Personally I think this adds no value to the social goodput of trading: its not market informing of real movement, its profit (rent?) seeking in dis-equal access to service.
I understand the profit motive. I just don't think that for trades, informing community of value (prices, derivitives, futures) that moving to femto-seconds is adding value.
"I made a lot of money because I have faster komputer than u"
It requires a great deal less work. No realtime, etc. And the interest on both sides is aggregated for the entire period, leading to fairer prices with less volatility.
Theory is the faster the market is, less the room for arbitrage, and better for the day trader. HFT has certainly narrowed the spread.
Big problem currently is moving large block (institutional) trades without affecting the market. Thats why a good chunk has moved to dark pools. (I work for a dark pool ATS).
Isn't it fundamentally impossible to order everything to an arbitrary precision? If the time between two events is less than the distance between them divided by the speed of light, then which one came first depends on the frame of reference. When you're talking about nanoseconds or less, you're inherently never going to be able to reconcile things that are miles apart, is my understanding.
The frame of reference is at the exchange. Having incredibly precise time synchronization among computers at the exchange would allow distribution of order acceptance for a given symbol without losing ordering. You could give each HFT a cabinet with your order acceptance port at the back and a processing computer exactly the same distance away from the port. Order processing could run asynchronously from order acceptance, and resort according to accpetance time in batches. This would eliminate complaints about such and such company being three fempto meters closer than me, so they get better execution.
> If the time between two events is less than the distance between them divided by the speed of light, then which one came first depends on the frame of reference.
It's not hard to pick a location and use its reference frame.
> When you're talking about nanoseconds or less, you're inherently never going to be able to reconcile things that are miles apart
Let's say four miles, for 20 light-microseconds. To be even .2 nanoseconds off, you'd need a time dilation of at least ten parts per million. That's a huge amount. Even GPS satellites, with their very high speed and distance, measure time dilation in fractions of a part per billion.
For things that are "miles" apart, you can completely ignore relativity at the nanosecond level. Or even between any two locations on the planet.
"A natural and very old question in relativity is: can such a global and meaningful time coordinate be constructed for the observers sitting on a disk that rotates at, say, constant angular velocity? It turns out that it cannot. The observers sitting on such a disk cannot agree with each other on the simultaneity of events. In a real sense, this means that no matter how perfect the clocks might be that we use on our rotating Earth, they will still disagree on the simultaneity of events. In fact, this disagreement amounts to tens of nanoseconds, globally."
> In fact, this disagreement amounts to tens of nanoseconds, globally.
I think that number is wrong. When I take 460 meters per second at the equator, double it, and put it in a time dilation calculator, I get a dilation factor of 5e-12. Multiply that by 150ms and you get less than a picosecond of dilation.
But if I made a mistake, and that number is correct, then my core point is still completely right. Locations on Earth that are miles apart are still synced within far less than a nanosecond.
It's the intersection of two different models of the financial universe.
The long-term investor has an interest in liquidity and a reasonable response time to market moving news. He's making decisions about a company based on expectations of the future. Typically, if his trade clears in 1 minute-- or even 1 day-- instead of 1 millisecond, there isn't that much in it except in some corner case situations.
The short-term investor is mostly interested in reacting to the noise in the signal-- getting a few nanoseconds ahead of other sellers and positioning accordingly. If you have enough of a "we got the info first" edge, you don't have to have any real opinion of the companies you're investing in, or even a preference of long or short, you can systematically ride the signals.
The justifications for ever-faster trading tend to break down as "more liquidity" and "less arbitrage", but we're not asking if the marginal increases in performance there are worth the costs-- both in the arms-race for faster market and informaion access, and in the impacts on market behaviour caused by large numbers of super-short-term investors throwing their inertia around tiny swings in the market.
it is right at the intersection of things you are more familiar with in tech:
aggregating user data and selling it.
people with dis-equal access to service and collecting everyone else's orders midstream and selling it:
the machine two racks away from the matching engine is selling information to the machine one rack away from the matching engine
and this goes all the way out to your broker, your etrade or robinhood who is selling all of your profile and trade data to the machine two racks away from the matching engine
it adds a lot of value for a lot of people. the actual femtosecond trade isn't what anybody cares about.
regarding the equality, it wasn't that long ago where you had dis-equal access if you couldn't physically be at the stock exchange, in fact thats where you were supposed to be if you wanted to trade. so I can't buy that argument, in isolation.
I do think it is a form of front-running, which is regulated in some capacity but very specifically worded such that your broker can't place a trade before you just because they found out you wanted to. what we have here is front-running but from different unrelated companies in the chain and all automated by machine.
Yes, good analysis. Whats the goodput (the net social benefit) and whats the badput?
Do I (for instance) pay more? I suspect I do. I suspect the intermediary sniping in, forces prices up to their advantage. The dis-equality is being abused, to extract rent from me because I am slow. HURRY UP! Bidness waiting.
Which means, overall, we all pay more. This is cost shifting to make money, by snatching goods in flight, and charging a minim to pass them on. The intermediary has no intention of doing anything except profiting. they aren't holding, they aren't helping.
I'm not a free market person btw. So, I'd probably say that about a lot of behaviour in 'the market'
I suspect that I (as an individual investor in my retirement account and personal investments, not as an HFT) pay less.
My mental model is that pre-HFT, banks and other market makers made a lot more money per share traded. HFT outcompeted them and the HFTs make less money per share. I, as an individual investor, get to keep more of my money and less of it goes to various intermediaries.
-you pay less for an order
-it takes less time for your order to get filled
-more efficient pricing
-liquidity ripples out into risk management products derived from thick order books
-easier to manage risk
-lower cost of transaction
"badput"
-hard to compete against the liquidity providers
-they have unparalleled view of order flow, which runs counter to a market with fair equal access to information
I recommend you first familiarize yourself with the subject. People are generally misinformed about this subject but feel obligated to opine on it before doing due dilligence, for some reason.
