London man arrested over $40 MILLION HFT flash crash allegations As I've mentioned around here I'm a bit of an aficionado of scams and scammettes: not because I partake in them but because the inventiveness of the human mind in hoovering up cash never ceases to amuse me. This morning we've a classic of the genre, as one bloke working out of his mother's basement* in Hounslow is alleged to …
Mmmmm. Man manges to make twenty something million pounds (according to the Americans) and yet still continues to live in Hounslow? Really? Presumably, the Americans think Hounslow is some sort of sleepy, lovely village in England where everyone with a few million would choose to live.
Sudden changes in personal circumstances? Errr, go read the companies house filings for his futures business - they've already been posted on the internet. I believe someone has already looked into it at noted that on the years where losses were reported they arose due to large unexplained withdrawals from the company. There's also apparently a company setup in a shadier tax realm with "Market Milking" in its name. I'm not so sure that living in Hounslow flies under the radar given the filings already made to the authorities.
HFT isn't the issue.
Spoofing the market is. Then all those 'home grown' algos take the head fake and in the edge case, the other HFTs force the market to keep moving. (gaining its own momentum.)
So it heads down and keeps going in that direction.
Looks like they were able to find trader 0.
Of course spoofing is dumb. They will ultimately get caught. (And yes, there are ways to sort of cover your tracks. )
The other thing is that its possible to correct the market to charge for each bid. Then you get a credit (multiplied) to balance it out. If you do enough spoofing to move the market, you'll end up always paying the exchange. In theory it wipes it out the problem,
"HFT isn't the issue"
Really? Being fooled by proposed sales that don't ever take pace, and you say that is not a fundamental failure?
What he may (or may not) have done may be dubious, but the real issue is just how much those automated traders were taken in by momentary data of sales that did not complete. You would have thought after one or two incidents they would have learned, but no, it seems to have been profit for years if the allegations are true...
I think you need to learn more about trading.
What he's accused of doing is spoofing. That is he placed orders/bids with no intention of having them filled. It was a head fake.
HFT is not illegal nor the issue.
Putting orders out and not having them filled because the market is moving away happens all the time.
This is why its hard but not impossible to show that the trader was 'spoofing'.
"HFT isn't the issue"
Really? Being fooled by proposed sales that don't ever take pace, and you say that is not a fundamental failure?
I see you have (currently) eight upvoters, so there are at least nine people here that don't understand that neither spoofing - "proposed sales that don't ever take place" nor "being fooled" are part of HFT.
This scheme did not depend on HFT. As Worstall pointed out, it's been done many times without benefit of HFT.
In cases like this, the "being fooled" part was based on algorithmic trading. That's necessarily a part of HFT, but it does not require HFT. Regular meatsacks can perform algorithmic trading, and many small investors do. Everyone who has a "system" that involves crunching numbers to make trading decisions is doing algorithmic trading.
And investors can be fooled even if they're not using algorithms. Indeed, many a fool has lost investments thanks to making trades based on intuition or whatever nonsense might have guided his or her decisions.
HFT may be a problem. It is not the problem here.
Incidentally, for those interested in more technical details about HFT, there's a nice intro to HFT algorithms here, and an overview of HFT systems here, both by Jacob Loveless. I recommend the latter in particular to anyone who wants to understand what HFT systems look like - regardless of whether you think they're evil. They're in ACM Queue which I believe does not require ACM membership.
That is not correct.
There are a number of things you can do to with HFT, including looking at the raw binary data as it passes thru your custom logic , whilst others have to pass it up thru layers of network cards, software and operating systems.
Thereby giving you the edge on a trade THEY are about to execute, so that you can out trade them.
Then there is the ability to deliberately spoof the market , all under the 'guise' of test orders and all perfectly legal, specifically because the markets have specific binary data to allow such things.
Just to the poor scum plebs it looks like real orders because you don't pay for the tagging information.
go check this out:
High Frequency Trading Acceleration using FPGAs :'Christian Leber, Benjamin Geib, Heiner Litz'
or
http://www.hoti.org/hoti20/slides/Lockwood_AlgoLogic.pdf
plus certainly read "flash boys" By Michael Lewis
Another reference point wrt HFT and FPGAs is the sad (?) history of Abingdon company Celoxica.
