Comment by tikhonj | Hacker News Reader

Comment by tikhonj | Hacker News Reader

A few immediate thoughts:

Transaction volume is ultimately not the important metric--revenue is. And, following [1], it seems that the total revenue for HFT was probably around $2Billion in 2013--for a whole industry, that's not very much! Measuring transaction volume is akin to comparing shipping between Amazon and Walmart ignoring the fact that Amazon ships directly to consumers while Walmart mostly ships to large Walmart stores.

"...increasing liquidity is the last refuge of bullshitters" is not an argument--it's an assertion. That was not really supported. Liquidity is a good thing; the article claims that HFT does not help much because most of the actual benefits happened before its advance. Of course, considering how limited HFT revenue is compared to other forms of trading, it's likely that the benefits are just smaller in proportion.

So I don't see, from the article, that HFT is necessarily socially useless. Rather, I see that it is likely useful in a moderately small way spread out over a lot of people (most people in the markets). The benefit is not obvious or concrete, but that doesn't mean it doesn't exist.

Similarly, the article complains about how bots just quote each other prices without necessarily making a trade. I don't see how this is a bad thing. All it means is that their quotes are at a much higher resolution than manual quotes, that's all. This seems like it would generally be a good thing.

Now, I'm not saying that HFT is not without its own risk or issues--they're just not the issues brought up in the article. Or, in fact, in most popular articles: popular reporters want to turn HFT into a moral issue and paint HFT firms as evil manipulators, when they really aren't. The actual risks of HFT are more structural and technical, which, I suppose, is not great for a broad audience or lots of pageviews!

It's also not immediately clear that HFT should be banned or how to deal with it. Many proposals I've heard would reduce liquidity beyond affecting just HFT, raising real costs for consumers. Ultimately, this is why there has not been much regulation in the space!

[1]: http://247wallst.com/investing/2013/03/24/high-frequency-tra...

See, here's the thing I don't understand about liquidity: If it's so valuable for trades to execute in microseconds instead of seconds, and the stock exchanges recognize this value and provide co-location etc to enable it, why are so many stock exchanges closed for half to two thirds of the day? [1]

Surely the 15+ hour shut downs are a much bigger limit to liquidity than a few microseconds here and there?

There's obviously no technical reason - I don't see Amazon or Google closing down their websites from 4pm to 9:30am. And if it's about the release of news, that only really needs a window of an hour or so.

[1] https://en.wikipedia.org/wiki/List_of_stock_exchange_opening...

I worked in the industry for a little while.

At this point, some companies depend on having that daily downtime. Their whole development is based around the fact that they will have guaranteed downtime. It's built right into their software stack.

Trying to fiddle with this expected downtime would throw (parts of) the industry into turmoil.

It's just a historical quirk, but it's probably here to stay.

Surely there are ways to get rid of it without the transition being so traumatic. For example, the change could be announced a few years prior, and the downtime could go down by an hour per year.

To put things into perspective, the company I worked for responded to bug reports by hiring a team of people whose role was to manually alter the database any time a customer reported a data problem. Dozens of times per day. In other words, instead of fixing the software, it somehow became a reasonable idea to dedicate employees to fixing the symptoms by hand. It made sense in hindsight: they were making so much money that they took the path of least resistance. Paying employees to fix the problem immediately was quicker than trying to hunt down a competent programmer to try to fix the problem without causing more problems.

The change would be traumatic no matter how long they are given to plan for it.

Oh, here's another reason I forgot to mention: You can capture huge profits by exploiting the opening and closing few seconds of the market. A huge portion of all daily trades happen within the first few and last few seconds of the day. So there's a financial incentive to leave things as they are.

or quote stocks in a separate "non stop" market

Going to a continuous cycle will end techniques like "banging the close". That would be disruptive.

>the downtime could go down by an hour per year.

Just my opinion but I think that would just increase the pain and complexity of the transition.

From a standpoint of market efficiency, I believe the weekend is a similar "test". Although, i agree that their are diminishing returns to a-synchronous opening hours.

It's true! A certain company has a trading platform that, when run, will wait until the markets open, do some setup stuff around the open, trade for one day, then do nothing forever. Some script comes by to kill -9 everything later in the night so it can be born anew.

All the people who work in finance I've spoken to have imputed to me that the industry is on the cutting edge, that they will and are able to go to any lengths to execute trades faster, and that they earn their bumper salaries by being the most talented technologists out there. They tell me stories of FPGAs and how they certainly couldn't use garbage collected languages and about people cutting holes in walls to shave valuable feet off cables.

Now, I'm sure that in the absence of any need to they haven't done the work to achieve 24-hour operation, but I didn't get the impression it would be beyond the abilities of the entire industry.

Do you think the people I've talked to misrepresented the industry, and it's not as advanced and competent as they made out?

You don't understand. Not everything is written in C. There's a backend system for reporting OATS that starts at 5 PM. If you need to make a non-trivial change you better not deploy it during market hours. Most of these systems are written in Java by guys who aren't at the firm anymore and didn't care when they were. You say cutting edge (which I would strenuously debate). But it's not magic -- in fact it's actually quite mundane when you get down to the nuts and bolts.