Eventually I have to believe even the financial institutions will decide that the current system is neverending, and they'll work out a system of simultaneous bidding. There's no reason why there can't be a tick-tock where trades queue until then next tick, and then are dealt with according to a fair algorithm of some sort.
I know people may down vote you, however, a new kind of auction system to slow things down could potentially make the market fairer for all participants. The average guy will probably never be able to compete with the big boys that are collocated in the same data centre as the exchange with full access to the order book.
That just changes the risk profile of deciding to execute a trade vs the predictable outcome. It may be more costly or less costly than existing approaches.
Has anyone considered the effect of general relativity on syncing clocks on a nanosecond level over large areas?
I previously read about this topic and it didn't seem to occur to the author (who may have been Levine in fact) that basic physics (not that I'm that knowledgeable about it) puts a limit on how finely you can order physically separated events.
I'm not an expert on this, but my understanding is that General Relativity is already taken into account for by the master time source.
I'm not one of the ops guys, but as explained to me, each of our NJ & NY datacenters, our PTP masters are synced to the DC master which is synced to GPS.
We also have processes that monitor if the SIP feed drifts from our time.
No, there is only a limit on an order of events that holds across all possible reference frames. If you define a fixed reference frame, there is no problem. Presumably, they define the same reference frame as whatever UTC uses.
The reference frame is the location of the server. Whoever is closer to the server has an advantage because someone half way around the world will take an extra 50ms or whatever to submit their order.
Syncing clocks implies to me that you're not just executing trades in the order of arrival. You're trying to determine which trade "actually happened" first. Hence the doubt I had about the relativity of simultaneity. Incidentally, I think I was mistaken about "general" relativity being applicable. This is probably just a "special" relativity problem.
I was reading something that says UTC is based on a non-rotating inertial reference frame, whereas locations of actual people are of course in reference frames co-moving with the Earth.
"UTC is a valid global time for the "Earth-Centred Inertial Frame" (ECI) in which Earth spins, because inertial frames do admit a global time. UTC is not a global time for the "Earth-Centred Earth-Fixed frame" (ECEF) in which we live our daily lives, because the ECEF is a rotating frame in the non-relativistic sense of the word, and so cannot support a global time coordinate with the same meaning as the time used in an inertial frame. That is, when two events have the same UTC time, they are simultaneous for all observers at rest in the ECI, but they are not necessarily simultaneous for any observers in the ECEF, and in fact those two events can be off by as many as 30 nanoseconds in the ECEF."
Something that confuses people is that trades don't happen that often. Batch auctions don't change this. You still need to decide who gets priority in the auction.
If you do time priority then it's not that different from currently existing American exchanges.
If you do size priority then it's not that different from currently existing American exchanges.
If you do random priority then it's asymptotically equivalent to size priority.
Batch auctions just make it harder to quote ETFs and other products whose prices depend on prices of other products. It's likely a net loss.
You claim that batch auctions are the same without addressing the arguments in the article, namely that (1) HFT is unfair to smaller traders, and (2) HFT is an arms race where the costs (materials, labor, and brainpower) exceed the benefits.
Here's the fairness argument: a severe negative report comes out about stock X. In a fair world, after the price updates, every investor loses the same per share of X that they held. In the real world, HFT can sell or short X before others can react, resulting in a net transfer of wealth from investors to HFT firms.
That's not my argument at all. My claim is that technology is still a huge advantage with batch auctions. It might even be a bigger advantage with batch auctions. More complex market structure usually benefits those who understand it best.
I also don't understand your argument about HFTs having an advantage over investors. Investors have advantages over HFTs too. Are those unfair? Why are you so concerned about the advantages that HFTs have over investors, but not advantages that investors have over HFTs?
I also hope you understand how small the HFT industry is. There are single very large investors who make more than the entire HFT industry.
How are batch auctions more complex? You still use the same rules for resolving priority. How can the rules be gamed?
Investors do not have any unfair advantages over technical traders. They take different approaches for price discovery, but both should be rewarded. HFTs have the unfair advantage of having faster access to information, including analyst reports.[1]
I'm not going to delve any further into technical details of batch auctions.
I never said anything about unfair advantages. You're the one who's ascribing fairness to this whole thing. I just fail to see how hiring smart engineers (HFTs) is an unfair advantage whereas hiring smart former investment bankers and research analysts (hedge funds, mutual funds, other investors) is not. It just seems like yet another baseless attack on technical nerds.
Also I hope I didn't give the impression that HFTs and investors are competing with each other. HFTs more naturally compete with banks and other market makers. It's just that the advantages that investors have are one of the reasons why HFTs don't expect to make half a spread every time they trade.
You could argue that the work of HFTs is less important than work of investors and that they should make less than the investors do. I would agree with you, but this is already the case and it's not even close how much more money investors make.
Arguments (1) and (2) run counter to both academic investigation[1,3] and industrial consensus[4,5]. Outside of IEX and its affiliates, you don't really find mainstream criticism of HFT from authoritative sources.
As a particular point of contention: most academic studies of the subject find that HFT activity has significantly improved liquidity and price accuracy across e.g. equity trading. But at the same time, HFT as an aggregate industry is both tiny and shrinking. Despite accounting for nearly 60% of equity trading volume last year, the HFT industry had aggregate revenues of ~$1B[2]. So no, it's not a net wealth transfer from investors to HFT firms, and their (ostensibly positive) impact is dramatically outsized compared to their activity (this is aside from the fact that HFT activity is virtually imperceptible to retail traders anyway).
As a meta comment, I think you should reexamine your assertion about what would happen in a "fair world", because it's game theoretically vacuous. Under your paradigm, how does price discovery function when informational and temporal asymmetry are eliminated? How would the market exist and accomplish its functions in a way substantially similar to the way it does presently?
Stated another way, what "fair" advantages would you allow market participants to compete with that are not already a function of information asymmetry or speed?
Thanks for the thoughtful response. I'm still reading through the efficiency analyses.
