Hacker Newsnew | comments | ask | jobs | submit | erichocean's commentslogin

> Direct feeds are available to anyone who wishes to purchase them.

Out of curiosity, how much does a direct feed cost, and is there a practical limit to the number of simultaneous buyers (for example, physical server space in a data center)?

reply

lrm242 13 days ago | link

Price lists are all public. Look at nasdaqtrader.com or batstrading.com.

reply


I get all that, but does that really apply when the time interval is sub-second? It's hard to imagine the discounted value of the trade changing in a sub-second interval, whereas it obviously does selling 11 years later.

In other words, sure, definitionally it's liquidity, but since the trade would have gone through a sub-second later anyway, it's not meaningful liquidity: the liquidity was already there on a reasonable timespan (again, sub-second) for the seller. Thus, nothing has been gained.

What am I missing here? Do sellers really need additional sub-second liquidity beyond what the market would provide without HFT?

reply

harryh 13 days ago | link

You're missing the fact that the trade would not necessarily have gone through a sub-second later anyways. There is a very real chance that the trade would never happen because the market moves away from you.

reply

aet 13 days ago | link

I think it is just the nature of the market mechanism i.e. the rules of the game. Since markets are continuous there is going to be competition. Tick sizes means that participants have to compete on speed rather than price. It seems to me that competition and continuous markets are maybe the best way to keep prices aligned across markets. And, just maybe, the best way to keep costs low for retail traders.

reply

hft_throwaway 13 days ago | link

If you think you don't benefit from someone else providing a quote, place a limit order of your own and update it throughout the day. Buying the market-maker's quote is basically giving them your trading problem. It's specialization like anything else. You get to do what you're best at. He gets to figure out the cheapest way to exit/hedge the position.

Also very few HFTs have sub-second average holding periods. There aren't many instances where the market trades both sides so quickly. Probably average in the 10s of seconds at least if not longer.

reply


> there are also many who genuinely believe that they are doing a service for the market

You make it sounds like HFT is some sort of non-profit, community service vocation, like teaching, or fire fighters. Do people actually think this?

Regardless of the merits of HFT, that kind of argument seems very disingenuous. I haven't met a single person engaged in HFT who is doing it to smooth the function of the market or "provide liquidity". They may think it has that effect, but no one is in it for that reason. If anything, it's what they tell themselves to justify skimming money from the other market participants.

reply

tptacek 13 days ago | link

You can replace the letters "HFT" in your comment with "any market maker" and come to exactly the same conclusion.

Liquidity has always, always been subsidized by "skimming". The rake market makers take from trades is the cost of liquidity: ie, of being able to buy shares immediately without getting shafted because the only orders on the market are from large traders who are holding out for very high prices.

Have you ever tried to sell a house? That's what crappy liquidity looks like: there are buyers and there are sellers, but nobody is anywhere near agreeing about the price of the goods being traded, and so it takes months to get in or out of a position, and you sometimes have to do it at extremely unfavorable prices.

reply

declan 13 days ago | link

>You can replace the letters "HFT" in your comment with "any market maker" and come to exactly the same conclusion.

You may even be able to replace the letters HTF with "any for-profit profession" and come to exactly the same conclusion.

I think it's fair to assume the HFT traders are rapacious capitalists (as someone who is pro-free market, this doesn't exactly offend me). But motivations don't matter here. It reminds me of the Adam Smith quote, which may only make sense if you've read the linked Scott Locklin essay first:

http://www.econlib.org/library/Enc/bios/Smith.html "It is not from the benevolence of the butcher, the brewer, or the baker, that we can expect our dinner, but from their regard to their own interest..."

reply

danielnaab 13 days ago | link

The problem is that market making is different than the critique, which is that having privileged access to exchange data centers is an unfair advantage that allows rent-seeking. To access information before the general public allows traders to inject themselves into transactions that don't require a market maker.

