High-frequency trading

A few MR readers ask about high-frequency trading.  Senators are calling it unfair because some traders have access to more powerful computers,and better quants, than do others.  The traders with the most powerful aids get there first and make more money.  Here is a typical critique.  Felix Salmon is also skeptical.

I do not worry about high-frequency trading.  Telegraphs and telephones also brought their own, earlier versions of high-frequency trading.  As did stock index futures.  There are second-best arguments relating to hockey helmets and the like but that is the case with most forms of progress and greater economic speed.  You don't have to think that the current profits measure the current social value of high-frequency trading to argue that the overall trend should be allowed.  The correct judgment of efficiency occurs at the system-wide level, not at the level of the individual trading strategy.  The short-run story is that private profits exceed social returns but in the longer run the trading activity and liquidity brings increasing social returns and better communication of information.

I'm not a believer in the strong versions of efficient markets hypotheses, so I do admit that high-frequency trading, like just about every other trading strategy, can bring short-run "whiplash" effects on market prices.  But if you don't like it, you can trade yourself at much lower frequencies, which is probably what you should be doing anyway.  At the same time high-frequency trading smooths out or shortens many other cases of price whiplash.  High-frequency trading brings more liquidity into the market.  Call it "low quality liquidity" if you wish, but it still looks like net liquidity to me. 

The complaint is that this liquidity sometimes vanishes.  Maybe high-frequency trading can scare other traders out of the market;
that charge has been leveled against every method of informed
trading.  In the short run it is sometimes true but markets respond by
upping the general requirements for quality trading and many market
participants rise to meet the new standard or else switch to longer
time horizons.

On the critical side there is lots of talk of "unfairness" and "manipulation," combined with snide references to the financial crisis.  I'd like to see a serious efficiency argument against high-frequency outlined and defended, without the polemics.  That would include a case that regulation will prove workable and catch only the "bad liquidity," while at the same time avoiding capture by envious and inferior competitors. 

If high-frequency trading is used to trick other traders into revealing their demand schedules, and then canceling orders, I can see a case for regulating that particular practice.  On that issue, here is background, from a critic, but note that these charges seem to be unverified.

The philosophical question is why it might possibly be beneficial to have market prices adjust within five seconds rather than within fifteen.  One second rather than five?  0.25 rather than one?  If you had been writing in the year 1800, what comparisons would you have chosen? 

Remember that old comic book where they had Superman race against The Flash?  The Flash won.  Someone had to, just keep that in mind.


i like these posts where you give us a clear, well-laid-out position on an economic issue.

i think that's what he was addressing with this paragraph:

"The philosophical question is why it might possibly be beneficial to have market prices adjust within five seconds rather than within fifteen. One second rather than five? 0.25 rather than one? If you had been writing in the year 1800, what comparisons would you have chosen?"

The argument against HFT is essentially the argument against insider trading and front running. Both of those arguably increase efficiency by quickly moving price, but hurt the market because of unfairness.

I see this as being trader vs. trader with no real impact on the average shareholder or the company. I would hope that traders could work things out among themselves without govt regulation.

I see no reason to scare us with another "crisis". This crisis became a global event once millions of Americans found themselves in a worthless home, unemployed and with their savings wiped out. HFT trading won't do that.

It benefits some traders at the expense of other traders. So what? We don't have to speculate. We can follow the teaching of Warren Buffett who only uses his computer to play bridge with Bill Gates.

High frequency trading is certainly not just US stock trading. Equities trading, however, is heavily regulated already--see in particular Regulation NMS. I do not think these regulations have the intended effect of protecting customers (I also do not believe the net effect of "high-frequency trading" is to harm customers). Dealing with these regulations is a large part of the job of a high-frequency equities trader.

Other very good firms are making hundreds of millions of dollars, but somehow Goldman is making 10's of billions. I don't see how the numbers add up. I know from friends at other feared HF firms that they are making tons of money, but not billions. These are firms that other people in the HF space fear and are also a significant percentage of stock and futures trading in the US. They are not making 10's of billions or even a billion. I can assure you their technology and well, everything about their firms is state of the art. We are talking NASA level quants and millions in technology for these other firms.

