Accurate article? Yes I think so. Hairsplitting? A bit. Any content about the big picture? Afraid not.
Algo trading has been around longer than HFT. It was invented to protect the information that that a big order was being executed. This avoided the risk of front running by handing the order to humans or scaring liquidity providers by executing it all at once.
HFT came about when computerized exchanges began to compete with each other for business. Nowadays you go hunting for liquidity. Speed differentials are more important.
I think the OP is likely to agree up until here.
Today HFT in its worst form amounts to high speed computerized rumor mongering. You game the market by bluffing orders and trading faster than your customers.
The regulators will never catch on and try to fix it even thought the remedies are many and simple. Moreover our attitudes about who owns information make it very difficult for us to even consider these simple solutions.
There are a massive number of HFT shops. Most of these are prop-shops rather than funds, as in they trade their own money and don't take investors. They are physically unable to front run their customers because they simply don't have customers.
I will give you that there are some big banks getting into HFT now, and that's a different story, but a statement like "HFT in its worst form amounts to high speed computerized rumor mongering." is wildly inaccurate and shows a complete lack of understanding of the industry.
Furthermore, there is a massive difference between algo trading and HFT. Technically speaking, yes, HFT falls under algo. However, HFT is about speed and making very little profit many times throughout the day, usually by providing liquidity.
Algos on the other hand, especially things like high end models aren't meant for HFT because they take larger amounts of time to run (ie: backtesting). These are used (for example) to determine misprices in the market that will pay off heavily in the long term, or (as others have mentioned below) to buy/sell a large quantity of shares in a way that it won't move the market in the other direction, rather than make an immediate, albeit tiny, profit.
I understand the industry quite well and from experience. BTW My only daggers were aimed at HFT.
Let me be clear. Anybody that puts an order in not expecting to get executed but to create favorable conditions for their next order is engaging in what I call "high speed rumor mongering". There are lots of variations of this game. Never happens? Always happens? You tell me.
But markets aren't chess. "Rumor-mongering" is a legal term I chose on purpose. It is generally prohibited because it destroys liquidity in markets, which hurts investors and issuers. The term is intentionally objectionable, but not inaccurate. Just don't trouble the regulators to figure it out.
OP's misunderstanding of prop trading vs funds aside, I still think it can be argued that "HFT in its worst form amounts to high speed computerized rumor mongering".
For instance, take a look at the Nanex article What is the Bid/Ask spread of this stock? [1]. Rumor mongering in this case being gaming the weakness of the NBBO.
On a side note, I highly recommend reading through Nanex's Strange Days research section if HFT related market anomalies interests you at all [2].
> The regulators will never catch on and try to fix it
Can you elaborate a bit (even at a high level) about why HFT is particularly bad? I always hear everyone deriding HFT, but never any specific reasons why aside from people complaining about them not contributing anything to society.
The major issue is HFT can create wild market swings with little to no basis in reality. It's actually possible for them to suck up all the outstanding bids over a few seconds using small amounts of capital and while a human might desire to sell if a stock goes up by 2% the seconds or minutes it takes US to make that choice is eons for the algorithms. The net result of this is actually less liquidity as someone buying or selling can't place large orders on the market or the algorithms with eat them alive. Also, they are often setup to simply stop all actions if the market deviates to far from the norm which pushes things even further out of whack.
Are HFT firms really doing enough volume to even move the needle, or are they simply targeting lower end stocks/securities? I guess I'm still in a forex mindset where billions is not considered an especially large amount.
For one who works with HFT systems, an extremely refreshing clarification.
Though I do get a chuckle out of the extremely bombastic stories ("MAN OBSOLETE? COMPUTERS TAKING OVER!!!"), it's nice to see the record set straight. Sadly this will get 1/10000th of the page views of the garbage articles it dissects.
Even if you must insist on being so wildly rude, couldn't you at least offer up some substantiation for your anger, something with a hope of a discussion?
Absolutely terrible blog post. Saying algorithmic trading isn't HFT is like saying a bird isn't an ostrich.
HFT is a subset of algorithmic trading. It's as simple as that. Not all algorithmic trading is HFT. But all HFT is algorithmic trading.
