however sophisticated.
And finally, let's suppose this trader is guilty of everything they allege... He still couldn't have crashed the market, it would have been a chain reaction of algos dumping stock at the same time, which he can hardly be blamed for...
2) You don't need to be have automated high frequency trading infrastructure and algorithms to spoof (though that would help a lot), you could do it by hand.
3) Agreed it would be ridiculous to blame this guy for the entire thing, but he did break the law (and the rules of the exchange) so there is something to be said about prosecuting him.
4) Laws and rules against spoofing are always going to be stealthily broken, and they are frustratingly vague, referring to the intangible "intention" of the trader. Exchanges would be better off removing such rules and let algo-traders take the responsibility of writing spoofing-resistant strategies. Allowing spoofing could even help improve execution. If a trader wants to execute a big order, it would be easier to just place it on the market and they wouldn't tip their hand as much. Other traders would have to guess if it's real or spoofed.
5) Another part of the problem is the asynchronous nature of order books. That's where 99% of the need for collocation comes from. You're trying to control your order in reaction to incoming information in an asynchronous way. Letting trading algos run in a round-robin fashion inside the order book would eliminate a lot of those issues.
https://www.chrisstucchio.com/blog/2014/fervent_defense_of_f...
Source: I've worked for "HFT" firms the past 9ish years.
He had negotiated a credit line from Credit Suisse to the tune of 30million. He would then use 30 million to open futures trades with a much much greater notional value.
His real positions would be in the hundreds of thousands long/short the market.
His credit positions would be a little deeper in the order book, knowing that other people and algorithms would front run his order because they think it is an indication of a big buyer or trend happening.
He'd close his real position at the better price to the people trying to front run him, for a profit. And simply remove the super big order from the order book.
This is spoofing
and yes this is what everyone else does.
He is being prosecuted for it.
During the flash crash, liquidity dried up and spreads got super big, front running algorithms were still responding to the spoofing and matching orders at really bad prices with wide spreads. It doesn't matter who "caused" it.
http://blogs.wsj.com/moneybeat/2015/04/22/flash-crash-trader...
I recall reading the charges from the government though, and I believe there it mentioned the size of credit line from Credit Suisse
What you're describing is just called trading. Someone says they wanna buy a million shares of something, you guess demand is going up. Just because they use special hardware and react fast doesn't make them any different.
And the UK laws did apply here, can we and should we allow the extradition to happen? UK laws said that he was liable under US laws so he would be extradited.
2. the UK is the international home of finance fraud, the main selling point of the UK for illicit finance companies and the USA don't want yet another shitty wrist slap
However, he absolutely was spoofing trades. That is not a high-speed activity, and can easily be done by point-and-click traders. There have been any other traders that have been known to spoof as well. His problem, however, was that he was dumb about it.
The law is about intent. When your order was placed, it must have been placed in good faith, with the intention of getting filled at the time of entry. However, it is completely okay to decide 10 milliseconds later that you no longer want to get filled (for example, you realize the market is moving against you.)
The problem here is that intention is extremely difficult to prove. You could say "I have a machine learning model that said it was a good price to buy when I placed the order and a second later, it changed the price". So, even if it is generally accepted that a trader is spoofing, the burden of proof lies on the prosecutor, and its a fairly high burden.
Sarao made it very easy, because he wrote in emails, exactly what he was doing, that he wanted a special order type that would never get filled. Then, the manner in which he placed those orders was obviously meant to drive fills into other opposing orders, and once those got filled, he would cancel the initial "fake" orders.
He also did a number of other things, and most importantly, wrote down what he was doing to make it easy for prosecutors.
I highly recommend reading the indictment: https://www.justice.gov/sites/default/files/opa/press-releas...
It explains everything that he did and the evidence against him.
Overall, the flash crash blame is the media/officials finding someone to blame, but he did commit a crime, and that's why he's being prosecuted.
The reason others are not being prosecuted is because others aren't stupid enough to write down their crimes.
They have him in emails literally saying things like "If I keep entering the same clip sizes, people will become aware of what I am doing rendering my spoofing pointless" and "I need to know whether you can do what I need, because at the moment I’m getting hit on my spoofs all the time and it’s costing me a lot of money."