I note that none of your sources refute the fairness argument. Source [3] concedes it, concluding that "HFTs impose adverse selection costs on other investors, by trading with them when they (HFTs) have better information".
> how does price discovery function [without] asymmetry? what "fair" advantages would you allow
Why wouldn't the market work without information asymmetry? Pricing would still be a difficult problem because in practice, it's difficult to predict the future even with "perfect" information. The market will reward good judgment.
I agree that the market also motivates useful information discovery, such as satellite imagery, but this ties back into the question of whether microsecond pricing actually increases efficiency.
Your argument presupposes that there is some constant per-share price to be paid for a piece of bad news, when in fact it is the function of the market, through trading, to arrive at a consensus valuation for that news.
I can rephrase without that assumption. The argument is that high speed traders are able to trade before others can receive or react to information. The time advantage therefore translates to an unfair information advantage. Compare insider trading, which we make illegal because it is unfair, even though it improves the price signal.
The direct benefit of the real-time market vs. batch auctions is that pricing information is updated faster in the real-time market, but is there any need for microsecond resolution pricing?
> Compare insider trading, which we make illegal because it is unfair, even though it improves the price signal.
Insider trading is arguably illegal because you're misappropriating company information for your own use. There a few other technical points, but they can be ignored in a casual conversation.
This applies if the CEO told you something that they shouldn't have or if they told you something in confidence and you abused that trust.
Trading on information that others don't have is generally legal and investors do it to a much greater extent and to much greater profit than HFTs.
> The direct benefit from HFT is that pricing information is updated faster in the market, but is there any need for microsecond resolution pricing?
The big direct benefit comes from maintaining fair prices for ETFs and other products whose prices depend on other products.
Furthermore, "microsecond" pricing is a very natural system. HFTs and other market makers put out passive limit orders whenever they want. Investors and other aggressive participants send marketable (usually limit) orders if it's the price that they want.
It's the far more complicated batch auctions that need to justify their existence.
Fairness is the "straightforward", "textbook answer" as to why insider trading is illegal. Misappropriating company information" isn't a great justification because insider trading is still illegal even if a company explicitly allowed employees to trade on nonpublic information.
Just to be clear, I'm disagreeing with you because I think you're wrong about the world as it is not because our interests are opposed. I would have far more money if the US had a super complicated batch auction system and everybody could only trade on public information. It would probably only have a tiny, barely noticeable, negative effect on your life in the form of higher spreads for ETFs and various index products.
> "Misappropriating company information" isn't a great justification because insider trading is still illegal even if a company explicitly allowed employees to trade on nonpublic information.
I am not sure why you are bringing up this specific and very complicated example. Has this been tested in court and I'm just not aware of it? I don't even see how a company could allow its employees to engage in insider trading without it being misappropriation. The company has a responsibility to its shareholders. At a minimum, this policy would have to be disclosed to investors.
Investors very frequently use data that is not publicly available. Sometimes it's mosaics that they've put together from smaller pieces of information that are not public, sometimes it's body language of executives in private meetings to which the public does not have access, sometimes it's satellite images, sometimes it's information from IR, and sometimes it's something else. None of this is illegal without additional clarification.
You're right, it was a poor example and I have deleted it. Perhaps we can agree that fairness would be an adequate justification for making insider trading illegal, and by analogy, that it would be an adequate justification for switching to batch auctions. If so, you have the burden of proof to show that the benefits of microsecond resolution outweighs our considerations of fairness.
> Perhaps we can agree that fairness would be an adequate justification for making insider trading illegal
I don't think that we agree on this. The way that I see it, investors always try to leverage information asymmetry and I don't see how you could realistically outlaw it.
That's what makes HFT so hard. The people who are trading against you inherently know more than you do. They've talked to the companies, they've researched them, etc. Meanwhile HFTs stand by willing to buy and sell just below and just above mid market without any of that information.
This is also why HFTs are willing to pay for retail order flow. At least you know that the retail orders probably don't know more than you so you're able to capture a greater fraction of the spread.
Of course retail orders are a small fraction of all orders and an even smaller fraction of orders on the exchanges.
You may be interested in reading Salman v. United States [1]. It's a dense read so here's a news article [2] that I'll quote.
> The justices rejected claims of defense lawyers who argued that there was no crime because no money exchanged hands between the insider and the stock trader.
> Instead, the court said exchanges within a family are like gifts and have value, even if no dollars are paid.
Sounds a lot like misappropriation is the concern and not merely having access to information that others don't.
This is a case that's widely seen as making it easier to prosecute insider trading.
True, it's easier to prosecute insider trading as a breach of fiduciary duty to the company. But why were insider trading laws passed, and why is insider trading prosecuted? The SEC answers that "moral imperatives have driven the development of insider trading law in the United States."[1] The prosecutor against Salman, Preet Bharara, stated that "today's decision is a victory for fair markets and those who believe that the system should not be rigged."[2]
> "Misappropriating company information" isn't a great justification because insider trading is still illegal even if a company explicitly allowed employees to trade on nonpublic information.
I can't think of a reason why it would ever be in shareholders' interests to allow insider trading. If a company did ever allow it, it would likely either be a case of company officers misappropriating information from shareholders or a case of majority shareholders misappropriating information from minority shareholders.
Developing proprietary sources of information about companies and trading on it isn't "insider trading". And again, it's pretty fundamental to how the market works: the whole idea is that people will learn things about companies, trade on that information, profit, and in doing so gradually nudge the price of those companies to their true value.
Insider trading is an agency problem where (simplifying a bit) you're brought into confidence by the company and trade against the interests of their shareholders. As an agent of the company and thus its shareholders, you're not allowed to do that. HFT traders aren't agents of anybody.
I said compare insider trading. I'm aware that it doesn't involve insider information; the point is that some information advantages can be unfair. We want the market to reward real research, and not front-running.
I don't understand what you're trying to argue. Front-running is also an agency problem: it occurs when you hire a broker/dealer to execute orders from you, and, upon receiving your order, the broker trades for their own account off it first. HFTs are proprietary traders.