Algorithmic trading is any type of trading done based on an algorithm, and not based on "traditional investing principles". For example, "buy when the 10-day moving average crosses over the 20-day MA, and sell when the 10-day it crosses below 20-day". One of the most famous types of this is Richard Dennis and the Turtle Traders, where a successful commodities trader took a bunch of ordinary people and tried to turn them into traders using this method.
I'm also an algorithmic trader. Not a wildly successful algorithmic trader yet, but I'm still learning and growing, and I love it more than any programming venture I've ever been involved with. And I haven't taken a catastrophic loss yet, so that's good. I trade on 3 separate markets using different algorithms, and for the most part it is fully automated. I'll hold futures contracts anywhere from 1 seconds to 20 mins.
HFT is several orders of magnitude more intense. For the most part, they are types of arbitrage, where they can arbitrage a few cents worth of difference in stock prices between different markets, and make a few pennies per 1000 shares. Or they will arbitrage between the price of an ETF or futures contract and the underlying basket of stocks that it represents. These are the ones that need colocation and trade on the millisecond. The more nefarious ones are the ones that game the system, by "quote stuffing", by frontrunning large orders by institutions, or by creating volatility through momentum-trading.
Is there a downside to algorithmic trading? Sure. Algorithmic trading is what has turned the stock market from a predictive market of future earnings, into essentially a casino, where the predictive nature of the markets is completely dead. Instead, it's about thousands of computers running random number generators and picking up nickels every 10 ms. This is why I believe there will be market crashes every 7-10 years, and long term investing is dead.
But unfortunately, this is the reality of the system that we live in, and we have to adapt or die. Or you can just buy bonds and get a stable 3-5% return every year, which is nothing to shake a stick at.
What evidence do you have that the stock prices have recently become less predictive of future earnings? Was the market performing its predictive duty in, for instance, September of 1929?
Previously the vast majority of investors were buy and hold, where they believed in companies and that their earnings would increase. Of course, there were always traders like Jesse Livermore that traded off of order flow, but those were the minority. Most were like Warren Buffett where buying and holding was for the best.
Now the majority of trades are done by algorithms with no biases at all towards the earnings growth of a company, be it HFT, or statistical arbitrage, or through other quantitative models.
HFT accounts for 75% of daily volume, and by definition, HFT does not take into consideration things like future earnings growth, etc. For the most part, they simply find arbitrage opportunities and profit from them.
So stocks being bought and sold are not done based on the earnings potential of a company. Case in point, Citigroup before the reverse split was trading hundreds of millions of shares per day, not because so many people believed in the company, but because it was dominated by rebate traders, HFT, day traders, etc. I believe companies like Fannie Mae and Freddie Mac were trading millions of shares before they went pink slip, even though it was known that they were defunct.
You're talking about volume, not price. Increased volume should, if a market is performing "properly", accompany new information. If there's new information, there's a reason to trade. The examples you cited seem like evidence of the stock market's predictive value - for instance, increased volume in Citi stock was a piece of evidence that something about the security was expected to change.
Moreover, it's not true to say that stat-arb guys don't care about future earnings growth - they just use statistical methods to project it. To use a simplified example, a stat-arb guy might automatically buy shares in some small-cap automotive supplier if Ford rapidly increases in price. If the increase in Ford stock represents positive fundamental information about the auto industry, they've applied that information to the price of the supplier faster than a human would have and they'd make a profit. If, on the other hand, it represents concern about some scandal involving the Ford CEO, they've contributed to the noise and lost money.
I probably wasn't clear with my original statement. I said the stock market used to be a "predictive market of future earnings". What I meant more precisely was that the stock market used to be a market where people would make prediction about future earnings about companies. If you thought a company was doing well, you would buy and hold it, a la Warren Buffett. Some people traded order flow and other things, but the vast majority traded it as if it was a proxy for future earnings growth.
Over the years, that has changed. Once the internet hit and day traders became more common, it became more and more about buying stocks that will go up and selling stocks that will go down. I think people forget that even during the 90s, commissions for buying and selling stock were in the hundreds of dollars, not $8.95 like today. It was expensive for retail investors to buy and sell stock.
The time period for holding stocks has decreased sharply since then, where a few years ago rebate traders would buy and sell stock to just get the rebate from the exchanges, and now to where HFT eat the lunch of those same manual rebate traders.