That fact that no one will defend this practice while vehemently defending "HFT" speaks to an impedance mismatch. We're not talking about the same thing.

reply

harryh 13 days ago | link

They have faster access to exchange data centers because they paid for it. Anyone can pay for this access. They use this privileged access to provide faster and cheaper service to their customers (purchasers of liquidity).

It's no different than McDonalds paying a lot of money to buy a prime corner location to build one of their restaurants.

reply

danielnaab 13 days ago | link

You can pay for a lot of things, but that doesn't mean it's in the interest of society or the market as a whole. You don't require faster access to provide liquidity, but it sure helps with arbitraging trading activity.

reply

tptacek 13 days ago | link

You can pay for a lot of things and that doesn't mean it's not in the interest of society or the market as a whole either. The trading in any one instrument on an exchange is a zero-sum game, but the structure of the market and the relationship between exchanges, buy-side firms, sell-side firms, and retail investors doesn't have to be.

Before electronic trading, the specialists who executed orders on the NYSE floor ALSO paid shitloads of money for their access.

One difference between today and the heyday of the specialist system is that in the specialist system, middlemen skimmed overtly from investors: the bid-ask spread was gigantic, and all that money went to the middlemen.

reply

gnopgnip 13 days ago | link

The increased liquidity/ reduced execution risk changes it from a zero sum game.

reply

hft_throwaway 13 days ago | link

You could not be more wrong. Providing liquidity at a razor thin spread is one of the toughest trades out there. Think about it: every passive order I place gives someone the option to trade with me until I can update it. If my price is wrong and I don't update, I will trade every time. If my price is right, I'll trade some of the time if I'm lucky. If I can tilt those odds in my favor by pricing more intelligently/quickly, my trades with uninformed traders slightly subsidize my losses to informed traders.

Someone taking out/arbing a market needs a source of alpha. I need to protect myself from every source of alpha, at least on average. If market makers weren't collocated and looking at many exchange/products for data to price their markets, they would be driven out by fast liquidity takers, forced to quote lower size or a wider spread until they simply stopped getting trades (other faster, smarter market makers could still quote tighter spreads and would take all your market share).

reply

danielnaab 9 days ago | link

The question is: do you require pricing data in advance of those you are trading with to earn your spread?

If you do, you are not providing a service to the market.

reply

harryh 13 days ago | link

What's wrong with arbitraging trading activity?

If MSFT is trading for $50.00 on one exchange and $51.00 on another we want someone to fix that price discrepancy! Even if the difference is much smaller we want that balance to be corrected!

reply

danielnaab 13 days ago | link

This seems like a back and forth that's quibbling about words. I didn't say there's anything wrong with arbitrage; the critique is the privileged access to market information and the ability to profit off it without adding any value to the rest of the market.

All things equal, I would prefer to trade on an exchange that makes an effort to provide equal access, and one that doesn't artificially inject middlemen into transactions that don't require them.

reply

harryh 13 days ago | link

The way that you think middlemen inject themselves into transactions that don't require them isn't what happens in the actual world. You're worried about a problem that doesn't exist.

reply

danielnaab 13 days ago | link

Well, Googling for research around latency arbitrage lends me to believe the opposite.

reply

hft_throwaway 13 days ago | link

WTF is "latency arbitrage"? All arbitrages are short-lived and latency sensitive. It's a marketing term made up by fear mongers.

reply

001sky 13 days ago | link

Along these lines, a note on so-called latency arbitrage:

http://web.eecs.umich.edu/srg/wp-content/uploads/2013/02/ec3...

reply

001sky 13 days ago | link

You're worried about a problem that doesn't exist.

Even if empirically it is non-observed, its illegal. So it is a problem, just one solved by law. Now, IFF your're willing to assume that nobody has ever/will ever transgressed such a law...you could hold this position 'logically'. But Few would be so unwise "in the actual world" to do so (and would likely prove foolish).

reply

harryh 13 days ago | link

It's illegal if your broker does it. He has a contractual (and legal) obligation to act on your behalf. This is completely different from the kind of thing that Michael Lewis claims is happening in his book.

reply

001sky 13 days ago | link

you think middlemen inject themselves into transactions that don't require them

I'm responding more to this (and more as a general concept).