From memory they were an Oxford University spinoff who devised a setup that allowed FPGAs to be programmed in a variant of C (Handel C?) rather than the usual extra-techy stuff associated with FPGAs.
They went through various finance rounds and sub-incarnations trying to sell the stuff as generic accelerators for specific sets of problems. They kept running out of money.
Then someone sussed that the kit could be used for HFT, and they're now primarily driven by HFT.
In passing: the job ads on here featured an ad for an HFT programmer a few weeks back.
Starting salary: £100K.
One man can, even momentarily, wipe out a large percentage of the value of the richest country in the world, and the author wants us to belive that there is no inherrent problem with the system?
I don't buy the idea that there's nothing wrong with HFT, and wish to see markets stabilised by abolishing it. I'd propose two simultaneous changes:
1. A minimum period of share ownership for listed companies of, say, 24 hours.
2. A tax (very small) on all trades. Fractions of a percent - perhaps 0.1%? Enough to make HFT unprofitable, but not to have any real impact on funds and investors actually, you know, funding and investing.
In fact you could drop point 1 altogether by starting the tax at 10% for shares sold sooner than one minute after purchase, and decreasing to zero after, say, a week/month/year.
This post has been deleted by its author
1) Wouldn't make any difference to this. He wasn't trading in shares, but in the index contracts.
2) The argument against such an FTT is very simple. It would shrink the economy overall. My one and only published peer reviewed paper is on this very subject. And the proof of the contention that it would shrink the economy is in the EU's own report into the matter. Where it says that.
Rather amusingly that paper of mine was used in evidence to the House of Lords subcommittee that looked into the Robin Hood Tax. That's not the amusing part: someone else submitted as evidence to the same inquiry a blog post of mine where I'd been saying much the same. Not sure how often blogs posts are used as evidence in the HoL......
Dicking around with figures on a computer isn't really 'the economy' though, is it? It's not real money. I'm skeptical that it offers any real value to anyone except sellers of luxury watches and cars and houses.
We don't have any empirical evidence of the effect or otherwise of a FTT so surely the sensible thing to do is to test it. Run it for a fixed period and see whether or not the economy is better or worse. If it's clearly made things worse for everyone (eg corporate tax receipts have fallen so there's less money to pay off the deficit, and it's put people out of work) then scrap it, because we'll have proved it doesn't make anything any better and makes an awful lot of stuff worse.
Your analysis, like that of most economists, assumes that benefits of the change will be shared (benefits here not meaning 'welfare').
It doesn't work like that.
Under the current regime in the UK and the US, the bigger part of the benefit goes to those who need it least. The rest of us get the joys of a failing economy, zero hours contracts, and similar delights.
Rather than a minimum period why not just make all markets turn based, so that trades are played out on the minute every minute. That seems like it would make no difference to the effect of the shares in terms of providing liquidity to companies, but reduce the maximum frequency of trades and consequently stabilise the extremes.
But everyone has access to the same developments, and the highest frequency is once per minute, so even where algorithms are ruling most decision making they don't have to be impossibly fast, so can take the wider situation into account and more sophisticated ( and consequently less liable to cascade failure ) algorithms can be developed.
But you would want to get your trade in at the last possible moment before the trading window, and have processed as much as possible of the new information available at that time.
For example if Mark Carney announced 10ms before the trading window that interest rates had been increased to 7%, some people may be trading without knowledge of that development, others would know about it and would be offloading their shares and bonds to people who didn't know.
One man can, even momentarily, wipe out a large percentage of the value of the richest country in the world
Ugh. The lack of basic critical thinking among the commentariat whenever this topic comes up is really a bit nauseating.
What does it mean to "momentarily ... wipe out"? Not much of a "wiping", is it? Congratulations; I believe you've coined an oxymoron.
"A large percentage of the value of the ... country"? Where the hell did you get that? The 2010 Flash Crash peaked at about 9% of the DJIA. Now, we'll pretend for the sake of argument that the DJIA actually means something, rather than being the result of a completely arbitrary formula. I wouldn't even call that "a large percentage of that particular market", much less "of the country".