The vast majority of trading done on the markets today is not done based on the quality of the company or the quality of the earnings, but based on how the stock will trade. Sure, there's still institutional trading, and I'm sure plenty of quant models make use of things like earnings growth, etc. And yes, things like news and fundamentals do cause prices to go up and down. But the vast majority of daily trading, 75% of volume, is done by trading entities that don't care about fundamentals, and only care about miniscule movements in the stock price. This is why I use volume as evidence, because it demonstrates that most trading done isn't done because of the predictive nature of the stock market for future earnings, but because of the extremely short term predictive nature of the stock price itself.
That's why I said that it's less about predicting future earnings grow and more about making nickels every 10 ms. C trading 500MM shares a day is like people rolling dice every millisecond in the alley way and exchanging money upon every roll. The other example that I was searching for but couldn't recall was when GM went bankrupt and the stock was still trading over $1. This was purely trading activity even though the "future earnings" of the stock was 0.
It's become a casino where probability theory dominate and less about "I drink Coke so I should buy KO".
BTW, I'm not saying this is good or bad. I just believe this is how the markets are. The same thing happened when daytraders entered the markets during the dotcom boom. I do believe gaming the system, trying to "break" the markets with quote stuffing, etc, is wrong, but fundamentally I believe that the nature of the markets have changed, and anyone who wants to get involved in it should understand the nature of the change. Anyone who thinks that they should keep their money in mutual funds and let the mutual fund companies sip 2-5% every year for doing worse than the markets, and then also exposing yourself to market crashes every 7-10 years, I believe, are fools.
I said the stock market used to be a "predictive market of future earnings". What I meant more precisely was that the stock market used to be a market where people would make prediction about future earnings about companies.
These two statements are not even discussing the same thing.
One statement discusses the practical computational power of a system. The other statement discusses the motivations of a majority of the people participating in that system.
Historically there have been people trading off various schemes other than fundamental value forever, whether they be the public rushing into the bubble before the 1929 crash, or the chart-trading technical traders who have been around since at least the 1980s. And dont forget the people manipulating and cornering the market. None of this is new...
Not sure exactly what you're asking, but the fact is that trading has been drying up. Since 2007, volume has plummeted, and August I believe was one of the worst, if not the worst, month for people withdrawing from mutual funds. Less and less retail are staying in the markets because of the volatility.
So I don't know if it will get cheaper to buy and hold. It will definitely get more volatile.
From the article and other sources I've seen before, it seems that algorithmic trading is not necessarily high frequency, but that high frequency trading is necessarily algorithmic.
In which case is this not a rather thin hair to split?
Not really. HFT is a subset of algorithmic trading, where many small orders are placed to take advantage of intraday (or intraminute, or even intrasecond) shifts in the spread. A large strategic order executed through an algorithm is a different creature altogether.
The big problem when you place a huge buy or sell order is that it shifts the price in a direction you don't want it to go. For a large sell order, the price goes down, as the market becomes skittish about the security. For a large buy order, the price goes up, as the market becomes bullish and arbitrageurs quickly buy up securities to resell to you. So, many traders use algorithms to hide their trades. The purpose of these algorithms is to avoid volatility, so they shouldn't be dangerous to the market, as long as they're designed correctly.
(Although, to be fair, algorithms may automatically stop trading when the market becomes too volatile, which contributes to flash crashes by reducing liquidity.)
"so they shouldn't be dangerous to the market, as long as they're designed correctly". Even if we accept the author's position, why would we also make the assumption that this software is designed properly and bug free?
I guess it depends on the stakes, and how well the traders understand (the risks of) software development to pay for the quality, testing, and maintenance.
"This isn't just some human lives we're playing with, this is serious! It could cost us billions!"
HFT is about making money by supplying liquidity or making money by exploiting the portions of the system which are designed to supply liquidity. You don't want to end up with a position. You want to net out the day with no position in any stocks. HFT is done by an HFT fund which is usually trading off credit extended by an agressive hedge fund who they split the profits with.