Front-running doesn't simply mean "someone who races to trade first". If you think about it, that definition doesn't even make sense. Does everyone who owns a Bloomberg terminal front-run everyone who doesn't?
The more expansive view of front running was most famously argued in Flash Boys, where "Lewis concludes that HFT is used as a method to front run orders placed by investors".[1] Under this definition, yes, why should traders need a Bloomberg terminal just to remain on a level playing field? The five-second batch proposal would make the market more fair. We should compete on real information and not on microsecond latency.
You say that this doesn't make sense, but I'm very interested in why. Of course, we want price updates to be timely, but why would microsecond resolution be useful?
At the point where you've argued that using a Bloomberg terminal means you're front-running everyone who doesn't have one, I feel like I can end this unproductive discussion by asking: "what's a definition of front-running that would be persuasive to someone who doesn't believe that simply owning a Bloomberg terminal is unfair?"
I don't want to argue semantics. I would like a straight answer on why you think that "fastest trader takes all" is a level playing field, or conversely, why implementing a batch auction format would be nonsensical. I'm agree that Lewis's book has many problems, and Kovac's book does not address batch auctions at all.
If your whole argument is that HFT is bad because it's "front running" or "insider trading", and you can't define the terms in ways that don't also invalidate the commonplace ways everyone accesses the market today, then what we haven't isn't a mere semantic argument; it's that you don't have a coherent argument at all.
Is HFT not front-running or insider trading? (I'd agree!) Then what's wrong with it? Otherwise: what's the useful definition of either of those terms that you're working from?
The problem with HFT is that it's unfair and discourages investor participation. A batch auction system would level the playing field. Obviously, HFT is currently legal and profitable; I'm talking about reforming the market.
In what way does HFT "discourage investor participation"? Before there was high-speed electronic trading, virtually no investors were out there trying to outcompete market makers or run elaborate pairs trading strategies.
That's the term the SEC uses to explain why a fair market is important. If ordinary investors and traders believe that low latency traders have systematic advantages over them, they will lose confidence in the market and choose not to participate.
There is no evidence that HFT dissuades ordinary investors from participating in the market. The Chief Investment Officer at Vanguard, the most important retail investor in the world, has repeatedly said that HFT is an unalloyed good for them: if what you do is buy stock and hold it, all HFT does is make it cheaper to execute your orders. You can also just, you know, look at stock prices and decide for yourself whether people are avoiding the stock market.
I'm sure HFT does dissuade a class of day traders from participating in the market. But who cares? Electronic trading also killed the human market makers, who simply got outcompeted. The result, again, was an unalloyed good: the human market makers were crooked as a wallet full of 3 dollar bills.
I think I understand your view that low latency trading is fair competition. I see that from the perspective of traders, it's unfair skimming that's literally impossible to compete with. (Compare in more corrupt days, you could literally pay NASDAQ for faster access to price information.) I am still interested in that competition improves spreads or efficiency.
There is obvious evidence that algorithmic trading has improved spreads: look at the spreads. Since the advent of electronic trading, they've fallen from almost a dollar(!) to pennies. Human traders colluded to keep spreads high; Google [odd eighths].
The microsecond level competition comes from the fact that the HFTs can't compete on price below one cent, because the minimum quote increment is one cent. So they have to compete with each other on speed. There have been proposals to allow sub-penny quotes, and thus let the HFTs compete more on price and possibly bring spreads down even further, but those proposals don't seem to have gone anywhere.
I'm long over arguing about HFT's fairness, but what I will say is that 'fastest trader takes all' is not a correct way to describe the current state of the world. The current state of the world is closer to 'soonest to best price' or 'soonest to best price with weird rules around quantity' for some markets.
No amount of speed on the traders side will allow them to beat me on a resting order* I put in 2 months ago, even if I did it on the phone with a human broker. Their speed is a risk management feature. They don't have to have the carry risk for those 2 months that I do. But they don't have any special access to the trade at that price.
*note that there is no such thing as 'the market'. There very well may be exchanges out there that have special order types that can trump mine, but they aren't common (or weren't last i was involved in HFT).
In context, I mean that the fastest trader takes all of the profit associated with the price correction in response to new information. Your example isn't relevant since you placed your order without access to the new information. In fact, after seeing the new information, you may realize that your order is now a losing trade--but too late, HFT has already arbitraged away your inefficiency.
No, that's false. He put a standing limit order on the market weeks ago in anticipation of the market correction you're stipulating. It doesn't matter how quickly you can respond to the news; his order beats yours.
You as another market participant (HFT or otherwise) , have no idea if I have access to the new information or not.
In fact, the vast majority of the new information that HFT participants are getting is the order itself. My order goes in and that itself is the information that causes reactions.
And thats the crux of the reason 'smart' money hates it. They want to take advantage of liquidity without it being priced correctly. Meanwhile liquidity providers (HFT) want to price it based on that order information. But note, the built in information asymmetry actually works against the HFT in that case. By the time the smart money order hits the HFT, the smart money already gotten their desires in at least part of the market.
For me as 'dumb' money its largely all upside. I can get execution right now for as cheap as its ever been if that is important to me, or I can set the price I actually want to pay without being impacted by this at all.
> You ... have no idea if I have access to the new information or not
You assume that you can obtain on information before HFT, possibly because you're talking about a different type of HFT than I am.
Can we use a clearer example? Let's say that we submit a limit order to buy 10 shares of AA (American Airlines). Before the order is fulfilled, two AA planes collide (the new information). In response to the new information, any trader would agree that the stock price should drop significantly, and it does. HFT will quickly sell to our previous uninformed trade, which we are not able to cancel in time--that's the unfair advantage. Please explain to me how this is not "fastest trader takes all".
I'm happy to discuss this using your clearer example supposing you agree to a couple of ground rules:
- your example has never happened. Its a thought experiment that is worth exactly that.