The usual estimate is that at its nadir the drop represented about $1T in nominal market valuation. In 2010, even the money supply dwarfed that - US M3 for May 2010 was around $14T. And that's just the (broad) money in the country. There are a few other assets here, like all the stuff produced and the services rendered and the property and the natural resources. Those add up to a few more dollars. The 2010 US GDP was also around $14T, for example; so even counting just the money supply and that year's productivity, the Flash Crash at its worst moment - and it was only a moment - represented less than 3% of the nation's "value".
And, as I and others have already noted, the FC was not due to HFT per se. The same thing can easily happen with high-volume algorithmic trading at more human frequencies, and variations of it without even involving algorithmic trading.
From the BBC: "The statement continued: "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."
This I cannot quite understand. To place a sell order which you don't want filled you must put it above the market (because if you place it below the market, you will get filled immediately and won't be able to withdraw). Then you cancel it if the market prints higher and closer to your number. That can only drive the prices up IMHO.
"This I cannot quite understand. "
Maybe because the Beeb don't understand it and have got their facts wrong. Or because the Feds' case is a load of old horse sh**. Or both.
In the wider scheme of things, you have to ask yourself whether you believe one day trader in a scummy part of London was able to bring down the financial might of the US of A, and to be smarter than than the crooks of Wall Street in turning nothing into money and manipulating markets?
My guess is that the flash crash was 100% prime American product, but the US regulators don't want to admit that their ineptitude and the greed of Wall Street are at fault. Far better to try and tar London (and if possible UK financial regulation). If some (relatively) ordinary joe has his life made hell, so what? Gary Mckinnon can talk to a similar story about the petty, enduring spite of Merkin bureaucrats, as can a whole assortment of US-based whistleblowers.
"Layering" refers to putting up multiple sell orders above (or buy orders below) market as far as I understand. Just one order above market might not look to strange to HFT algos (or perhaps even other traders) but multiple ones from ostensibly different market participants can appear to indicate that the market will move in a different direction. e.g. (in a very simple market)
Market before layering:
10000 sell orders at 100 and 20000 buy orders at 101
Usually you would expect new trades to settle between the two prices.
Market after layering:
10000 sell orders at 100 and 20000 buy orders at 101
20000 sell orders at 101.1
15000 sell orders at 101.2
5000 sell orders at 101.3
The algos \ traders originally trying to buy at 101 could then adjust their prices upwards (maybe to 101.1) to take into account this new information from the additional sell orders. I think the part about the "fake" orders being large in relation to the "actual" orders on the market is important here because that might potentially fool algos. Not sure this strategy will work against an actual person on the other side of the trade.
That's just speculation though - the actual strategy was probably a lot more complicated.
There's no real problem with layering - it is expected behaviour for certain participants especially in markets where orders are filled on a first in first out basis. It is common for market makers to do it. There are also informational reasons for using them by placing a 1 lot bid/offer at the back/middle of the order book - in a market move you'll get notified of your order being filled quicker than the public dissemination (via the exchange market feed) of that trade, and hence price, occurring. This gives you a speed advantage over other players provided you are happy to take whatever hit may occur from the 1 lot being filled.
With spoofing where you place your "never intended to be fulfilled orders" in order to move the market you'd need to be damn sure that you're one of the fastest kids on the block because if you're not the market could move and your orders get hit before you have time to pull them. I've seen HFT players caught by bigger faster fish doing this and in less competitive markets than the US.
Not on the same scale, but the big online retailers use pricing algorithms, and they can go crazy too.
Hmm. So much unrevealed as yet. First one that jumps out at me, having had some experience in trading software platforms, is credit limits. All clients of the major exchange participants have credit limits. Orders, even those immediately pulled, are first credit checked. Placing orders that you don't expect to be filled are still subject to credit limits before being added to the bank's positions on the markets - so that's either broken or someone was extending too much credit here.
His offence seems to be being a small guy doing what the larger guys whose HFT algorithms make the rules are doing and bragging about it. His choice of name for his offshore company seems to give the game away a bit too much for their liking: "Nav Sarao Milking Markets Limited."
His offence seems to be being a small guy doing what the larger guys whose HFT algorithms make the rules are doing and bragging about it
I think it's simpler than that. His offence is basically not being in Wall Street.