Goldman's HF operation was an offshoot of the Blair Hull trading firm purchase. They paid 900 million, I think. The word on the street 6 months later was that they overpaid. I don't see how they make a few billion unless they are frontrunning an order book.

Tyler rather misses the point. There's nothing wrong with a technology that lets the market incorporate information faster. But we're really talking about an institutional rule that allows the HFTrader to see other people's orders before they buy the stock, and make trades on that basis:


The NADQAQ lets HFTraders see other people's orders before they're made.

So using the poker analogy, we're not talking about a new technology that allows players to calculate probabilities based on the visible cards; we're talking about a rule that lets some favored players look at the hands of other players.

"But hey," some people say, "it's not about fairness between traders. It's about efficiency."

Those people got it all wrong. Efficiency depends on fairness between traders. Markets converge to a collectively true probability estimate because anybody willing to bet money on their belief can participate. But if you give some players an institutional advantage, then other players realize a lower return from participation, and your market then reflects a narrower slice of beliefs. This is especially true if that narrow group of advantaged traders all reside and work in close proximity to one another, and face similar institutional and social incentives to hold similar beliefs.

"The philosophical question is why it might possibly be beneficial to have market prices adjust within five seconds rather than within fifteen. One second rather than five? 0.25 rather than one? "

The social benefit is the net present value of having "accurate" prices a few fractions of a second earlier. This benefit is a function of the discount rate, which is customarily expressed in annual terms, and the value of the changes in decision-making that might result under the new, more "accurate" prices.

I submit that the social benefit is trivial.

The private benefit is in the billions, but that is almost 100% rent-seeking. It is gained almost entirely at the expense of another party who's computer or trading algorithm might be fractionally slower.

This is an nice example of almost pure rent-dissipation. It's a case where the invisible hand of the market does more harm than good, by directing enormous capital resources and some of our most brilliant minds toward an activity that creates essentially zero value.

We should set up the rules of the game to maximize incentives for value creation and minimize incentives for rent dissipation. That suggests banning or heavily taxing HFT.

"Hey, Warren Buffett has "people" to use computers to do stuff other than play chess with Mr. Bill. A big FAIL on that analogy."

So, you think Warren Buffett became the richest and best investor in the world by trading strategies? No. You are wrong. A big FAIL on any credibility you have for anything else you say. That's fine, most people don't understand it. He makes money buying ongoing businesses. He's made money on finance business probably none of which use trading to make their major profits. Maybe Goldman Sachs, but I doubt it.

Sounds like high tech "front running" to me. I think the SEC has taken strike three: Madoff, naked shorting and now legalized front running. Time to terminate the SEC with extreme prejudice.

most high frequency traders are not attempting to move the market towards a true price. There are definitely some algorithms that do stuff that, but most of them are just attempting to trade against order flow and earn the bid/ask. Prior to decimilization in 2000 it made sense to do this manually, but now that you are fighting for a penny rather than an eighth, you have to be a little more sophisticated, and you have to trade a lot more and take on more risk to generate the same revenue. What we saw after decimilization was a huge decrease in the profits of equity market makers. Now, after 8 years they have figured out new ways to make that money and are doing well. In order to do well again they have had to trade a lot more so now we see much more volume and liquidity in US equities than even a couple years ago. The net result for institutional and retail investors is substantially lower net transaction costs.

The discussion on HFT seems to assume that you need enormous resources to do this stuff or compete in this space. That is just untrue. There are tons of small companies doing this kind of trading. The technology component (having co-located servers and the best datafeeds) costs around 5-50K/month depending on the exchange. Not exactly a domain that is reserved for only the biggest players. If you aren't willing to pay for that level of access, then you are presumably making the judgement call that it is not worth it.

I do HFT for a living (in futures, not equities, though I traded equities for years) and I have very little knowledge of what other successful high frequency shops are doing. I am always trying to figure it out, but no one is going around explaining the details of these algorithms. So, it is a little ridiculous for people who have much less involvement in the markets to think that they have this stuff all figured out and they know what the consequences would be to banning this activity.