Algos are about getting a trade done at the least cost. You want to buy 100M shares for fund XYZ or sell 10M shares for fund ABC. Algos are run by brokers on behalf of their buy-side clients. A simple algo is something like TWAP, the time weighted average price, which attempts to buy a large position over the course of the day gradually. The purpose of this algo is to avoid moving the market. This is suitable for the "build up a large position because I'm bullish on this industry" type of trade.
Another algo might be IQx, a package offered by CovergEx which tries to quickly execute the trade by sending it out to multiple dark pools and the major exchanges at gradually worse prices until the whole order fills. (Or at least that's my reading of the marketing material.)
Example: there are algorithmic strategies designed to make large trades.
Let's say an institution wants to buy 10K shares of a thinly traded stock (lets say, 100K shares per day). Placing a single order for 10K shares will send a signal to the market that someone wants to buy, and other traders will see it.
There are special algorithms designed to purchase the specified number of shares without making too much of a market impact. For example, in this case, the strategy would send 100 share orders rather than presenting the full 10K interest at once.
for the googlers: there are all kinds of technical terms like implementation shortfall to give more info
I always find it interesting how much vitriol there is against automated trading, even among programmers. Too many people seem to believe that a small number of, ultra resourceful, nefarious folks are using unfair means to "game the system." The truth, as usual, is less interesting.
Doing this type of trading doesn't require millions of dollars and teams of PhDs. You don't have to know the right people and you don't have to know any secret handshakes.
Critics of high frequency trading are almost always misinformed. Some of the most informed critiques I have read about this stuff are the following books:
"A Demon of Our Own Design: Markets, Hedge Funds, and the Perils of Financial Innovation" by Bookstaber
"Traders, Guns and Money: Knowns and unknowns in the dazzling world of derivatives" by Das
And Nasim Taleb's work.
-------
More to the point, the author is explaining something very basic (which journalists don't seem to understand):
-Algorithmic trading is NOT a general term for all trading done with algorithms/computers. It refers to telling a computer to EXECUTE a specific trade. In other words, when your retirement fund decides to buy A LOT of AAPL, they naturally need to spread that trade over the whole day (or even several days). In the old days, human traders used to do it. Now it is mostly done by computer programs.
This is different from the kind of trading where a computer decides WHAT to trade (NOT HOW to trade). This kind of trading involves so many different strategies that it is silly to lump them together.
There seem to be other misconceptions:
-The best and the brightest are working in Finance, instead of doing things more beneficial to society.
A quant colleague of mine, who has a PhD in Physics from an Ivy League school told me that he, and many of his friends, left academia because there were simply no positions for them.
-75% of trading is now automated, it is just computers trading with each other.
I hope someone will correct me if I'm wrong but I have never figured out if this 75% includes algo trading. If it does include algo trading (my guess is that it does), then I'm surprised it is not 100%. That is like saying 95% of TV channels are controlled by remote-control devices.
-People seem to think that wall-street traders make their money by "trading ahead" of mom & pop investors: your Dad buys 1000 shares of microsoft, a wily trader puts your dad's order on hold, buys it for himself, raises the price, sells his shares to your dad at a higher price...thereby making money off your dad.
Your broker is not allowed to 'trade-ahead' of you. At least in the places where I have worked, this is taken very seriously. Interestingly, high frequency traders (who are most frequently accused of this) don't actually have access to customer order-flow. High frequency trading hedge funds don't generally have any client orders. Places where the two co-exist are forced to have seperation. Traders from one department cannot share information with the other. As more and more client facing firms (sell-side) become automated, the chance of them actually coding up such cheats is even dumber.
Flash trading is often given as an example of people, in cahoots with exchanges, trading on others' order information. As far as I know, "flash" functionality exists to help clients trade more effectively. Large traders are VERY concerned about letting the whole market know that they are interested in some stock. Some exchanges offered the following functionality: if you are interested in buying a stock, you have the OPTION of giving other members of the exchange a chance to trade with you. If no one takes you up on the offer, then the order goes to the wider market.
I have to admit that a laywer friend told me that he opposes this functionality at his firm. If someone _really_ needs to know more, I suppose I could ask him to explain.
-High frequency traders trade so fast that mom & pop simply can't compete with them. Their computers/networks are just too fast and they can get closer to the exchanges than anyone else.