I have over time noticed a trend with US regulators (and government) in that they are incredibly quick and harsh to rule on anything they can sell as done by people abroad (them foreigners), especially if they are Wall Street competitors, but take out the wet noodle for anything that has been dreamt up in the US.
If you are genuinely investing in a company, then you'll spend a week researching, and hold shares for a couple of years or so. The "flash crash" lasted a few hours, and was interspersed by lots of people standing around saying: "what the heck, something has obviously broken today".
If you've programmed your computers to do dumb things automatically, well heck - maybe humans are better suited to this problem. Especially if your employment mostly consists of finding someone else to blame when something goes wrong.
In any case, no one doing anything of worth in the financial sector would be damaged by this temporary glitch.
If you are genuinely investing in a company, then you'll spend a week researching, and hold shares for a couple of years or so.
Often, yes, but certainly not always. Take banks for example - Once Barclays and Lloyds dropped to being penny shares, they became a nobrainer. Little research required, and a holding period of just a few months would have seen you well into the black. [1]
A recent example might be Tesco shares. 50% off? Dropped to less than £2? Sometimes the market just oversells due to fear. Sell those same shares today and you're looking at gains of about 15% in just a few months.
Holding periods of milliseconds however, would be highly suspicious when ran out of a house. You're trying to out smart the rocket scientists and out compute people who think literally nothing of spending more on computer power than your whole street costs.
[1] Lloyds still is a penny stock but its worth more than double its oversold value.
Mr Sarao is being regularly picked out as the 'orchestrator' or similar of the 2010 Flash Crash. However, the SEC investigation report in to the events on May 6 indicate that the primary trigger for the crash, in a context of general volatility and poor market sentiment, was the single selling handled by automated algorithm of a large number of the S&P E-Mini contracts. That's real selling, not spoofed, and not even by HFTs.
So the SEC report does not attribute the trigger or primary cause of the crash to an HFT let alone someone spoofing. The SEC report does imply that the presence of HFTs contributed to the acceleration of the crash, and the complaint filed against Mr Sarao does allege that his algorithms were active immediately prior to the crash (hence contributing to it).
Mr Sarao seems to have a case to answer regarding spoofing (which to the untrained eye looks like a hair's breadth and a good lawyer away from what the big HFT firms do everyday). But it looks more like he was a bit of a wide boy making a dodgy (if possibly impressive; possibly Hounslow) living off the markets who happened to get caught inside the building when it collapsed.
The Mr Big behind the Flash Crash? Seems a bit too convenient to ignore the inherent instability in the whole system and easier to blame on one ne'er do well - good for the extradition though.
Flash crash occurs and although the SEC doesn't directly name a cause, it does point a finger in the general direction of Standard & Poor. That's not good enough for the decent, hardworking, God-fearing folks in America. Apart from anything else, how could a respected US institution possibly do anything wrong?
So, we need a scapegoat. Found one? OK then - go through the list:
1. Foreign? Check.
2. Based outside the US? Good.
3. Located near a financial rival? Plusgood.
4. Asian name? Doubleplusgood. Obviously a paedo-trrrist IS mastermind and financier.
OK, that'll do boys. Get that mofo over here and waterboard a confession outta him. Justice Will Be Done.
Spoofing is hard to prove.
Putting bids out and then canceling them if the market moves away from you is legal. Putting bids out as a head fake and misdirection is illegal.
According to the articles on this individual, his program entered bids that represented 1/2 the volume of contracts being traded. This is also why its easy to charge him. He doesn't have enough cash to cover his trades. He's a small fish making outlandish bids. And once you find a guy trying to move the market, you can then go back and review any and all bids he may have made. Now you can establish the pattern of spoofing.
He's going to lose his extradition appeal.
Whether or not he's trader 0 is irrelevant. If he did put out and cancel the bids, that large a number of contracts... it could easily have been the tipping point. I seriously doubt these algos include a PID function.
Robert Peston of the BBC did an entire series on this subject on Radio 4. It was both highly revealing and quite frightening. I think it was called "A Dark Magic" but it doesn't appear to be on iPlayer at the moment. He basically suggests that the way the markets operate at the moment are a complete disaster for pretty much everyone. We lose 10s of % points of our pensions in transaction fees each year. The traders pretty much always lose money. I'm going off memory from 2 years ago but something like 5% of hedge fund managers actually make significant profits. However, those 5% make such significant profits that the remaining 95% of break-evens and losses are insignificant. The really scary part was that there is almost no overlap between the 5% who made money last year and the 5% who made money this year. The only people making reliable money are the 13 layers of traders/brokers/etc that perform each transaction and who each take a fraction of a fraction of a penny.
He points out that financial markets, rather than being huge revenue makers as suggested by both Labour and Conservative governments for 20-30 years actually just take money out. Every penny they "make" is actually money they have skimmed off our pensions, savings and whatever our industry generates.
"Every penny they "make" is actually money they have skimmed off our pensions, savings and whatever our industry generates."
Yes, and every penny farmers make is skimmed off the bread and milk on your table, and every penny a teacher makes is money skimmed off the education of your children. Where did you think it was coming from? I don't invest in hedge funds or "star fund managers" so I couldn't care less how many lose money. I invest mainly in cheap tracker funds. And the money is to be left there for years or decades so whether some coked-up barrow boys cause the index to dive for a few minutes before immediately going back up again is also of no interest to me.
In the long run this will do me better than stuffing my money under the mattress. Investing would not be possible without financial markets; without financial markets I would have the choice of either stuffing it under a mattress, perhaps putting it all in a mate's business (which could go tits-up and ruin me), or going down the dogs.
The assertion that we lose 10%+ in transaction costs every year is nonsense. Whatever funds you are invested in must by law declare the total of all their costs, including all trading costs, in their disclosure documents under "Total Expense Ratio" or "Ongoing Charges". If it's more than 2%pa you have chosen a very expensive fund.
"
Yes, and every penny farmers make is skimmed off the bread and milk on your table, and every penny a teacher makes is money skimmed off the education of your children.
"
The huge difference being that the farmer and teacher have actually used their labour to produce the commodity you are buying, with the amount they are paid being roughly proportional to the amount of work they have put in. i.e. they have *earned* their money. A percentage charge levied on a transaction is in no way proportional to the amount of work put in by the person who "earns" that percentage.
Oh what a comfortable world you live in. But that it were truly so :)
Damn near everything gets passed through the markets, whether you like it or not. You've chosen to invest some of your money in a way you are comfortable with and that's great. But if you have a personal pension, your pension company IS wasting huge amounts of it just moving money around. Because, every time they move YOUR money YOU lose some of it in those 13 layers of transaction charges. Company pension? Same thing. Pay National Insurance? Same thing?
ALL of our money is in the system and being moved around and therefore the value of what the world produces is being constantly diminished by an industry that exists entirely to tempt others into moving money around in order to generate "profits". But there are no profits because they don't produce anything. If one fund makes a profit, someone else makes a loss. And the sum of the two is always a bit less than the whole thanks to those 13 layers of fees and charges.
I know what the transaction costs are and I know what they're paying for. They're paying for the fact that I don't have to buy and keep certificates for hundreds of different shares to ensure my money is sensibly diversified. You should try actually reading the details of what it costs to invest your money in the markets, it's extremely tedious but much more relaxing than the miasma of fear you've surrounded yourself with.
"Pay National Insurance? Same thing?" - Er, no, I know exactly what happens to my National Insurance, it goes straight into my parents' bank account via their State Pensions, along with all the other pensioners, the unemployed, benefit claimants, etc. It never goes anywhere near the markets.
@Cynic: Not remotely true. A teacher in a sink school teaching 30 disruptive children puts in more work than a teacher teaching 10 attentive kids in a fee-paying school, they still get paid less. A serf working a strip of land with an ox puts in the same work as a farmer with a hundred acres and a fleet of machinery, they still get paid less. Remuneration is not about a central planning committee looking at how much work you've done and handing out wages accordingly.
Liability, insurance, capital adequacy and regulatory costs all go up in proportion the more money you look after. As, to a certain extent, do research and staff costs. (If your fund doubles in size from £10m to £20m you can not simply chuck that £10m into the companies you already hold without hitting issues with diversification, the cost of acquiring that many shares and owning too large a chunk of one company - even if you wanted to, there are regulatory limits.) Costs have to be paid for. If you don't think they're worth paying, well, it's a free market and there's always the mattress.