- you stipulate that the actual information that HFT systems take advantage of is the market microstructure itself. I'm going to demand that you do this because of my experience in the HFT space and my lack of experience in the 2 plains colliding in mid air space
Even given those stipulations, HFT is not the issue. If I am on the ground observing the crash. I have a unique and privileged access to the information. If I am a air traffic controller observing their flight plans before they hit, I have a unique and privileged access to the information. If I am a hedge fund that have been detailing the long history of AA job boards and notice that they spend waaay less on QA engineering than their competitors, I have unique and privileged access to the information. It is not the point of the markets to penalize those participants. It is the point of the markets to reward them. We want that information incorporated into the price as soon as possible and the markets are how we incentivize that.
That is, having data that other market participants don't is a fundamental state of the markets that generates the outcomes we want. Any external manipulation of that information need have a sufficient justification. In the cases we've said are out of bounds, the sufficient justification comes from an issue of agency. We have said that brokers can't take advantage of your order flow because it will destroy the markets if they did. Insider trading represents one party stealing from another (1 set of shareholders who are beholden to another set).
HFT does not represent any of those things. Quite the opposite. They are placing honest bets on their predictive ability, just like every other market participant, the only difference is their information flows and timeframes. Frankly, the markets hold HFT to a higher standard than other participants. No one bailed out Knight when their advantage became their undoing and no one cares now when they are providing liquidity for basically free and the entire industry is a non-event. If an investment bank or an automotive company succumbed to market pressures like HFT did, a president would step in.
My example would apply whenever an analyst comments on AA or whenever any polarized headlines is published about AA.
> It is the point of the markets to reward [information]
But that's exactly what we're debating: why is updating information a couple microseconds earlier something that is valuable? Conversely, if it doesn't make the market more efficient, it shouldn't be rewarded.
Since you mainly want to talk about market microstructure, how do you think a switch to batch auctions would affect the market?
I presume you are talking about Budish style batch auctions?
I think it is an interesting paper but it ignores the reality of the market place, which is that there is no such thing as 'the market'. There are a collection of exchanges, spanning the globe, under many different jurisdictions offering either directly fungible symbols or things that are very tightly correlated.
Even if every exchange in the world agreed to the same schedule and time interval there would be arbitrage opportunities available to 'mechanical' HFT at the time scales mentioned in that paper due to the nature of physics and global communication latencies. You'd need to increase their interval value considerably even in that highly unlikely scenario. If they did that, I'd presume that in the short term spreads would go up as the current market makers reduced their risk exposure as they figured out the new system. In the long term either the existing players would figure out their tolerances or new players would come around and things would largely settle back to where they are now. My intuition is that the games they would end up playing in that world would be largely more goofy than the fairly straight forward latency arbitrage they do now (I think it would look more like dark pools than like our current exchanges).
But thats not even realistic. In any realistic scenario a batch auction exchange would exist in an eco system with other non-auction exchanges. It would still have latency based arbitrage opportunities and would largely be a non-event to the markets. HFT firms would continue to do the yeomans work of making sure the prices on it match the prices on the other exchanges. My bet is that the pricing on that exchange would largely be worse for everyone and that most customers would not care enough to make that exchange more than a curiosity. IEX had one of the best marketing plans of all time and haven't changed the market in any significant way with their goofy speed bump concept.
All of that to fix something that I don't view as broken. HFT is one of the most purely Darwinian capitalist systems in the world. Margins are razor thin. I'm happy to have them invest all that money and pay for my order flow so that I can get free trading and near instant execution. Most of the major commercial players like Vanguard agree with me. That a couple of hedge fund whales don't like it doesn't bother me much.
A random sample of shares is naturally biased to whichever shares are part of the same large order. (If I offer to trade 9 shares, and you offer to trade 1, I have 9x the change of one of my shares being traded instead of yours)
I don't know. I think ~15% of the daily volume is traded at the opening/closing auctions. Why do you think the end-of-second (or end-of-tenth-of-second) auctions won't be popular?
> If people want to trade stocks every hundredth of a nanosecond, why not let them?
> but if that meaningless-seeming competition between high-frequency trading firms helps subsidize basic research to improve our understanding of time itself, isn’t that kind of cool?
Seems to me this guy is really struggling to reach the conclusion his editor wants..
92 comments
[ 2.4 ms ] story [ 142 ms ] threadhttps://news.ycombinator.com/item?id=17429625
I understand the profit motive. I just don't think that for trades, informing community of value (prices, derivitives, futures) that moving to femto-seconds is adding value.
"I made a lot of money because I have faster komputer than u"
Big problem currently is moving large block (institutional) trades without affecting the market. Thats why a good chunk has moved to dark pools. (I work for a dark pool ATS).
It's not hard to pick a location and use its reference frame.
> When you're talking about nanoseconds or less, you're inherently never going to be able to reconcile things that are miles apart
Let's say four miles, for 20 light-microseconds. To be even .2 nanoseconds off, you'd need a time dilation of at least ten parts per million. That's a huge amount. Even GPS satellites, with their very high speed and distance, measure time dilation in fractions of a part per billion.
For things that are "miles" apart, you can completely ignore relativity at the nanosecond level. Or even between any two locations on the planet.
http://math.ucr.edu/home/baez/physics/Relativity/SR/rotating...
I think that number is wrong. When I take 460 meters per second at the equator, double it, and put it in a time dilation calculator, I get a dilation factor of 5e-12. Multiply that by 150ms and you get less than a picosecond of dilation.
But if I made a mistake, and that number is correct, then my core point is still completely right. Locations on Earth that are miles apart are still synced within far less than a nanosecond.
The long-term investor has an interest in liquidity and a reasonable response time to market moving news. He's making decisions about a company based on expectations of the future. Typically, if his trade clears in 1 minute-- or even 1 day-- instead of 1 millisecond, there isn't that much in it except in some corner case situations.
The short-term investor is mostly interested in reacting to the noise in the signal-- getting a few nanoseconds ahead of other sellers and positioning accordingly. If you have enough of a "we got the info first" edge, you don't have to have any real opinion of the companies you're investing in, or even a preference of long or short, you can systematically ride the signals.