High frequency traders are not competing with mom & pop, they are competing with market makers. If two people hear a news item, the one closer to the exchange will naturally get the trade done faster (presumably at a better price). The same is (generally) true of people on the East Coast vs rest of the country (let's assume US financial system). The same is true of people who can click their mouse faster. Besides, there is no moral reason your Mom should be able to dump her Enron stock faster than Joe Trader.
etc., etc., etc.
-------------
I should add that I am actually not at all comfortable with the role finance plays in world economy. I can't call myself a critic since being critical requires more complete understanding.
I am specifically opposed to things like direct market access. This is where any Joe Blow can use an API to setup his trading system. If he accidently leaves an infinite loop in his code, he can cause real problem. I remember sweating bullets (and almost trembling) when my boss asked me to flip the switch on the trading system I wrote. In reality, there are at least some protections built in to keep this from happening. However, I would like to see more uniform, consistent and better advertised rules.
I am also against the ability to trade by borrwing money from brokers (margin trading or leveraged trading). If an individual trader screws up, they wipe themselves out. If they borrowed money, then the consequences of their bad trades starts to seep out to others. If more than a handful of traders, trading on margin, go belly up, the lender could be in trouble as well...you can see how this could ripple across a system.
Closely related to allowing trading on margin is reliance on models. Say you have calculated that two stocks always move together. You _and your lender_ are so sure of this correlation that they think of it as the truth. What if your calculations or your assumptions were wrong? The consequences of this mistake may not be linearly related to the risk you thought you took. Read Nasim Taleb's work on this for more.
Finally, those who smell something fishy should broaden their concern beyond just modern trading system or even complex derivatives. I can see no principal, within the framework of free markets and individualism, which leads to condemnation of ever more automated and faster trading, more complex instruments and more dependence of finance. The best moral principal, I can think of, which opposes the current state of affairs, is the one uttered by Martin Sheen's character in the movie Wallstreet: "Create, instead of living off the buying and selling of others."
Re: the "best and brightest" it's more at the undergraduate level then the PhD level. My brother has tippy-top grades in physics at one of Harvard/Yale/Princeton and legit research experience in nano-tech, and he like many of his friends in similar positions are choosing between going into industry or R&D and going into finance. The lure of $120k the first year out of school and guaranteed admission to a Harvard/Stanford/Wharton MBA is hard to compete with.
A big part of the problem is systemic. The researcher that discovers new technology gets a nice $30k bonus. The owners of the capital get the millions of dollars resulting from that invention. So if you're a bright physics student in the US, why on earth would you pursue R&D? It is far more remunerative to work for those who own the capital figuring out new ways to move money around.
There aren't jobs in academia, but there are jobs in industry. Intel, IBM, etc, hire people with physics backgrounds to work on new types of memory cells, etc.
But why would you? If you're a 1 in 100 physicist, you might make a discovery that will net your company tens of millions of dollars, but our laws are such that you won't see any of that money. Anything you invent will automatically belong to your employer. Meanwhile, even a 1 in 10 trader can stick with an investment bank long enough to bring in some $1 million/year paydays until exiting to a cushy corporate finance position.
In other words, as long as it's more profitable to play with the government's Ponzi scheme we're going to suffer brain-drain from R&D and other actual industry.
"-75% of trading is now automated, it is just computers trading with each other.
I hope someone will correct me if I'm wrong but I have never figured out if this 75% includes algo trading. If it does include algo trading (my guess is that it does), then I'm surprised it is not 100%. That is like saying 95% of TV channels are controlled by remote-control devices."
From the Foresight project homepage (http://www.bis.gov.uk/foresight/our-work/projects/current-pr...): "For example, today, over one third of United Kingdom equity trading volume is generated through high frequency automated computer trading while in the US this figure is closer to three-quarters."
>A quant colleague of mine, who has a PhD in Physics from an >Ivy League school told me that he, and many of his friends, >left academia because there were simply no positions for >them.
Then why not work in any other industry? Going from the worst paid to the best paid occupation is not really something you have to push most people to.
A good friend of mine became a quant simply because they where the only people remotely interested in hiring someone with an MSc in math and then letting them actually work with interesting high level math. The money or any actual desire to work in finance never really factored into it.
btw, the work is not always less interesting. Think of it this way, this industry has ambitious people from pure science PhDs, mathematicians, statisticians, programmers, MBAs, idiot nephews of rich uncles...everyone is competing.