If given the opportunity, people would do this. So it's useful to dispel the notion that because its "not seen" it's not a problem. It might not always a be problem, but many it is and others it's also "annoying".

Stepping outside of markets for a second, this is why M&A deals have exclusivity and breakage terms. The theory/practive on why you do this involves assymetric information and something economists refer to as "opportunism". Opportunism[1] is exploiting legal abiguity for narrow self interest.

The more complex the legal/transaction structure the wider the attack surface for "opportunism" (because it's proportional with deal complexity). In slow and complex deals like M&A at a certain point it becomes an issue and so safeguards are taken.

In fast and simple deals like buying stock on the box or what not, the scope is very limited. The glaring example is the broker issue you highlight. But as the deal structure gets more complex, this risk rises.

Now, the attack surface for opportunism is proportional to deal complexity as mentioned above. This is legal/structure but also now needs to consider execution and # of counterparties. Limiting counter-parties (ie, middlemen) reduces this "attack surface" (if you will). Because each of them needs to be monitored/paperwork delath with etc. This is why such uncecessary interventions are structured out of deals. Typically by way of convention/courtesy if not expressly by things like ethics codes & stronger (ie, exchange regs+laws).

This is a long way of saying that the institutional integrity of markets hides a lot of problems. This is a good thing. But not to be taken for granted. The cost of undermining these institutions/trust etc. is not insignificant. Because it means you have to solve all these problems again in new ways.

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

reply

harryh 12 days ago | link

We've got a system specifically designed to prevent the kind of thin you're talking about.

It's also a system where every single trade is recorded in detail that anyone who wants can look at. No one is pointing at this data showing the smoking gun of the kind of front running you describe.

You're just saying "well, people are shady, so it's probably happening anyways." That's an incredibly weak argument in the face of a stack of evidence on the other side the biggest piece of which is the dramatically reduced cost of liquidity. If such shadiness was happening the cost of liquidity would be going in the other direction.

When talking about crimes people refer to means, motive, and opportunity. You've certainly got the motive part, but you're completely ignoring means and motive.

reply

001sky 12 days ago | link

Then I think we're just getting hung up on the labguage concerning what is 'necessary' or 'un-necessary'. I'd actually be interested to hear about the level of abstraction of your systemic data, just to be more clear on where you are coming from. Again, my comment up-thread was not in any way impugning a particular trade or trade structure. If you are a practitioner, you're well aware of the level of detail required to opine on something so specific. As a general rule though the more complex any transaction is (whiteboard+lawyers) the more legal grey there are. That is a statement I'm comfortable with in general.

reply

msandford 13 days ago | link

OK but if the HFT traders only hold stocks for say a few seconds (or milliseconds) then that means that there is plenty of real liquidity to be had and thus HFT isn't adding value. It's adding volume, but not value.

If the HFT guys are willing to hold stocks for minutes or hours or days then yeah I might concede that they're adding liquidity. From what I understand that's not HFT anymore. But for a second? I can wait an extra second for an order to get filled.

reply

harryh 13 days ago | link

You can wait an extra second and maybe get it filled.

Or maybe the price will move away from you and you'll never get filled (at least at the price you want).

reply

patrickg_zill 13 days ago | link

One of the quotes either from the book, or one of the TV interviews, was "if it is illegal if a human does it, why is it not illegal if a machine does it"?

I think it is from an interview on The Daily Show, see here: http://www.zerohedge.com/news/2014-04-02/jon-stewart-hft-its...

The quote seems to be 'The HFTs "function on volume and volatility" alone and "they know the prices before you do... which is illegal if it's a person, but as a computer, meh?"'

So my question to you is, i) explain how HFT serves the same purpose as market-making ii) explain how HFT is not front-running the valid trades

reply

tptacek 13 days ago | link

That quote drives me nuts, because it is completely unclear on how anything that computers are doing would be illegal if humans did it.