That's definitely not representative. Perhaps you've spoken only to guys doing HFT. A lot of trading systems are written in Java, and something like 20ms between seeing a trigger and sending an order to the market is fast enough for most needs.

I think for 24/24 operation though, the problem isn't really a programming one but simply the sheer number of people who'd have to work in shifts. Traders, engineers, middle office, maybe even compliance or quants... all these people are really expensive. You can't just leave a system running without people to monitor risk, check for reporting breaks, approve transactions, etc. The end of day reconciliations and reports are also much easier done with the exchange down.

I found a similar question on Quora [1], with an answer by an ex-quant. It seems to boil down to tradition and bureaucratic inefficiencies by the involved parties. So I agree, this makes the whole liquidity argument seem weak.

[1] http://www.quora.com/In-the-24-hour-world-why-do-the-New-Yor...

the stock exchanges recognize this value and provide co-location etc to enable it

Not true. The exchanges charge hefty fees to colo in their datacentre. What you do with it is completely up to you. It's just more revenue as far as the exchange is concerned.

why are so many stock exchanges closed for half to two thirds of the day

In practice, this doesn't matter. When NYC closes, trading moves to Tokyo, then onto London, then back to NYC. Anything you want to trade, you can do so 24 hrs a day if you really want to.

You say you can trade around the clock by trading around the globe. But a position in New York can't exactly be liquidated in Tokyo. You can hedge for an approximation, but then you have cost of carry. Most markets are also completely closed on Sunday. It's not obvious to your parent why this should be. And don't even mention the half-day for stocks in the US around Thanksgiving, which is just silly.

Important things are 24 hours a day. Spot and forward FX, commodities, sovereign bond futures, Equity index futures, etc. See for example:

http://www.cmegroup.com/trading_hours/

CME tends to be closed for an hour a day for cleanup and that's it. US equities aren't a big deal compared to the size of bond or FX markets.

Liquidity during trading hours is important because an open market with little liquidity makes it difficult to to sell and buy assets in significant quantities. Even reasonably sized transaction start having massive effects on price. Short-term price stability gets lost and what you ultimately have is a significant increase in transactions costs.

Ultimately, liquidity during trading hours is a different question from what hours the exchange is open for trading in the first place.

Trades executing in microseconds is about beating the other HFT. It's useless for consumers.

Price competition (one HFT bidding 10.01 instead of 10.00) and depth of book (the ability to buy 10,000 shares in one shot) is what helps traders.

Multiple reasons... History being one. They clear the books after market too, it's a big part of catching the crooks and frauds. Almost certainly makes auditing easier.

Is executing trades quickly bad? Or is flipping an equity quickly bad? I cam see no good that comes from buying and selling in milliseconds; the tax should inversely exponential to the hold time or something.

Can't agree more (focusing on microsec vs nights/weekends). It's not an excuse to say they need to train/test their algos, they have entirely separate machines for that. For any other tech company live rollouts are a fact of life. I'd suspect the reason is historical and based on a single person's practical workday, if you extend it to 24/7 then naturally you'd have to have several shifts. Same reason why they are closed on holidays, to give people a break. So I'd suspect it's to benefit the people involved and not the machines which I'm certain can be adapted.

Excellent point! Being able to trade on information revealed overnight would have much greater value than being able to trade at an even smaller sub-second interval.

Since the marginal value of additional liquidity decreases rapidly, I've often wondered if having a fixed resolution (say, one trade per minute) would actually be beneficial.

Nobody can realistically trade a stock based on sub-minute changes in information anyway, and having a fixed resolution would eliminate the advantage some players have by having more servers/etc. After all, if we are chasing liquidity, allowing some players in the market to have an advantage restricts the number of players able to participate which lowers liquidity.

> popular reporters want to turn HFT into a moral issue and paint HFT firms as evil manipulators, when they really aren't.

First, almost every issue is a moral issue, especially one dealing with the value of a certain endeavor (isn't that what ethics is about? Trying to find the value of things?).

Second, claiming that HFTs aren't evil is as much of an assertion as calling them bullshitters. Most "popular reporters" as you call them (I assume pejoratively) at least support their claim. They say that HFT has little social value, and then claim that putting so much effort into something of little social value is at least morally questionable.

It is claiming that this is not a moral issue that is the more powerful moral assertion here, and quite suspect, at that. Whatever economic risks HFT may entail, its mere existence is first and foremost a problem of ethics.

Much of HFT does have little social value. We could eliminate much of the worst of it by allowing subpenny trading, so trading firms could compete on price instead of on latency.

Maybe you think "sub-pennies? that's just a different kind of insanity." But remember that we're talking per share pricing. So imagine every transaction of every share ever being wrong by an average of half a penny. Imagine you could compete for the money represented by that error, and you could win it, just by having the fastest computers which put in the orders first. Behold: Wall Street as you know it.

I agree that we could and should treat it as a moral issue, and that avoiding those questions is often a sign that something immoral is going on.

But we can also look at it as a system design question: if we're trying to build an efficient, robust marketplace, what activity do we permit and forbid? What do we encourage and discourage?

