9 ms·

> High-frequency trading doesn't need to exist Neither does most of the stock market, for many definitions of “need”. But on the flip side, nobody has ever rea

by 542458 3y ago

> High-frequency trading doesn't need to exist Neither does most of the stock market, for many definitions of “need”. But on the flip side, nobody has ever really demonstrated strong negative effects from HFT and many studies show positive effects, so bans seem very kneejerk to me.

stefan_ 3y ago
The point is that whatever function you think HFT serves can be done just as well at fixed tick rates. This whole latency optimization mini game is just a pure waste.
- super256 3y ago
  And in which order would orders be executed when they come in and get queued? A delay wouldn't fix anything because the queue would still be FIFO. The only way to remove HTF would be removing the trading session, and instead only offering the market auction. Btw, a "tick" in trading is the minimum up/down movement of an asset (e.g. 0.25 points on NQ, or 0.01 on TSLA).
  - denotational 3y ago
    > Btw, a "tick" in trading is the minimum up/down movement of an asset (e.g. 0.25 points on NQ, or 0.01 on TSLA). This is true, but the term is overloaded: if I say "the exchange slid the price one tick to comply with Reg. NMS" then I'm using it in the sense you mean; if I say "the model recomputes on each tick" then I'm using it in the sense of the GP.
  - elictronic 3y ago
    Randomize the queue.
    - growse 3y ago
      That'll make everything more expensive, as every trade now has to price in the risk that it won't actually happen. Increasing market friction is not a good idea.
      - cosmojg 3y ago
        Given that randomization need only apply to orders placed simultaneously at identical prices, I don't see how it would increase market frictions enough to significantly limit their utility in price discovery, long-term investing, and the raising of capital. Individuals and corporations aren't making tradable decisions at subsecond intervals.
        growse 3y ago
        By simultaneously, I assume you mean "within one time interval". I suspect you'd be surprised how many orders at the same price can happen in a very very short time window. As for price discovery, a lot of information can happen in one second. Quantizing just adds risk, which adds cost, which harms everyone. > Individuals and corporations aren't making tradable decisions at subsecond intervals. No, but they're benefitting significantly from the fact that the cost of trading is now minuscule compared to what it used to be.
        cosmojg 3y ago
        > Quantizing just adds risk, which adds cost, which harms everyone. No, it adds risk at the scale of the quantum, adding cost for those trading at the scale of the quantum, harming a very small subset of traders who would normally benefit from trading at or below the scale of the quantum, like HFT firms and short-term market makers. Should such institutions cease to exist because the cost of trading at such scales becomes prohibitively expensive, I highly doubt that society will suffer.
        growse 3y ago
        We already have evidence from history (where trading latencies and intervals were much higher) that the spreads become much higher as trading frequency goes down. The spread represents the amount of uncertainty and risk associated with a price. Every bip away from the midpoint your bid/ask sits is pure cost to the participant, and that's true for all participants. Even if I only come to market once a year to splurge my bonus on GME, I'm worse off paying 20c over the midpoint than I am paying 1c.
  - cosmojg 3y ago
    > And in which order would orders be executed when they come in and get queued? That's easy! Come time to match and fill for orders placed at identical prices, there's no reason a random number generator couldn't do the trick. Trading second-by-second need not be reliably profitable in the extremely short term to ensure an orderly and efficient market. The whole point of a static tick rate is to disincentivize such myopic trading in the first place! I agree, though, that the logical conclusion of all this would be to replace the trading session with a market auction. However, while I appreciate the superior allocative efficiency of a Vickrey-style sealed-bid second-price multi-unit double auction[1] as much as the next guy, I have more faith in iterative improvements coming to pass before revolutionary ones. > Btw, a "tick" in trading is the minimum up/down movement of an asset (e.g. 0.25 points on NQ, or 0.01 on TSLA). Right, I apologize if I confused anyone with overloaded terms, but I think it's appropriate to expect the average Hacker News reader to correctly parse "tick rate" in the context of telecommunications and networking rather than confuse it with the trading concept of "tick size." [1] https://www.cs.cmu.edu/~softagents/papers/Emarket.pdf https://www.cs.cmu.edu/~softagents/papers/Emarket.pdf
    - growse 3y ago
      > > And in which order would orders be executed when they come in and get queued? > That's easy! Come time to match and fill for orders placed at identical prices, there's no reason a random number generator couldn't do the trick. Trading second-by-second need not be reliably profitable in the extremely short term to ensure an orderly and efficient market. The whole point of a static tick rate is to disincentivize such myopic trading in the first place! Adding deliberate randomness to the market makes it neither more orderly nor more efficient. And you're confusing "myopic" trading with market making. Buying now and selling 10 minutes from now brings together two other participants who want to trade but not at the same time. That seller may have needed to have accepted a worse price in order to execute when they needed to. But if you just look at the market maker, they briefly held onto something they didn't need for 10 minutes. What's the point!
      - AlbertCory 3y ago
        You are intentionally exaggerating. He didn't say to add "deliberate randomness" and he didn't say "10 minutes." A one microsecond or one millisecond time window would serve all legitimate trading purposes.
  - AlbertCory 3y ago
    Someone else here said "random" for choosing which of two or more identical trades to execute. As it is now, it's "first one to arrive, measured in nanosecond time." I think the SEC decreeing "random" for trades within some time window (is the window of millisecond size? microsecond size? not sure) would be totally legit, and it would redirect financial resources to better trades, not faster ones.
  - Ekaros 3y ago
    Disallowing fill or kill(or deprioritising) could solve this. So any sell order is split between buy orders matching.
- Enginerrrd 3y ago