For decades, humans did exactly what the HFTs did, except much more brazenly: where HFTs take pennies in compensation, the humans took dimes, quarters, even dollars.

That premium wasn't outlawed. No rule was passed to eliminate the profits the middlemen took. No, the premiums were competed down to where they are now.

(I didn't downvote you).

reply

ericd 13 days ago | link

Are the HFTs actually marketmaking, though? Lewis is accusing them of front-running transactions that were already possible, driving up the price slightly of transactions that were already going to happen based on telegraphed intent. This is different from what an actual market maker does, which is provide a counterparty for all transactions and hedge their resulting risk, in exchange for the spread.

EDIT: Perhaps this is just a small piece of what HFTs do, and maybe it's limited to a few bad actors, but being blown out of proportion by the book?

reply

kasey_junk 13 days ago | link

It is a small piece of what they do, and there are a few bad actors in HFT. That said, Lewis literally just gets it wrong on what is happening in his front running accusations. Either he doesn't understand what is actually happening or in an effort to simplify and/or make a compelling narrative he misrepresents the actual market movements that are happening.

reply

patrickg_zill 13 days ago | link

The HFT firms have statistically significant deviations in terms of profitable trading days.

For instance, Virtu had 1238 straight trading days of profitability, with no losses.

There are no other firms in the history of Wall Street with such success. Such an outlier deserves a lot of scrutiny.

It might help if you expanded on your views of why Lewis is wrong.

reply

kasey_junk 13 days ago | link

We have to be very careful about our terminology when we talk about profitability of trading days. I have not inspected Virtu's trades, pnl statements, or legal documents related to them, so this is less about them specifically than about trade pnl generally.

When most firms talk about profitability of trading day, they are talking about the the profitability of just their trades. That is, if they bought for 1 million dollars and sold for 1 million and 1 dollars. That is technically a profitable trading day. If the cost of entering and exiting that trade cost them 2 billion dollars they will still mark that as a "profitable" trading day, even though it clearly isn't.

That is to say, it is trivial under this accounting scheme to come up with a strategy that has no losses. Just don't trade.

I apologize for not going into more depth about why I think Lewis is wrong. I've explained it a lot in my comments on HN lately and can't be bothered to do it again. That said, if the linked article was more professionally presented and didn't have the weird racial language the points are basically what I would argue against Lewis' premise.

reply

jhart3333 13 days ago | link

I've looked through your comments on this post and did not find anything that looked like an explanation or refutation of Lewis' claims that HFT were front running. Maybe I missed it or you were talking about another post. Why don't you create a blog post? That way you can point people to it and you won't have keep repeating yourself.

Edited for clarity.

reply

harryh 12 days ago | link

https://gist.github.com/harryh/9996284

reply

hft_throwaway 13 days ago | link

Virtu basically only does market-making which is a low risk trade. You put on small positions with a tiny statistical edge hundreds of thousands of times a day to make the bid-ask spread. It's like flipping a coin that comes up heads 50.05% of the time and betting $1 on it a million times a day. If something breaks or loses too much you exit your risk and turn off for the day. If a stock doesn't work in the portfolio, you take it out. If strategies or markets stop making money, you turn them off. Even if they buy a company that goes bankrupt, their notional position is less than their daily profit firm-wide.

Firms that make markets basically make money trading every day absent technology issues. That doesn't mean the company itself is profitable every day. It costs a lot of money to build the technology and strategies, and for every Virtu out there, there's a dozen failed HFT startups.

reply

throwawaay 12 days ago | link

While it's true that increased competition among market makers have driven spreads down, decimalization was also a factor. For example:

  http://www.sec.gov/news/testimony/052401tslu.htm
"For example, OEA estimates that, from December 2000 to March 2001, quotation spreads in securities listed on the New York Stock Exchange ("NYSE") narrowed an average of 37%, and effective spreads narrowed 15%. An even more dramatic reduction in quotation spreads was observed in Nasdaq securities, with spreads narrowing an average of 50% following decimalization, and effective spreads narrowing almost as much."

