There is nothing like a flash event to get people excited and for news outlets to dust off and update the old “blame the algos” stories for publication. Equally, there are members of our industry of a (ahem) certain generation who are quick to jump on the bandwagon with the rejoinder “it wasn’t like that before the machines”.
As a member of that “certain” generation I can assure readers that it is utter nonsense and without wishing to get all “Oliver Stone” on you all, I am fairly confident the decision(s) taken that led to last week’s price action (as it was with the sterling flash crash in October 2016) emanated from a human.
The easy route is to blame algos, and if people are willing to argue that algos might exacerbate events I would have more sympathy (not much – I still think all they do is make things happen quicker), but to blame them? I simply do not get it, because this market is not exclusively driven by machines – it is not an AI market. Most of the time the computers are executing the commands of a human and when they are not they are managing risk, most often over a short time horizon.
Yes, there is a chance that risk management functionality could have gone wrong, but I think in the one week since the event we would have heard if a firm had a rogue algo. All I have is a few stories of firms getting hurt through their pricing activities and a few from traders who made their first couple of months’ budget for the year thanks to the opportunity presented by the irrational move.
If it was an algo that triggered the move then that algo was chosen by a human to exit a chunk of risk, it did not pick itself and it did not (assuming there is no malfunction as discussed) suddenly decide to sell USD/JPY and AUD/JPY. I would, as I noted earlier, accept that some algos may have exacerbated events through their use of correlations, and that is why we saw events in other markets that were used to hedge out but had just as little liquidity – which is the key to everything that too many commentators are ignoring.
Last week’s shenanigans happened at what has always a witching hour for foreign exchange markets, that two or three hour window between the New York close and the mainland Asia open. Markets have always done occasionally silly things in that window and traders have always known that. Indeed, in the past there have been suspicions that some were not above targeting stops in that window, all you had to do was contact a bank in the region, hit them with enough risk and the stops would be run.
And that’s before you get the concentration of orders in a few hands during that window. It may have been unfair, but there were times in the FX market when a trader in London would not go home with risk over a weekend not just because there were a lot of events out of hours but because their chances of surviving the Monday open in Australasia were thin – one of my old colleagues used to joke that any order left in that time zone shouldn’t be called a stop loss it should be called a target!
It shouldn’t be accepted either that human traders would react any differently to the algos, they reacted the same, just slower. FX markets used to move a lot more than they do now because the information channels were slower and therefore selling and buying was more sustained – it didn’t happen in bursts like it does now. It did happen though, regularly, and again ask the old guy on the end of the desk about how a voice broker could get everyone excited by giving it the large one (technical broking term) when shouting “GIVEN” or “TAKEN” down the line to 10 banks?
Equally if you thought you were dealing with an informed trader, the voice trader would not only clear out the risk they had been given, they could have gone with it. The fortune for the FX market was that back then there were so many more players with risk appetite and a view, or who at least had to make a price because they were obliged to. In today’s markets there are not the risk assumers nor the number of market makers there were, but the basic behaviour of the market is still the same.
So there is nothing new in markets moving in the Australasian window and often they did so as the result of a deliberate act. I am not directly suggesting such a thing happened last week – I have no evidence either way – but it should be an avenue of exploration.
I have long argued that the report into the sterling flash crash in 2016 was ineffective (not deliberately so I should stress) because it only focused on the event itself and not the triggers. Sources have told me that their observation of the activity last week indicates to them that a limited number of participants were involved. There could be several explanations for this. Firstly it could be that some players were, as my colleague Galen Stops pointed out in our weekly podcast, in that window when they shut down to roll to the next day.
Secondly, there are only a limited number of participants active in the markets in that window – liquidity recyclers tend to disappear for a few hours so we probably have a more realistic picture of actual liquidity than at any time of day – and often it’s not pretty.
A third explanation could be someone misjudged conditions and tried to execute a customer order that was simply too big for the market at that time; and fourthly someone was targeting a level.
The last is obviously a no-no under the FX Global Code, but as with so many areas in markets, proving intent is extremely difficult, if not impossible. This is, however, what I think happened – it’s either that or we have institutions with hopelessly inadequately trained staff who actually believe the hype about FX being 24/five-and-a-half and that liquidity never changes!
The Global Code itself recognises the issues around this window. I am not sure if my copy is bang up to date, but the fourth (negative) example for Principle 12, which deals with orders likely to disrupt the market, says the following:
A hedge fund is long an exotic Euro put. The currency has been weakening towards the option’s knock-in level during the New York session. Knowing that liquidity will be lower during the Asian session, due to a major holiday, and intending to knock in the option, the hedge fund leaves a large Euro Stop Loss sell order for the Asian open with bank A at a price just above the knock-in level. At the same time, the hedge fund leaves a limit buy order with bank B for the same amount of Euros but at a level just below the knock-in level. Neither bank A nor bank B is aware that the hedge fund is long the exotic Euro put.
The examples in the Code were developed by experienced market professionals who had, in all likelihood, seen, or experienced, such action at some stage in the past (as indeed I have). That means it has happened and whether we like it or not probably still does.
Where that example doesn’t match what happened last week is the second order, which clearly wasn’t triggered in the mayhem of the initial move, my point is more generic – there are people out there who, for a variety of reasons, target levels and they know that if they pick the right time (and last Thursday was a Japanese holiday – not that Tokyo was awake when it happened) they will get some serious bang for their buck.
The creators of the Code know that there is a liquidity problem at a certain time of day, so therefore you’d have to think that market participants would as well.
Personally, and I might as well throw in my tuppence worth, I think this was yet another clear out of retail carry trades in AUD/JPY, which is a favourite of that group. Who did it is open to question but all it takes is a little knowledge of how the market is positioned and how vulnerable it is to a move, before picking the right time to attack and there you have it.
The moves last week were no different to those of years gone by when we would see (slower) big figure moves in the Australasian window before Asia got in. The real question is, do the authorities want to do anything about it? I am not sure they should, after all this is meant to be a consenting adults market and if you’re in a crowded trade you’re always vulnerable.
Aside from that though, to mind mid the chatter about algos is white noise. Algos are not the problem and to use an analogy I would simply point out that people used to kill each other with clubs and other blunt instruments and then guns came along. They still kill just the same, just quicker.