The justifications for ever-faster trading tend to break down as "more liquidity" and "less arbitrage", but we're not asking if the marginal increases in performance there are worth the costs-- both in the arms-race for faster market and informaion access, and in the impacts on market behaviour caused by large numbers of super-short-term investors throwing their inertia around tiny swings in the market.
aggregating user data and selling it.
people with dis-equal access to service and collecting everyone else's orders midstream and selling it:
the machine two racks away from the matching engine is selling information to the machine one rack away from the matching engine
and this goes all the way out to your broker, your etrade or robinhood who is selling all of your profile and trade data to the machine two racks away from the matching engine
it adds a lot of value for a lot of people. the actual femtosecond trade isn't what anybody cares about.
regarding the equality, it wasn't that long ago where you had dis-equal access if you couldn't physically be at the stock exchange, in fact thats where you were supposed to be if you wanted to trade. so I can't buy that argument, in isolation.
I do think it is a form of front-running, which is regulated in some capacity but very specifically worded such that your broker can't place a trade before you just because they found out you wanted to. what we have here is front-running but from different unrelated companies in the chain and all automated by machine.
Do I (for instance) pay more? I suspect I do. I suspect the intermediary sniping in, forces prices up to their advantage. The dis-equality is being abused, to extract rent from me because I am slow. HURRY UP! Bidness waiting.
Which means, overall, we all pay more. This is cost shifting to make money, by snatching goods in flight, and charging a minim to pass them on. The intermediary has no intention of doing anything except profiting. they aren't holding, they aren't helping.
I'm not a free market person btw. So, I'd probably say that about a lot of behaviour in 'the market'
My mental model is that pre-HFT, banks and other market makers made a lot more money per share traded. HFT outcompeted them and the HFTs make less money per share. I, as an individual investor, get to keep more of my money and less of it goes to various intermediaries.
-you pay less for an order -it takes less time for your order to get filled -more efficient pricing -liquidity ripples out into risk management products derived from thick order books -easier to manage risk -lower cost of transaction
"badput"
-hard to compete against the liquidity providers -they have unparalleled view of order flow, which runs counter to a market with fair equal access to information
So prices are some pennies more efficient, but some of the trading money is going to the pockets of high frequency traders.
This is good! To a point. Eventually the price tag of improved liquidity could be higher than the value of that liquidity.
We are long past trading adding any actual value to society.
I recommend you first familiarize yourself with the subject. People are generally misinformed about this subject but feel obligated to opine on it before doing due dilligence, for some reason.
everyone else would think of that too
I previously read about this topic and it didn't seem to occur to the author (who may have been Levine in fact) that basic physics (not that I'm that knowledgeable about it) puts a limit on how finely you can order physically separated events.
https://en.wikipedia.org/wiki/Relativity_of_simultaneity
I'm not one of the ops guys, but as explained to me, each of our NJ & NY datacenters, our PTP masters are synced to the DC master which is synced to GPS.
We also have processes that monitor if the SIP feed drifts from our time.
"UTC is a valid global time for the "Earth-Centred Inertial Frame" (ECI) in which Earth spins, because inertial frames do admit a global time. UTC is not a global time for the "Earth-Centred Earth-Fixed frame" (ECEF) in which we live our daily lives, because the ECEF is a rotating frame in the non-relativistic sense of the word, and so cannot support a global time coordinate with the same meaning as the time used in an inertial frame. That is, when two events have the same UTC time, they are simultaneous for all observers at rest in the ECI, but they are not necessarily simultaneous for any observers in the ECEF, and in fact those two events can be off by as many as 30 nanoseconds in the ECEF."
http://math.ucr.edu/home/baez/physics/Relativity/SR/rotating...
If you do time priority then it's not that different from currently existing American exchanges.
If you do size priority then it's not that different from currently existing American exchanges.
If you do random priority then it's asymptotically equivalent to size priority.
Batch auctions just make it harder to quote ETFs and other products whose prices depend on prices of other products. It's likely a net loss.
Here's the fairness argument: a severe negative report comes out about stock X. In a fair world, after the price updates, every investor loses the same per share of X that they held. In the real world, HFT can sell or short X before others can react, resulting in a net transfer of wealth from investors to HFT firms.
I also don't understand your argument about HFTs having an advantage over investors. Investors have advantages over HFTs too. Are those unfair? Why are you so concerned about the advantages that HFTs have over investors, but not advantages that investors have over HFTs?
I also hope you understand how small the HFT industry is. There are single very large investors who make more than the entire HFT industry.
Investors do not have any unfair advantages over technical traders. They take different approaches for price discovery, but both should be rewarded. HFTs have the unfair advantage of having faster access to information, including analyst reports.[1]
[1] https://www.bloomberg.com/professional/product/news/
I never said anything about unfair advantages. You're the one who's ascribing fairness to this whole thing. I just fail to see how hiring smart engineers (HFTs) is an unfair advantage whereas hiring smart former investment bankers and research analysts (hedge funds, mutual funds, other investors) is not. It just seems like yet another baseless attack on technical nerds.
Also I hope I didn't give the impression that HFTs and investors are competing with each other. HFTs more naturally compete with banks and other market makers. It's just that the advantages that investors have are one of the reasons why HFTs don't expect to make half a spread every time they trade.
You could argue that the work of HFTs is less important than work of investors and that they should make less than the investors do. I would agree with you, but this is already the case and it's not even close how much more money investors make.
I started with an example where investors lose money due to adverse selection from HFTs, and I don't think that you've really addressed that argument.
As a particular point of contention: most academic studies of the subject find that HFT activity has significantly improved liquidity and price accuracy across e.g. equity trading. But at the same time, HFT as an aggregate industry is both tiny and shrinking. Despite accounting for nearly 60% of equity trading volume last year, the HFT industry had aggregate revenues of ~$1B[2]. So no, it's not a net wealth transfer from investors to HFT firms, and their (ostensibly positive) impact is dramatically outsized compared to their activity (this is aside from the fact that HFT activity is virtually imperceptible to retail traders anyway).