Incidentally *transaction* costs are often in pounds and pence depending on your trading platform, though management costs are usually a percentage.
We lose 10s of % points of our pensions in transaction fees each year.
I lose about 0.49% in fees due to low cost trackers.
The traders pretty much always lose money. I'm going off memory from 2 years ago but something like 5% of hedge fund managers actually make significant profits. However, those 5% make such significant profits that the remaining 95% of break-evens and losses are insignificant. The really scary part was that there is almost no overlap between the 5% who made money last year and the 5% who made money this year.
Robert Peston has all the economic credibility of my aunt Sally, however on this issue he is very much correct. I think the numbers you give are actually generous, its more like a 99-1% split.
I work in the City and have done for decades. My pension invests ONLY in low cost trackers, which ought to speak for itself.
There are some fund managers / investors that have outperformed the market in the long term, but I'll not list them here as I have some ISA investments in their funds and don't wish to "talk my book" as it were.
Yankee comedian Chris Rock made the point well.
Alcohol is made by white people: no problem.
Pot and cocaine made by brown people: illegal.
White people sell guns: no problem.
Black rapper says "guns": congressional investigation.
As for your pommie trader who made a packet by, so far as I can see, bluffing some of Wall Street's trading algorithms, more power to him I say.
Clearly you haven't a clue.
If the allegations of his bids are true, he's in serious trouble.
What don't you believe, that the markets can record every bid that's being made?
That they can't record the open volume of contracts at any given moment?
That the guy is incapable of being foolish enough to try this stunt?
Or that the bulk of the HFT algos don't consider edge cases in their strategies?
(Hint: Trying to buy/sell a volume equal to the volume traded in a normal day is a stupid move.)
If your algo is running on a financial service provider's co-lo kit the latency is way less. If he was doing what is alleged he would have needed to be very quick in order to get out of the way of the big boy's hitters which operate in the <=1ms range. This points to his algo running on co-lo kit provided by someone else rather than on a gaming rig at his mum's place.
So, it's already for HFT systems to manipulate the market, make those fractions of a cent that they haven't REALLY earned, but if someone else manages to do it they get in trouble?
To elaborate: 1) The HFT systems usually use exploits in the trading platforms to look at trades everyone else has put in ALREADY IN THE TRADING QUEUE, stick theirs AHEAD of the ones ALREADY in the queue, so they can stick themselves in the middle of buy/sell orders and "split the difference", making a profit for themselves that they truthfully did not earn in any way. When anyone complains about HFT traders, they refer to this as "adding liquidity to the market", and imply that if the HFTs weren't there suddenly nobody would be willing to make trades.. which of course is complete bull.
This guy figures out how to "trick" the HFT systems (no real people) and all of a sudden he's in trouble? I honestly think this is a load of crap, it's the HFT system writer's problem if they can be tricked by this. HFTs put up and pull trades without executing them ALL THE TIME to both try to have trades ready for things they think may or may not happen, and to try to manipulate other HFT systems behaviors (after all, multiple investment firms have HFTs and they'll be competing with each other for this money.)
2) The HFTs did cause the flash crash. The fact that his unusual input may have triggered it (if it even did) doesn't change the fact that their software malfunctioned and caused the crash.
"2) The HFTs did cause the flash crash. The fact that his unusual input may have triggered it (if it even did) doesn't change the fact that their software malfunctioned and caused the crash."
Isn't that like saying "The lorry driver did cause the crash. The fact that the boy running out into the road in front of him may have triggered it (if it even did) doesn't change the fact that the driver swerved and caused the crash."
I'm not saying that I think he's in the wrong, but you cannot use arguments like that to exonerate him.
Look,
Traders are trying to make money from buying and selling contracts over time. How long a time is now down to sub seconds as a way to make a small amount with little risk and then do it enough times that it adds up. That's not illegal.
Placing a trade and then making bids you have no intention of filling just to get the market to move in a certain direction? That is illegal.
The waste of time thinking you could stop the extradition.
He traded in the US and is bound by the US laws, specifically the Dodd-Frank law which outlaws his tactic.
Pretty dumb move on his part.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