It's not a huge reach for the government to claim he was spoofing when he himself claims he was spoofing.
How could a small trader like that possibly move enough volume to even budge the market as a whole, never mind crash it?
I think people who aren't familiar with trading (especially futures trading) don't understand that Sarao was spoofing massive size. He was placing orders for 500+ contracts at multiple levels in the book... that's on the order of $100M notional value. He was more of a pro than people like to believe.
Oh yes. However...
> How could a small trader like that possibly move enough volume to even budge the market as a whole, never mind crash it?
He wasn't that small, but in any case the argument is basically "by spoofing, which scrambled some dumb HFT algos and made them crash the market".
> They claim he was spoofing trades... That's more the domain of HFTs,
Quite the reverse; HFTs and spoofers are basically the exact opposites. The archetypal spoofer would be a small trader, working alone, trying to make misleading orders to fool the HFT algos. From a spoofers point of view, they're trying to use cleverness to beat the speed advantage the HFTs have and make the markets fair. From an HFTs point of view, a spoofer is trying to use dishonesty to unfairly beat the speed advantage HFTs have. The legal system sides with the HFT view, but that's not entirely uncontroversial.
> let's suppose this trader is guilty of everything they allege... He still couldn't have crashed the market, it would have been a chain reaction of algos dumping stock
You are 100% correct! However, spoofing is illegal whereas dumb algos are not illegal. IF he triggered the mess by doing something illegal, then yeah, legally he can be blamed for the mess.
That being said...the timing doesn't really work. Sarao's trades don't quite line up with the flash crash. (Roughly speaking, his spoofing was causing the market to make significant moves, but he was doing it in bursts, and the flash crash happened right after he's stopped one burst and the market had already recovered. And then after the flash crash happened and then recovered, he kept right on doing it, causing no additional crashes. Current research suggests it probably wasn't him. Worst case he might have spooked some HFTs algos and made them react poorly to some other triggering event. ...of course, his spoofing was still illegal, even if it didn't cause a flash crash.)
The Chris Stucchio article opens with a cute take on the South Park underpants gnomes list ("2. ???") and suggests that nobody has outlined why this strategy would be profitable ... which is weird because this is exactly what Nanex did:
"What is the benefit of a feed delay? For one, latency arbitrage is enhanced, especially if the delay is predictable, which it would be for someone running a Quote Stuffing trading strategy. The type of firm that would benefit the most from latency arbitrage would be an internalizer - someone that matched retail stock trades based on the slower SIP, but could buy and sell the same stocks on the faster direct feed, a strategy that became visible during the Nasdaq Blackout. By the way, Citadel's bread and butter is internalizing (matching) retail stock trades"
This is extremely easy to understand. Reading on there are a number of xkcd-style graphs that basically say "testing is hard!" but still don't really provide a solid rebuttal. I stopped reading at that point, I'm afraid :(
I'm far from an expert on this, and my opinion really doesn't matter - but I'm far more persuaded by the Nanex explanation.
So, it's all about initial intent, which is extremely difficult to prove. And that's why most traders who are well known to spoof markets haven't been prosecuted (see 3Red's Igor Oystacher), because they are smart enough to hide their intentions.
Sarao, on the other hand, wrote emails where he stated exactly what he was doing. For example, in the indictment, they have emails where he wrote that wants an order type that will never get filled, and if it looks like he is close to getting filled, to immediately cancel. Without those emails, he would've been able to at least make up some defense. When he has stated in writing that he is placing orders that he does not want to get filled, well, that's a very prosecution.
Source: I got run over a couple times in the Treasury market.
HFTs do the first two, but not the third, and it's the third which is the definition of spoofing.
By the same token there are probably a bunch of other traders doing the same thing who are smart enough to keep mum.
This is another example of how wrong the current models based on traditional assumptions and academic theories are.
If a market could be nearly destroyed by one single trader, there is no market at all, but a fraud, smoke and mirrors.
It doesn't seem unusual at all to me that someone charged with fraud in US markets would get extradited to the US.