As a meta comment, I think you should reexamine your assertion about what would happen in a "fair world", because it's game theoretically vacuous. Under your paradigm, how does price discovery function when informational and temporal asymmetry are eliminated? How would the market exist and accomplish its functions in a way substantially similar to the way it does presently?
Stated another way, what "fair" advantages would you allow market participants to compete with that are not already a function of information asymmetry or speed?
_________________
1. https://onlinelibrary.wiley.com/doi/full/10.1111/j.1540-6261...
2. https://www.ft.com/content/d81f96ea-d43c-11e7-a303-9060cb1e5...
3. https://www.ecb.europa.eu/pub/pdf/scpwps/ecbwp1602.pdf
4. https://www.bloomberg.com/view/articles/2014-06-20/why-i-lov...
5. https://www.wsj.com/articles/high-frequency-hyperbole-139639...
I note that none of your sources refute the fairness argument. Source [3] concedes it, concluding that "HFTs impose adverse selection costs on other investors, by trading with them when they (HFTs) have better information".
> how does price discovery function [without] asymmetry? what "fair" advantages would you allow
Why wouldn't the market work without information asymmetry? Pricing would still be a difficult problem because in practice, it's difficult to predict the future even with "perfect" information. The market will reward good judgment.
I agree that the market also motivates useful information discovery, such as satellite imagery, but this ties back into the question of whether microsecond pricing actually increases efficiency.
The direct benefit of the real-time market vs. batch auctions is that pricing information is updated faster in the real-time market, but is there any need for microsecond resolution pricing?
Insider trading is arguably illegal because you're misappropriating company information for your own use. There a few other technical points, but they can be ignored in a casual conversation.
This applies if the CEO told you something that they shouldn't have or if they told you something in confidence and you abused that trust.
Trading on information that others don't have is generally legal and investors do it to a much greater extent and to much greater profit than HFTs.
> The direct benefit from HFT is that pricing information is updated faster in the market, but is there any need for microsecond resolution pricing?
The big direct benefit comes from maintaining fair prices for ETFs and other products whose prices depend on other products.
Furthermore, "microsecond" pricing is a very natural system. HFTs and other market makers put out passive limit orders whenever they want. Investors and other aggressive participants send marketable (usually limit) orders if it's the price that they want.
It's the far more complicated batch auctions that need to justify their existence.
[1] https://www.investor.gov/additional-resources/general-resour...
Edit, previous response:
Fairness is the "straightforward", "textbook answer" as to why insider trading is illegal. Misappropriating company information" isn't a great justification because insider trading is still illegal even if a company explicitly allowed employees to trade on nonpublic information.
[1] https://blogs.wsj.com/law/2011/10/26/why-exactly-is-insider-...
> "Misappropriating company information" isn't a great justification because insider trading is still illegal even if a company explicitly allowed employees to trade on nonpublic information.
I am not sure why you are bringing up this specific and very complicated example. Has this been tested in court and I'm just not aware of it? I don't even see how a company could allow its employees to engage in insider trading without it being misappropriation. The company has a responsibility to its shareholders. At a minimum, this policy would have to be disclosed to investors.
Investors very frequently use data that is not publicly available. Sometimes it's mosaics that they've put together from smaller pieces of information that are not public, sometimes it's body language of executives in private meetings to which the public does not have access, sometimes it's satellite images, sometimes it's information from IR, and sometimes it's something else. None of this is illegal without additional clarification.
I don't think that we agree on this. The way that I see it, investors always try to leverage information asymmetry and I don't see how you could realistically outlaw it.
That's what makes HFT so hard. The people who are trading against you inherently know more than you do. They've talked to the companies, they've researched them, etc. Meanwhile HFTs stand by willing to buy and sell just below and just above mid market without any of that information.
This is also why HFTs are willing to pay for retail order flow. At least you know that the retail orders probably don't know more than you so you're able to capture a greater fraction of the spread.
Of course retail orders are a small fraction of all orders and an even smaller fraction of orders on the exchanges.
> The justices rejected claims of defense lawyers who argued that there was no crime because no money exchanged hands between the insider and the stock trader.
> Instead, the court said exchanges within a family are like gifts and have value, even if no dollars are paid.
Sounds a lot like misappropriation is the concern and not merely having access to information that others don't.
This is a case that's widely seen as making it easier to prosecute insider trading.
[1] https://supreme.justia.com/cases/federal/us/580/15-628/opini...
[2] http://www.latimes.com/business/la-fi-supreme-court-insider-...
[1] https://www.sec.gov/news/speech/speecharchive/1998/spch221.h...
[2] https://www.nytimes.com/2016/12/06/business/dealbook/supreme...
I can't think of a reason why it would ever be in shareholders' interests to allow insider trading. If a company did ever allow it, it would likely either be a case of company officers misappropriating information from shareholders or a case of majority shareholders misappropriating information from minority shareholders.
Insider trading is an agency problem where (simplifying a bit) you're brought into confidence by the company and trade against the interests of their shareholders. As an agent of the company and thus its shareholders, you're not allowed to do that. HFT traders aren't agents of anybody.
Front-running doesn't simply mean "someone who races to trade first". If you think about it, that definition doesn't even make sense. Does everyone who owns a Bloomberg terminal front-run everyone who doesn't?
You say that this doesn't make sense, but I'm very interested in why. Of course, we want price updates to be timely, but why would microsecond resolution be useful?
[1] https://en.wikipedia.org/wiki/Flash_Boys
https://www.amazon.com/Flash-Boys-Insiders-Perspective-High-...
Lewis' book is not well-regarded.
At the point where you've argued that using a Bloomberg terminal means you're front-running everyone who doesn't have one, I feel like I can end this unproductive discussion by asking: "what's a definition of front-running that would be persuasive to someone who doesn't believe that simply owning a Bloomberg terminal is unfair?"