Putting ethics aside, HFT will be the next financial tower of Babel to fall in grand fashion. See portfolio insurance of 1987 and the smartest guys in the room at LTCM. To paraphrase Charlie Munger, having a bunch of smart finance guys compete with each other isn't quite the same as having a bunch of restaurant guys compete with each other.

Paul Wilmott weighs in: http://www.nytimes.com/2009/07/29/opinion/29wilmott.html

Andrew, your implication was that Mr. Buffett eschews the use of computers except to play a particular computer game, and that QED it doesn't take computers to garner a lot of wealth. I would hazard a guess that Berkshire Hathaway has a lot of staff and silicon doing the fundamentals analyses. I was not in any way saying that Buffett and B-H are involved in program trading, just highlighting what looks like apples and oranges to me in your exemplar.

Wait, why are you so dismissive of the "talk of "unfairness" and "manipulation?"

That's precisely the issue! It is an unfair advantage!

You reference telephones, but telephones quickly became cheap enough for everyone to use and so it very quickly afforded NO unfair advantage to the very wealthy.

However, I really do not think these crazy machines are going to be accessible to the average trader any time soon. So until it DOES become accessible, why not ban it? And when, like telephones, it to becomes common, you can unban it.

And, since you yourself are wondering what actual social benefit having these supercomputers trade brings, it seems that banning this practice would have no net costs to society, while all the time ensuring an equitable and just trading floor.

How am i wrong?

I'm a daytrader, and with Keith I don't see any problem with computerized trading except in two cases.

First, if the boxes (or their operators) screw up and their owners get crushed, I don't want to see those trades broken, but I see these trades broken every time. If the computer crushed you, that's your loss, just as if I have a keystroke error and give up $2/share on a big spread I have to take that loss.

Second, quote availability should be uniform for everyone. It's fine if your computer/connection is faster and therefore you are able to access that quote more quickly, but if you get privileged access to flash orders this is a problem.

Other than that, keep up the good work, HF traders!

A reader sent me this very excellent comment, for some reason he could not post it himself:

"Is it possible to stop talking about front-running in a context of "flash" orders? It's an _additional_ order type available to the party entering order, so by deciding to use flash order the submitting party _prefers_ it to the traditional one. What is the problem with optional functionality which doesn't remove an ability to send regular limit orders?

First it was GS reported profit which seemed unjustified to some people, then someone noticed Program Trading numbers on NYSE and decided that GS made profit only on that exchange (although even firms with, say 20 people, trade globally), then where are "program" there is "speed" which is HFT and where there is a speed there are front-runners :)

My company is a software vendor in algo trading business and I'm looking at all this stuff around HFT as on typical reaction of people who need someone to blame, while these people don't understand how the world is actually work.

First, people confuse Program Trading (PT), Algorithmic Trading (AT), High-Frequency Trading (HFT) and Front-running (FR).

And really, HFT/Algo systems is not THAT expensive, so they are
available for "elite" participants only. The prices quotes above are correct for off-the-shelf systems.

I'm not very familiar with the US regulatory definition of PT, but I always thought that the Wiki explanation is correct:

"Program trading is casually defined as the use of computers in stock markets to engage in arbitrage and portfolio insurance strategies. However, the New York Stock Exchange (NYSE) defines the term as "a wide range of portfolio trading strategies involving the purchase or sale of 15 or more stocks having a total market value of $1 million or more" without any direct reference to the use of computers[1]."

So, let's say I want to buy a basket - this will be regarded as PT from this POV and included into widely quoted GS numbers. And imagine arbitrage between S&P basket and index fund like SPDR. But even forgetting about ETF arbitrage - almost any basket purchase/sell will be considered as PT.

Algorithmic Trading - this is trading involving computer system and it's really a lot of different stuff, for example:
* Order execution. People typically don't send huge orders to the market to avoid moving it (which leads to huge transaction cost). Instead they use strategies to execute these orders, be it VWAP/TWAP or something more complex, especially involving buying stocks traded on multiple venues. Noone really manages such orders manually.
* Arbitrage (stocks, options, ADRs, indexes, baskets etc)
* Hedging - similar to order execution. Hedging with huge orders is quite expensive...
* Market making in broad definition (not only option MM, but, say, MM in ADRs or original stock). This overlaps with arbitrage in some sense, since arbitragers act as market maker for the instrument they quote.
* Automatizing day-to-day operations. For example, you might get an FX exposure which you don't understand and want to get rid of it (similar to hedging).