Technically, that's an example of a rule that was passed to eliminate the middlemen's profits.

reply

pash 13 days ago | link

"It is not from the benevolence of the butcher, the brewer, or the baker that we expect our dinner, but from their regard to their own interest."

Adam Smith's insight helped launched the enterprise of modern economics because it turned out attention to what matters: results. Benevolence, like self-interest, is neither necessary nor sufficient to produce a social benefit.

So you're right: the motivations of high-frequency traders are irrelevant, except insofar as we can change their behavior by changing the incentives they respond to.

reply

sseveran 13 days ago | link

If teaching were as competitive as market making major cities might graduate more than half their students from high school while simultaneously driving down costs. Non profit is not a badge of honor.

reply

smoorman1024 13 days ago | link

Yes. Kind of a good example how competitive systems lead to a better end result for customers. Competitive teaching is something we should actually encourage. Tyler Cowen often argues how the availability of systems to teach to a huge number of students is going to make teaching superstars and bring the cost of teaching down to almost zero because the marginal cost of distributing teaching is almost 0 with enough students.

reply

gshubert17 13 days ago | link

If more students and their families in major cities understood the benefits of learning in high school and were able and motivated to pursue them -- that is, if being knowledgeable were a badge of honor -- would schools be more likely to respond to them, with an ultimate result of higher graduation rates?

reply

steveax 13 days ago | link

Yes of course. All markets are the same. High frequency teachers FTW!

reply

panzagl 13 days ago | link

Teaching is plenty competitive- starting teacher salaries are low because there's lots of competition. If you want results though, you have to pay.

reply

lmm 12 days ago | link

> You make it sounds like HFT is some sort of non-profit, community service vocation, like teaching, or fire fighters. Do people actually think this?

They're doing it to make money, but that doesn't mean they're not helping other people out. My uncle is an importer; most of his time is spent finding something that's cheaper in Korea than it is here and drawing up a contract with a supplier there and a buyer here. Does he do it as a community service? No. Does he take his cut? Yes. But every deal he makes benefits both the supplier and the buyer.

reply


…which is, on average, $5K/month. Not exactly a ton of money.

reply

Kudos 13 days ago | link

An average without context is not useful information.

When I last checked, a sample price for Cloudflare is $3k for 100TB. Cloudflare is far from the cheap option when it comes to serving content.

reply


A relatively easy way to change that would be to use Intel's ISPC compiler.

http://ispc.github.io/

reply


And it's been around forever. If you're interested in high-speed processing, there's lots of good papers from that project (e.g. how to avoid creating branches, stuff like that).

reply


Not ironic, I'd love to see it too. I like knowing how things were done originally. :)

reply


> Isn't that what LLVM tries to solve dynamically?

No. And in fact, even writing directly in LLVM IR, there are certain kinds of code (high-speed interpreters like LuaJIT) that cannot even be expressed.

LLVM is awesome, but it's a language (LLVM IR) that does not support the full range of what can be computed efficiently with a particular CPU.

reply


The reason you're wrong is because the semantics of C restrict what compilers can do. Humans know the actual semantics they require, and thus, when writing assembly language, can retain only those semantics, dropping everything else. Thus, "ingenuity" as you put it is a real advantage that humans actually posses over what a compiler can do—even if we gave that compiler an infinite amount of time.

This is quite unlike chess, where the rules are known up front, and are the same for both the computer and the human.

reply

zwegner 13 days ago | link

The reason you're wrong is that I wasn't talking about C :)

Yes, for any current programming language, humans do have some advantage by being able to utilize both low- and high-level semantic information that simply cannot be communicated to the compiler. The trump card, I believe, is coming up with new languages where this is not the case.

reply


> A Basic Income policy is totally Socialism, which is a system in which the economy is cooperatively managed.

Disagree. Provision of money and "co-operative management" are two separate concepts. You can do either, both, or none.

-----

More

Lists | RSS | Bookmarklet | Guidelines | FAQ | DMCA | News News | Feature Requests | Bugs | Y Combinator | Apply | Library

Search: