Australia May 'Pause' Trades To Tackle High-Frequency Trading 342
angry tapir (1463043) writes "The Australian Securities and Investment Commission (ASIC), a government financial watchdog, is reportedly contemplating the idea of implementing a 500 millisecond delay on trades in an effort to put the brakes on high-frequency trading. ASIC last year knocked back the idea and stated that fears about HFT were overblown. However, in a government inquiry today representatives of the organization said the idea of a 'pause' is still on the table."
Better article (Score:5, Informative)
There's a gripping article over at the NY Times (adapted from a just released book) that explains very well the pitfalls of HFT, where the problems are mostly due to the haves and have-nots, just like in most things. The article is at http://www.nytimes.com/2014/04... [nytimes.com]
Not having a level playing deck in an exchange is a major problem for the correct functioning of said exchange.
Article is not very clear. (Score:4, Informative)
The Australian Financial Review today reported that ASIC had told the inquiry into Australia's financial system chaired by David Murray the regulator would consider introducing a half-second clamp on trades to remove HFT's speed advantage.
HFT work by seeing the order in one exchange at one price, and the same thing is available in another exchange for a slightly different price, simultaneously buy in one, sell in another and pocket the difference. Plain arbitrage, something Commander Vanderbuilt apparently did back in the days when news traveled on horseback during the day time. And he traveled at night in his sailing ship and raced ahead of the news, dumping bonds from bankrupt New York corporations.
These exchanges communicate the prices between themselves and take slightly longer than 350 milliseconds for the news to travel between exchanges. These big trading companies have faster access to both exchanges and are able to act on them. Would it be enough to delay all orders by 0.5 sec? Even if one trading firm sees the price difference, before it could act on it, the news would have traveled and it could no play micro second arbitrage.
This is my understanding. It might mean any trader must hold the instruments for 0.5 sec before trading it again. Not really sure what the article means by clamp.
HFT = a cost to society (Score:5, Informative)
What really annoy me with HFT, besides not being "fair", it that it as a cost and that the society doesn't benefit from it.
Building a stock exchange with top-notch computers if fine, since there is a need fulfilled here for our society.
But building new warehouses as close as possible to stock exchange computers to house top speed fiber connected computers, just to lower the delays from 600ms down to 10ms or so, to allow HFT, is a waste of resources.
No one needs that, it's just a smart way to build a sucking vampire over information systems. And this cost is always somehow reflected to society.
One big bank of my country paid a lot to move all its crucial infrastructure abroad, in such new buildings, to be able to compete in HFT.
Who's paying for those efforts? The company, the bank, instead of doing something more useful to society (investments to improve their services, etc).
Re:Yikes (Score:4, Informative)
The concept is that the market is supposed to be for investing. Investing implies certain loss of liquidity (no idea what you mean by loss of value). That said, see my response.
Re:Won't work (Score:5, Informative)
"... the illegal practice of a stockbroker executing orders on a security for its own account while taking advantage of advance knowledge of pending orders from its customers."
serious problems with networking equipment in HFT (Score:4, Informative)
http://www.nytimes.com/2014/04... [nytimes.com]
this article explains in depth what the problem is. the SEC has now been alerted to the problem, and is investigating. the people who found the issue believed originally that this was deliberate, but it actually just turned out to be a systemic problem of the speed differentials between different routes that high-frequency trades come in at.
what they originally discovered was that they could see a price on a screen, but the moment that they put in the bid to a number of brokers, the price would DISAPPEAR. they thought that this was deliberate, that someone was scamming them: it turned out that this wasn't true, but it took a couple of years of investigation to find out. what they did was they put in *individual* bids *directly*, and found that they were accepted. they then investigated various combinations, introducing delays into the bids, and found, amazingly, that it was down to the *time of arrival at the exchange* of their bids as they were sent via numerous brokers.
so it was only when they invented a tool (which they called "Troy") that *deliberately* introduced networking delays, such that the bids would (as best they could manage) arrive within milliseconds of each other at the exchange, that they managed to trade successfully.
if however any one of those bids happened to go via a different ISP, or a different router, or any other random combination, then the bids would *FAIL*.
the problem it turns out is that these delay effects are well-known. most of the money in high-frequency trading is therefore made by seeing a slightly slower broker's prices, then putting in an undercutting bid *knowing full well* that the other broker has a slower network. and this aspect of high-frequency trading is what is currently under investigation by the SEC.
*this is why the introduction of networking delays is so absolutely important*.
the people who discovered this phenomenon basically had to set up their own independent exchange in order to solve the problem. they needed to introduce a delay of 350ms as a way to make things fair for everyone. they did this by basically putting in 38 miles of fibre-optic cable in a shoe-box in the basement of the server farm that they leased.
it turns out that once investors discovered this, they began *specifically demanding* that their trades *exclusively* be brokered through this new exchange that had this 350ms shoe-box delay. it actually caused a lot of embarrassment for a number of brokers and trading houses because the brokers were explicitly disobeying their client's instructions, because the brokers didn't understand how important this really is.
anyway: you really have to read that article (or the book) fully because it's quite complex, and it's basically an inherent flaw down to the fact that the internet (TCP/IP) is routed randomly, thus introducing gross unfairness that has become the subject of intense investigation, very recently.
so yes, *all* trading should be done with at least a 350ms delay.