If you are successful, you can justify building a huge Hadoop cluster, experimenting with hardware TCP/IP processing, buying (or storing) petabytes of data for statistical analysis. Pretty interesting stuff for a geek.
Obviously, not every one in the industry gets these chances and all this not necessary. I know of people who earn their living doing automated trading in ... visual basic (not VB .NET) :)
We all can agree that increased liquidity is good for everyone, mom and pop, traders, investors, America, etc., Of course one can argue whether we really need sub 100ms liquidity but that is not the biggest problem.
The biggest problem with HFT provided liquidity is that it is far from a sure thing as the Flash crash proved. The liquidity dried up so fast, because most players algorithms "said" the situation was too unpredictable so the easiest thing to do was to close up shop temporarily.
In older days, market makers on the floor were allowed to make the money from the spread with the understanding that if the things got rough they would HAVE TO stay in the game. Some got rich, some died broke, some jumped out of windows, but overall the game continued.
Increasingly, the role of a market maker has been delegated to HFT firms, but without any obligations placed on them.
Who is to blame for such state of affairs is a question someone else can answer better than me.
As far as I know, market makers still have this obligation.
I believe Taleb has argued that HFT liquidity disappears when it is most needed, like seat-belts which work all the time, except during accidents (his metaphor may have been different)
Paul Wilmott argued that this much liquidity is actually not necessary. If someone will have trouble getting out of a stock, maybe they will think twice about getting into it.
Of course, an HFT practitioner doesn't need to prove to anyone why their activity is beneficial to society. The burden of proof is on the critics to show why this activity is harmful.
> I am also against the ability to trade by borrwing money from brokers (margin trading or leveraged trading). If an individual trader screws up, they wipe themselves out. If they borrowed money, then the consequences of their bad trades starts to seep out to others. If more than a handful of traders, trading on margin, go belly up, the lender could be in trouble as well...you can see how this could ripple across a system.
In theory, shouldn't those giving the loans account for the risk and thus be protected from wiping out themselves?
This is exactly what happened during the stock market crash of 1929, and regulations were put into place in the 1930s to prevent excessive margin leverage that might result in liquidity problems at brokerages (and in the banks that lend to them). These regulations have been in place since then and probably mitigated the effects of the dot com crash in 2000. See http://en.wikipedia.org/wiki/Regulation_T as a good starting point for research.
Interestingly, I read somewhere that there were similar regulations regarding residential mortgage loans that were repealed during the 1980s, does anybody have a reference to this? I think that requiring a 20% equity/debt ration when originating or refinancing a mortgage loan probably would have made the 2008 real-estate crash look a lot more like the dot com bust and would have saved a lot of economic pain.
Even theoretically there are agency problems in many situations re: evaluating credit risk. And as we have learned in practice, people are just bad at gauging credit risk. Finally, the traditional theory doesn't incorporate behavioral economics, which IMHO turns a lot of the traditional precepts on their head.
This sounds reasonable, but I simply don't know enough about trading to determine whether it's true or not. Anyone care to comment on it? It seems that if there were no trading at all, the world would be worse off.
It's completely wrong, because it ignores both time and risk. It's worth real value to me to have money now rather than later. It's worth real value to me to give you some money, and you take some risk off my hands. And on the other hand, if you think my risk is lower than I think it is, and if you have money to spare, let's make a deal.
Remember, if both parties aren't better off, why would the trade even happen?
> Remember, if both parties aren't better off, why would the trade even happen?
That certainly is true for goods and services, but it just seems awfully abstract from, say, high frequency trading as a profession. How much value is that adding to the world, compared with, say, going out and creating something? I'm not saying it should be banned or anything along those lines, just that I'm not convinced that it's adding much to the world. Even things like games or movies make people happy, even if they're not 'productive'.
Well, party A says "I will be hurt if the price of this asset falls" and party B says "Well, I don't think it will, so if you pay me a smaller fee upfront, I'll cover any losses you make for the next T time". Multiply by thousands of assets (which a big mutual fund may well hold) and thousands of updates a second as everyone else in the market trades to their own ends, and you have the essence of HFT.