HFT - this is a part of Algo Trading, since not all parties using algo trading systems care so much about _speed_. Especially if they trade on slow markets, if they automate their operations, like FX hedges etc. However, most algo trading vendors to promote speed as a key feature of their products, although few of them really say what speed do they have.

Front-running. I would say this is a small part of trading (I don't remember anyone doing this, but maybe they just don't talk about this). It's something with unknown social value (I personally doubt that there is any) and is always around anticipating future order flow and taking advantage of it. Front-running doesn't have to be HFT - it can be done even manually if someone enters, say, huge buy order which is visible to everyone.

However, not all order flow anticipation and not all algos doing "pinging" the market are bad things:
* Market makers do need to anticipate order flow so they can balance their quotes to not take large speculative positions against more informed participants.
* Order execution strategies, run for, say, institutional investors like pension funds may need to "ping" markets in order to discover hidden liquidity in them. The problem with the hidden liquidity is that a lot of parties do want to trade, but they don't want to reveal information to all participants to avoid front-running of their orders, that's why a lot of dark pools appeared and different venues like Turquoise have added "dark" functionality to their systems."

I see the comparison to insider trading and remember that not everybody is against insider trading ( but most people are).

It isn't about making a lot of trades, it's about Goldman Sachs and others seeing Bids/Asks before everybody else and making can't lose arbitrage trades that "skim" the entire market. It IS front running, it IS insider trading, and they ARE making billions of risk free dollars off of other investors.

Tyler Durden on Seeking Alpha seems to be leading the war cry against GS and the other HF traders while CNBC takes pot shots at him and "bloggers" on Seeking Alpha. Jul 29 01:58 PM |Report abuse| Link | Reply +10

Many, many firms have servers in the optimal co-location facilities. Only a few of those firms make anything like a billion at high frequency trading.

I take back the statement that Getco makes "most" of their money prowling the markets. That was emotion getting in the way of logic. But its pretty well known that they bribe exchanges to receive preferential access and execution. Thats crossing the line.

I think you should talk to a control theorist about this. I will at least.

We can view the markets as a control system which attempts to equilibriate price to value. In those sorts of systems there is always a trade off. If you get speed of reaction, you might lose bandwidth, or get steady state errors. The financial markets are far more complicated than the PID systems I am used to and have the disadvantage of imperfect information - if I am running an experiment I know what temperature my device should be; I don't know how much a stock should really be worth, and even if I did there a lot of idiots out there doing things like... investing in Lucent in the 90's.

I thought the main problem that the "flashes" of information where going to the big players first so their processors could react slightly ahead of anyone else (with a super fast computer.

Tyler, in the event that, during shocks to the system, that high frequency liquidity providers stop providing liquidity, there is a profit opportunity for the right trading strategy with risk capital to capitalize.

And for those of you who haven't checked it out yet, read Eric Falkenstein:


Michael Wellman: I read your blog posts on the call market idea. I'm familiar with the mechanism, as I utiltize market-on-open and market-on-close orders all the time in my business. But how did you arrive at 1-second intervals? Why not 100 milliseconds, or 10 minutes?

From an engineer's point of view.. Increasing frequency in a complex system can spark unintended changes in the system's behavior.

Man, for such a lucid article, your commenters (excepting the HFT dude) are all drinking mass media bongwater. No, HFT isn't going to be the next "too big to fail" -HFT is *small.* HFT dudes do not hold large positions, by definition. Duh.
And no, it's not "front running." Front running involves self-dealing by a broker attempting to rip off his clients. Changing the price to reflect many buy or sell orders: that's what a market is supposed to do. That's how the market encodes information: price. If someone's selling a lot, that means they probably know something and the price should go down. If it's just someone dumping a block to buy a new set of titanium golf clubs, the price will only move a little and mean revert the way it's supposed to.

Investment banks, hedge funds and institutional investors apply high frequency trading as a means of automating trades via the capabilities of high speed computers, normally to enable transact a huge amount of orders at record speeds.Los Angeles Web Design

Comments for this post are closed