Re:Won't work: Nope the only way it will work. (Score:2, Informative)
I read through most of the replies. Most of them don't consider the 500ms is both for placing an order and canceling an order, which is why it will work. Read further below. This post goes the extra mile and suggests we don't need any intraday trades. I will counter this post and clarify the others.
This post completely misses the point of HFT. "HFT" is a "good" thing in certain instances. It produces liquidity, within the range of prices. Now remove computers from HFT; it used to be real humans who did the HFT stuff instead of algorithms. So you could hit the exchange with a million dollar order and not see the price move from 100 to 150. Some other exchanges handle this with market makers or what have you in the pink slip world. So, every single market participant including the guy who holds stuff for years, needs the liquidity at least twice in his lifecycle or if you are a hold till your grave guy, at least once.
Before anyone accuses me of being in favor of HFT. Let me quickly point out why the 500ms delay will work. There is an exchange in the US which is privately run which uses a 350ms delay, by using a 30 mile box of fiber. Many of the answers below talk about arbitrage between exchanges. All of them or most of them missed why it will work even within the exchange!
If the current state of the market order book is:
10000 shares at $10
20000 shares at $10.1
30000 shares at $10.2
Remember the 350ms delay? Orders which are on the exchange stay for at least 350ms. The computers cannot have orders that stay for 20ms and then leave. HFT's can absorb or are designed to absorb a hit on the market book. So the first 5000 or so orders or whatever the market book is at a price may result in a loss due to some investor pumping in cash at a price point. It is usually the big fish which have bigger orders. When such big orders hit the market the HFT's are turned off. Why? The HFT's are just trying the squeeze out 0.1 or 0.4 or whatever minimal profits they are aimed to get out. So a big move in a particular direction due to big funds operating is not the right time for the HFT's!
Now all of that is tangential. Let's get back to the 500ms delay.
Someone wants to buy 60,000 shares at 10.2 The order arrives at the exchange and is on the market books and is executing at 10.1. Everyone can see this. The HFT's in the above case will see this @10.1 the orders at $10.1 and $10.2 will simply vanish within the milliseconds it takes for the program to realize it is time to vacate the floor to the GoldMans of the world. So HFT's "cheat" the market off liquidity. The other arbitrage context applies here, but I won't go into it as other posts explain why. Since there is a 350ms delay, the order book can't change for 350ms even after the 10000 shares have been bought at $10. So the playing field is leveled and the HFT's will have to provide a "useful" market function as well. Provide liquidity even within the exchange and can't just turn off when there is a huge order. Please note that 10,000 shares may mean nothing to these HFT's. Scale accordingly.
So the key is it takes 350ms to send and order to the exchange and 350ms to cancel the order! There in lies the leveling of the playing field. It is actually the best way to prevent the HFT's from being what they are today. Playing on the markets without contributing liquidity! Works brilliantly!
So it is not a game of 350ms + few seconds out. The computers are stuck for 350ms! even when they have further information. They cannot influence the markets by pulling out trades! Practically, the orders will remain in place for 350ms + latency time of the the fiber outside the market for a cancellation request placed as soon as the other order hit the market. So you induce a delay of almost a second back and forth. 350ms for a order to go in. 350ms for an order to be cancelled. 700ms will kill the HFT's which don't contribute to liquidity. But will keep the good human controlled HFT's in play.
I am sure, some of you will point out
Re:Won't work (Score:4, Informative)
Utter rubbish.
I used to work for a high frequency firm, and I can tell you the downwards strategy is as important as the upward strategy. While in theory it is true that potential gains on the upside are unlimited and potential gains on the downside are limited to the share price this presupposes that a stock price can go to infinity. Realistically, this is not true, and realistically barring a catastrophe a stock does not go to zero. Even then a stock does not go to zero immediately when becoming worthless (see Dead Cat Bounce [wikipedia.org])
HFT is focused in on short term moves. On a short period of time, say the order of seconds, it will dominate. On a longer period of time, say hours, days, or weeks, investor sentiment will dominate. Even when HFT goes completely wonky and accidentally manipulates the market, over a period of days investor sentiment will still dominate.
Additionally, stock ownership represents ownership in an underlying company, that on average will experience growth. As profits grow, the value of the shares grow. Therefore growth in the present value of stocks is the natural state. Additionally, while you might take the counter argument that not all stocks should grow, most indices will cut out underperforming stocks. Therefore, an index is a bad indicator for the stock market at large because you are artificially selecting for winners and losers.
Big Red Button (Score:4, Informative)
This is currently the problem. Zero liability currently. There have been a number of LARGE examples of this, where things have gone awry, and the company loses like 500 Million. The response has been to halt trading, and reverse all the trades. To me this is cheating. They may have lost, but that just means that someone else was the winner.
If people want to use these methods, then they take the risks. They don't get to call a "redo" because things didn't work out in the way they thought it should.
After a couple of big losses like this, people might think twice about using such a service, or at least account for it within their threshold of risk. They do not own a licence to make money.