The material facts that stock values are supposedly based on do not change 1000's of times a second, though - it seems like it's trading for the sake of trading, rather than trading to best distribute goods and services to where they're desired and will be most effectively used.
One thing is deciding that IBM has a brilliant future and that the stock is a steal at the current price, and trading with someone who feels the opposite, another entirely to have thousands of transactions a second. Even the former transaction seems a bit zero-sum in that one of the people involved has the wrong idea and is going to either lose money or lose potential money.
That said, maybe I'm missing something - I don't know that much about HFT and stocks/finance in general, so I don't claim to have everything figured out.
It's not clear to me that the financial system as an enabler of liquidity/commerce/trades/etc. would be significantly worse if the granularity were slightly reduced, though. Say, run exchanges in a discrete-time world of 100ms timesteps. Are there real-world use cases where this would make the finance system unable to facilitate the economy?
If there are ways to make profit by trading at sub-100ms resolution, but a 1ms-resolution exchange is not any better at facilitating the outside-finance economy than a 100ms-resolution one would be, then it seems like HFT is solving problems of the exchange's own creation. It could even genuinely be solving those problems, but if they're problems that only arise in the context of extremely-high-granularity exchanges, and there is no practical benefit to such high time resolution, then why not just axe the problems?
I'd argue that there's more good than bad about algorithmic trading. People making decisions based on fear and adrenaline is much more dangerous than setting a pre-determined course and sticking to a mathematical model. Besides, it's not like they just set up these programs and forget about them. If there's some sort of flaw in the algorithm the trader isn't just going to bang his head into the wall while he loses millions; he's going to fix the algorithm.
Approx 99% of what is written about algo and high-freq trading is written by people who don't have a clue, have never worked in a trading environment (let alone actually traded) and think they're some kind of expert because they've read an article or two. Just remeber that the next time you read an article (or a comment about an article) on algo/high-freq trading.
Here is a video of an extremely interesting talk, entitled "Human Traders are an Endangered Species!", by Dave Cliff who's involved with the Foresight project:
Are there not hedge funds that attempt to use algorithms to spot longer term (ie. days, weeks, months) market trends and then buy and sell on that information? I think it's pretty crazy to say that humans will always make the decisions, computer can process more information faster then any human can. At some point computer programs will be able to make more accurate market predictions then humans, and at that point the "robots" really will be in control.
Isn't the claim that algorithmic trading will never replace human decision making basically equivalent to the claim that humans will never construct strong AI?
I'm not Jeff but perhaps I can provide some insight into the flash crash.
Also, what IS algo trading?
Is a margin call algo trading? Is a stop loss algo trading? What about technical analysis?
(Flash) crashes should be expected anytime you over leverage your entire economy.
See: George Soros and the pound.
Imagine everyone you know has widgets, and you realize that most people like to keep their widgets in a warehouse. You setup a warehouse that stores widgets and people pay you to store their widgets. Everyone loves it, no more widgets around the house cluttering things up.
Now since the widgets are identical and interchangeable you stop tracking whose widgets belong to who and throw them all into a big pile, and when someone asks for their 20 widgets you give them a random 20 widgets.
Now you realize that keeping all these widgets around is a massive waste of time and money since only about 1% of widgets are actually in use before being returned to the warehouse. So you tell your customers, "hey, I'm not going to bother actually keeping all the widgets, and I'm going to stop charging you to store your widget and instead I'll pay you to keep your widgets, if you store your widgets for a year I'll give you an coupon you can bring back to me at the end of the year and I'll give you all your widgets back plus 10% more".
This system works great and everyone is happy, after a few decades the coupons for the widgets outnumber the widgets by a factor of 10,000.
Now some asshole with a basic grasp of mathematics invents a computer program to manage widget coupons and it realizes that if it buys 1/10000th of the widget coupons and redeems them for widgets that no one else can actually redeem any other coupons. So your computer takes delivery of all the widgets in the world and then redeems one more widget coupon and everyone loses faith because the coupons for widgets no longer get you widgets and suddenly widget coupons are only worth 1/10000th of a widget. Suddenly everyone is mad at the guy with the computer and basic grasp of mathematics because their widget coupons only buy 1/10000th of a widget, and he's making bank selling everyone their widgets back.