Since most everything we do today involves computers and the internet, wire fraud has become a catch-all charge that can be selectively enforced (see Aaron Swartz) to suit a prosecutor's needs.
Despite the fact that many large players spoof on a regular basis, only this guy has been actually criminally charged. When a government anointed HFT firm does it, they only get a small slap on the wrist. Ironically, this guy pointed out to regulators how everyone is spoofing, and only after they did nothing to stop the spoofing did he say "well if everyone is doing it, I will too".
The corrupt financial elite have found their scapegoat, and the message is clear: if you're a connected Wall Street firm, the laws do not apply to you, but if you're a regular guy day trading from home, the book will be thrown at you.
Key quote: "This manipulative practice [spoofing] is illegal under the U.S. securities law, yet it has been frequently discovered in both equity and futures markets."
Here's an example of the penalty a middling HFT firm sees: http://www.reuters.com/article/us-sec-enforcement-spoofing-i...
$1.9M in fines - no jail. By contrast Sarao is facing up 380 years in prison.
Others faced fines of only several hundred thousand dollars: http://www.wsj.com/articles/how-spoofing-traders-dupe-market...
Note that the most connected HFT firms (Virtu, Citadel - oh hey, Ben "Revolving Door" Bernanke works there) don't even get fines, despite engaging in spoofing, and even being banned from China for spoofing: https://www.ft.com/content/d4b61458-398d-11e5-8613-07d16aad2...
Also note that there have been many "flash crashes" triggered by spoofing HFT's. There's nothing special about the Sarao one, other than the fact that he was an individual operating out of his home.
Furthermore, the study is based on the premise that there is lots of spoofing going on. That's because EVERYONE knows that spoofing is rampant. Everyone has acknowledged it, including the SEC and the Fed, so to dismiss the study by simply saying "it used fake data" with no other context is not a forthright appraisal of the situation.
That's fine, but you can't then claim the study as evidence of spoofing, since it was injected into the simulations explicitly.
> which they then compare to real data to test its accuracy, which they claim is 90%
The validation with real data was just comparing things like trade volume, volatility, autocorrelation, etc. They didn't put a specific number on it (the 90% accuracy mentioned in the abstract is about detecting HFT in simulated data). And they added in the spoofing after the validation.
There definitely is spoofing going on (some of my job involves trying to detect it), but I'm deeply skeptical of anyone who claims they can detect it accurately or quantify the amount. 92% accuracy for detecting spoofers in simulated data is insanely high, and makes me doubt the applicability to the real world.
As a former trader, I think it is bullshit and he is a scapegoat, but I think he'll get due process.
Unfortunately, the search you suggested yielded no obvious results.
I would expect that a savvy financial type who can afford to pay a competent lawyer would be significantly less impacted by this phenomenon (if at all) than the typical person charged with a drug possession offense represented by a public defender who has too many clients to properly serve each one.
Really, it seems to me that the issue isn't necessarily with plea bargains in general so much as with plea bargains with massive differentials to the original charges.
> Really, it seems to me that the issue isn't necessarily with plea bargains in general so much as with plea bargains with massive differentials to the original charges.
I agree with you. I think the root cause is completely unreasonable penalties for certain charges (which is especially true of drug charges). With a smaller differential, the bargaining power of the prosecution is drastically reduced (as it should be).
As proof of misconduct, note that when evidence is proved systematically unreliable, prosecutors argue that the rulings should remain. (eg, hair testing, etc) There's no desire to get the right answer, just to get the most convictions.
It's a prosecutor willingly not performing their duties (charging people with appropriate crimes for their actions), because it makes their tracked stats look better (conviction rate).
Im of the camp that believes plea bargains should only be allowed to negotiate sentences on the charge, not involve prosecutors charging a different set of crimes.
note that the UK does not elect its judges
Austria, Brazil,the Czech Republic,France, Germany, Japan, the People's Republic of China, the Republic of China (Taiwan), Russia and Switzerland [1]
To me it seems clearer cut in a case which is more obviously criminal.
Now unbalanced extradition, which is the case for us in UK with regard to the US, seems odder to me. Our politicians should sort that out.