Is HFT not front-running or insider trading? (I'd agree!) Then what's wrong with it? Otherwise: what's the useful definition of either of those terms that you're working from?
I'm sure HFT does dissuade a class of day traders from participating in the market. But who cares? Electronic trading also killed the human market makers, who simply got outcompeted. The result, again, was an unalloyed good: the human market makers were crooked as a wallet full of 3 dollar bills.
No amount of speed on the traders side will allow them to beat me on a resting order* I put in 2 months ago, even if I did it on the phone with a human broker. Their speed is a risk management feature. They don't have to have the carry risk for those 2 months that I do. But they don't have any special access to the trade at that price.
*note that there is no such thing as 'the market'. There very well may be exchanges out there that have special order types that can trump mine, but they aren't common (or weren't last i was involved in HFT).
In fact, the vast majority of the new information that HFT participants are getting is the order itself. My order goes in and that itself is the information that causes reactions.
And thats the crux of the reason 'smart' money hates it. They want to take advantage of liquidity without it being priced correctly. Meanwhile liquidity providers (HFT) want to price it based on that order information. But note, the built in information asymmetry actually works against the HFT in that case. By the time the smart money order hits the HFT, the smart money already gotten their desires in at least part of the market.
For me as 'dumb' money its largely all upside. I can get execution right now for as cheap as its ever been if that is important to me, or I can set the price I actually want to pay without being impacted by this at all.
[edited for clarity]
You assume that you can obtain on information before HFT, possibly because you're talking about a different type of HFT than I am.
Can we use a clearer example? Let's say that we submit a limit order to buy 10 shares of AA (American Airlines). Before the order is fulfilled, two AA planes collide (the new information). In response to the new information, any trader would agree that the stock price should drop significantly, and it does. HFT will quickly sell to our previous uninformed trade, which we are not able to cancel in time--that's the unfair advantage. Please explain to me how this is not "fastest trader takes all".
- your example has never happened. Its a thought experiment that is worth exactly that.
- you stipulate that the actual information that HFT systems take advantage of is the market microstructure itself. I'm going to demand that you do this because of my experience in the HFT space and my lack of experience in the 2 plains colliding in mid air space
Even given those stipulations, HFT is not the issue. If I am on the ground observing the crash. I have a unique and privileged access to the information. If I am a air traffic controller observing their flight plans before they hit, I have a unique and privileged access to the information. If I am a hedge fund that have been detailing the long history of AA job boards and notice that they spend waaay less on QA engineering than their competitors, I have unique and privileged access to the information. It is not the point of the markets to penalize those participants. It is the point of the markets to reward them. We want that information incorporated into the price as soon as possible and the markets are how we incentivize that.
That is, having data that other market participants don't is a fundamental state of the markets that generates the outcomes we want. Any external manipulation of that information need have a sufficient justification. In the cases we've said are out of bounds, the sufficient justification comes from an issue of agency. We have said that brokers can't take advantage of your order flow because it will destroy the markets if they did. Insider trading represents one party stealing from another (1 set of shareholders who are beholden to another set).
HFT does not represent any of those things. Quite the opposite. They are placing honest bets on their predictive ability, just like every other market participant, the only difference is their information flows and timeframes. Frankly, the markets hold HFT to a higher standard than other participants. No one bailed out Knight when their advantage became their undoing and no one cares now when they are providing liquidity for basically free and the entire industry is a non-event. If an investment bank or an automotive company succumbed to market pressures like HFT did, a president would step in.
My example would apply whenever an analyst comments on AA or whenever any polarized headlines is published about AA.
> It is the point of the markets to reward [information]
But that's exactly what we're debating: why is updating information a couple microseconds earlier something that is valuable? Conversely, if it doesn't make the market more efficient, it shouldn't be rewarded.
Since you mainly want to talk about market microstructure, how do you think a switch to batch auctions would affect the market?
I think it is an interesting paper but it ignores the reality of the market place, which is that there is no such thing as 'the market'. There are a collection of exchanges, spanning the globe, under many different jurisdictions offering either directly fungible symbols or things that are very tightly correlated.
Even if every exchange in the world agreed to the same schedule and time interval there would be arbitrage opportunities available to 'mechanical' HFT at the time scales mentioned in that paper due to the nature of physics and global communication latencies. You'd need to increase their interval value considerably even in that highly unlikely scenario. If they did that, I'd presume that in the short term spreads would go up as the current market makers reduced their risk exposure as they figured out the new system. In the long term either the existing players would figure out their tolerances or new players would come around and things would largely settle back to where they are now. My intuition is that the games they would end up playing in that world would be largely more goofy than the fairly straight forward latency arbitrage they do now (I think it would look more like dark pools than like our current exchanges).
But thats not even realistic. In any realistic scenario a batch auction exchange would exist in an eco system with other non-auction exchanges. It would still have latency based arbitrage opportunities and would largely be a non-event to the markets. HFT firms would continue to do the yeomans work of making sure the prices on it match the prices on the other exchanges. My bet is that the pricing on that exchange would largely be worse for everyone and that most customers would not care enough to make that exchange more than a curiosity. IEX had one of the best marketing plans of all time and haven't changed the market in any significant way with their goofy speed bump concept.
All of that to fix something that I don't view as broken. HFT is one of the most purely Darwinian capitalist systems in the world. Margins are razor thin. I'm happy to have them invest all that money and pay for my order flow so that I can get free trading and near instant execution. Most of the major commercial players like Vanguard agree with me. That a couple of hedge fund whales don't like it doesn't bother me much.
Why? I immediately thought of the Central Limit Theorem but I am not sure
Substantially lower spreads, possibly.
> but if that meaningless-seeming competition between high-frequency trading firms helps subsidize basic research to improve our understanding of time itself, isn’t that kind of cool?
Seems to me this guy is really struggling to reach the conclusion his editor wants..
https://vimeo.com/214855845