Having worked for market-makers, I get the value of liquidity. It's not at all clear that HFT firms actually provide liquidity [1], but even if they did, we'd want to ask, "What is the cost of different sorts of liquidity provided, and which ones do we choose to maximize the value of the market to participants and society as a whole?" So far I haven't seen any evidence that HFT activity isn't purely parasitic. In which case it's reasonable to ask whether we should still reward it.

[1] The most profitable ones apparently remove liquidity from the market: http://faculty.chicagobooth.edu/john.cochrane/teaching/35150...

> "...increasing liquidity is the last refuge of bullshitters" is not an argument--it's an assertion.

I believe it's technically an observation, a claim that in the writer's experience, bullshitters fall back on that argument. It's true that he didn't explicitly give evidence for that, but expecting writers to justify every single statement in a short piece that is one of many they write on a topic is another refuge of bullshitters. He's right, though. Bullshitters use that claim because it's a vague, hard-to-verify positive claim that can be made about almost any market activity.

He does support the implied assertion that the claim is bullshit in this case. The only reason we care about liquidity is that you want people the market serves to be able to execute productive trades more quickly and cheaply. If it hasn't gotten cheaper, that's good evidence that HFT trading is not socially useful.

You interestingly also provide evidence that the claim is bullshit. In the article you link, it mentions that the most profitable HFTs aren't liquidity-generating; they are liquidity-taking. That is, they aren't coming into the market with open orders that sit their waiting for other people to take them. They are coming in with orders that match existing offers, removing liquidity from the market. That's from an academic study linked in your article: http://faculty.chicagobooth.edu/john.cochrane/teaching/35150...

"The only reason we care about liquidity is that you want people the market serves to be able to execute productive trades more quickly and cheaply. If it hasn't gotten cheaper, that's good evidence that HFT trading is not socially useful."

That serves as some evidence that it is not socially useful. It completely ignores "more quickly", and for the strongest argument you should also show that nothing else has happened in that time period that would have increased costs and slowed transactions but for HFT.

"the most profitable HFTs aren't liquidity-generating; they are liquidity-taking"

Sure, the easiest way to be among the most profitable HFTs is to cheat, and the way you act on early information is active orders. That doesn't mean they are the most prevalent - I was not able to find actual statistics on the number of each class (aggressive, mixed, passive) represented in the market, but the sample count for trades of aggressive firms was less half of that for the other two.

Edited to add: Actually at a slightly closer glance, it looks like they were looking only at firms that didn't lose money, which seems pretty worthless. "Active strategies have more divergent outcomes" seems a better explanation of the data then "active strategies make more money" - particularly if the reason for the lower sample count is more "active" firms lost money.

None if this is to say that I think there's nothing worth fixing in our financial system, I just want to be sure we're using the evidence we have appropriately.

On the contrary, the idea that HFT increases liquidity is pretty much obvious if you think about it. It's the assertion of the opposite that needs extraordinary proof, and since the author doesn't provide anything at all, he sounds like a bullshitter to me.

You are seriously suggesting that a professor of finance who specializes in studying high-frequency trading is just bullshitting? In a 60-page academic paper on his area of professional expertise? And your view is based on nothing other than the causal intuition of an anonymous commenter called "tutufan"? Gosh, color me convinced.

Yes, market participants on average increase liquidity. Which is why you have that intuition. But it isn't specifically true in all cases. For example, take a simple commodities market where buyers and sellers show up in person to trade wheat. Farmers show up to sell; flour-makers show up to buy. With me so far?

If I place people on the main roads into town and have them buy up all the grain before it reaches the market, I will be reducing market liquidity, because anybody who needs wheat will be totally fucked unless I decide sell to them.

It's not about speed for the sake of speed. That speed allows insider trading and frontrunning.

Insider trading means[0]:

  buying or selling a security, in breach of a fiduciary duty or other
  relationship of trust and confidence, while in possession of material,
  nonpublic information about the security.

How do HFT traders get "material nonpublic information"?

  The Wall Street Journal reports that HFT funds buy early access to data
  from third-party distributors—everything from corporate earnings to
  the Philadelphia Fed's manufacturing survey.

If an analyst at the Philly Fed tells me the results of the manufacturing survey two days in advance of its release, and if I profit from that information and give a kickback to the analyst, that would be clearly illegal.

But if the Philly Fed gives the information to Reuters ten minutes early so they can write a story, and if Reuters sells electronic access to HFT traders two seconds before the public can trade on it, how is that different?

And don't get me started about using HFT for frontrunning client orders[1, 2].

[0] https://www.sec.gov/answers/insider.htm

[1] http://blogs.barrons.com/stockstowatchtoday/2013/05/03/charl...

[2] http://www.nanex.net/aqck2/4442.html

Early access to research reports is not insider information and is not in general illegal.

First, presumably the report was generated from public information so it is not non-public. Second who are the parties involved in a breach of "fiduciary duty or other relationship of trust or confidence"?

But they're profiting off making the 2 second early access to data more liquid, at best.

Reply on news.ycombinator.com