Is the cause of the devaluation of widget coupons the fault of the algorithm, or the fault of the system that allowed more coupons than there are widgets?
The 2010 flash crash was caused by "blackhat" HFT traders trying to game the system.
It was shown that one, some, or many HFTs were involved in "quote stuffing" which is bidding for stock and then pulling the order, something like 100k times per second. This gave the appearance of liquidity and demand, but it was fake, because as soon as someone would bid for the stock, they would pull their order. But another use of this was to essentially slow down the "ticker tape" of the NYSE. What was happening was that the "ticker tape" that showed the current bids and asks was slowing down, and by doing this, some HFTs could make use of the latency arbitrage. Colocated HFTs got their quotes for the best bids and asks directly from the exchanges, but other people were getting their quotes from the NYSE, so they were behind. I believe they were something like 30 seconds behind, so what would happen is that the HFTs had full reign to take advantage of others being blinded like this.
Of course, the side effect of this was that they "broke" the markets. Because of this latency, the NYSE suspended the markets temporarily, which then had the unintended consequence of forcing all the bids and asks to flow into the smaller exchanges, which didn't have the liquidity to handle the orders. There were so many sell orders, that basically all the buy orders for some stocks got taken out, causing the prices to plummet down to 1 cent or something like that.
I'm expecting another flash crash to occur at some point, so whenever I see heated market action, I place a bunch of trades around 25% below the current stock price, which I believe is just above the limits that the exchanges would use to roll back bad trades. (Un)fortunately, it hasn't happened yet, but I'm sure at some point it will.
All HFT is blackhat, leeching off flaws in the way the market clears. There's no legitimate value to be offered by interposing between a buyer and a seller who are in the market simultaneously and would have done the trade unhindered. Traders are a net win for society if they cause better resource allocation, not the same allocation a fraction of a second sooner thanks to a greater misinvestment in network hardware.
Batches of buy and sell orders which are in the money should be executed hourly. The fundamental values of companies don't change more quickly than that. The rest is noise, not signal.
> I place a bunch of trades around 25% below the current stock price
Funny, we sat around trying to figure out the right price and this is what we came up with as well.
Too bad this doesn't show up in second level quotes, it would be a good leading indicator of what funds in general thought the chance of a crash was on any given day:)
I place a limit order, which I guess you could see if the depth were deep enough, but I could very well have programmed an algorithm to just monitor the prices and do market orders instead, which you definitely wouldn't see. I'm sure most traders do it that way.
The interesting thing is that on the day of the Flash Crash, if I'm not mistaken the ES futures contract bounced exactly off the 200 day MA. So, one thing that the flash crash revealed is a lot of algorithms programmed into the markets that would normally never have been revealed. So if you want, keep a floating order just above the 200 day and you might make a few dozen points in a few mins, just like May 6, 2010!
I guess you aren't the only one who started putting in trades within a reasonable margin around the market price after the flash crash. Thus it will be hard to repeat like that.
Perhaps a flash crash could happen in the other direction as well? I.e. flash boom, maybe by squeezing the shorters? In that case buying way out of money call options and putting in automatic orders to sell those options if the stock price goes 50% (or so) over last minute's market price would be a viable strategy?
There was something similar to this in August 2007. A bunch of quant shops blew up, and I believe if you look at the volatility index, it spiked up hugely during this time.
by the way, the original article (now) opens with "Algorithmic trading, including high frequency trading (HFT)" and not "Algorithmic trading, also known as high frequency trading (HFT)"
I think the author is actually wrong on this algo trading is defined as programs taking the decision, how fast this happens determines whether it is high frequency. Computer aided execution is called smart order routing..... just saying :)
Algo trading has been around longer than HFT. It was invented to protect the information that that a big order was being executed. This avoided the risk of front running by handing the order to humans or scaring liquidity providers by executing it all at once.
HFT came about when computerized exchanges began to compete with each other for business. Nowadays you go hunting for liquidity. Speed differentials are more important.
I think the OP is likely to agree up until here.
Today HFT in its worst form amounts to high speed computerized rumor mongering. You game the market by bluffing orders and trading faster than your customers. The regulators will never catch on and try to fix it even thought the remedies are many and simple. Moreover our attitudes about who owns information make it very difficult for us to even consider these simple solutions.