  I don't think this would really get rid of HFT. You'd still have price discovery and a fight to reduce latency happening on some HFT time scale before each tick.
  - cosmojg 3y ago
    And that's fine. Most of that activity would be happening off-exchange, meaning less market noise, less price volatility, and less need to consume limited resources like radio spectrum and New York City real estate.
hhs 3y ago
> But on the flip side, nobody has ever really demonstrated strong negative effects from HFT and many studies show positive effects If interested, this is a paper published in January 2023 called "Does high-frequency trading actually improve market liquidity? A comparative study for selected models and measures" [0]. And the abstract notes: "By comparing a variety of regression models that associate various measures of market liquidity with measures of high-frequency activity on the same dataset, we find that for some models the increase in high-frequency activity improves market liquidity, but for others, we get the opposite effect. We indicate that this ambiguity does not depend only on the stock market or the data period, but also on the used HFT measure: the increase of high-frequency orders leads to lower market liquidity whereas the increase in high-frequency trades improves liquidity. We hypothesize that the observed decrease in market liquidity associated with an increasing level of high-frequency orders is caused by a rise in quote volatility." [0] [0]: https://www.sciencedirect.com/science/article/pii/S0275531922002586 https://www.sciencedirect.com/science/article/pii/S027553192...
cosmojg 3y ago
The stock market provides an efficient medium for the allocation of capital. This means that corporations can more easily and safely raise capital from the public and that the public can participate in the positive-sum growth of publicly traded companies and the economy at large. Moreover, the stock market, like any market efficient enough to be worth its salt, provides a uniquely effective information aggregation mechanism via price discovery, facilitating more accurate and more precise forecasting and prediction for nearly any process which interacts with the greater economy. Studies of high-frequency trading have yet to show benefits to these processes above and beyond those provided by a slower static tick rate, and in fact, more recent studies seem to suggest that HFT may even exist only to the detriment of these processes, all while wasting undue resources, both physical and intellectual, on an endless arms race over ever decreasing trading latency[1]. [1] https://en.wikipedia.org/wiki/High-frequency_trading https://en.wikipedia.org/wiki/High-frequency_trading
- j-pb 3y ago
  > The stock market provides an efficient medium for the allocation of capital. Only if P=NP.[1] Besides Efficiency: the ratio of the useful work performed by a machine or in a process to the total energy expended or heat taken in. This also begs the question: "To efficiently allocate capital (energy) for what goal (work)?" And it doesn't look like the answer to that is "the betterment of humanity", but to push the bell curve of wealth as far to the left as possible with the longest tail possible, externalities be damned. The stock market isn't a usefull tool for society like kickstarter, but a wealth extraction mechanism for the wealthiest. So I'd argue that HFT isn't any more immoral than T. 1: https://arxiv.org/abs/1002.2284 https://arxiv.org/abs/1002.2284
  - Dylan16807 3y ago
    > Efficiency: the ratio of the useful work And then you merely link a proof that the efficiency is not 100%. That's effectively meaningless.
    - consilient 3y ago
shw1n 3y ago
I thought they did find negative effects? I’m not necessarily for or against but didn’t the creation of dark pools lead to front-running and worse prices for customers? Michael Lewis covered this beautifully in Flashboys
- denotational 3y ago
  Can you explain how "dark pools lead to front-running and worse prices for customers"? I would agree that the price discovery works best when liquidity is lit up on a single CLOB, but claims about HFT leading to front-running or worse prices for customers (I assume you mean retail investors?) need justification. What do you mean when you say "front-running"? It has a fairly specific definition, but it is often used inaccurately in discussions around HFT. If by front-running you mean "HFT firms accept instructions as agent and then deal as principle prior to executing their client's instructions so as to benefit from the price move their client's order may induce" then you would be using the term correctly, and the SEC would be very interested, but this isn't what HFT firms do.
  - AlbertCory 3y ago
    That's my (naive) understanding of front-running. But I'm not familiar with the argument that HFT facilitates it. The best argument is that HFT serves no legitimate purpose, and is just a pointless arms race. Money should flow to better trades, not faster ones.
    - denotational 3y ago
      Usually when people claim that HFT strategies “front-run” they are talking about aggressor/taker strategies that are trying to shave off every nanosecond of latency to increase the chance of a fill at the best possible price. Because these strategies are actively racing other participants for a fill, some people (inaccurately) call this “front-running”, presumably because they are “running in front” of other participants, but this isn’t what is usually meant by front-running. As another commented has pointed out, front-running (in the sense I described it in my previous comment) is explicitly prohibited by regulation in the US, so to misuse the term in this way is to accuse someone of acting illegally without actually understanding what the terms mean.
tmn 3y ago
It’s my naive understanding that HFT makes money at the direct expense of other market participants. Is this not a downside? I’m genuinely curious to the arguments against this.
- qeternity 3y ago
  You can’t state it’s at the expense without stating how.
  - Ekaros 3y ago
    They make money/"wealth". That must come from either the seller or buyer or both. Unless existence of HFT directly leads to higher appreciation of assets traded.
    - qeternity 3y ago
      All trading is zero sum. If I own shares in a company for 20 years, that means someone else who sold them to me doesn’t own those shares and doesn’t earn those returns.
    - rcxdude 3y ago
      In the absence of a market maker (human or algorithmic), a much larger fraction of the value of a transaction is effectively randomly distributed between the buyer and seller. And it turns out buyers and sellers don't really like this much, and are happy to pay a small fraction of the transaction for a more predictable outcome.
  - bobbylarrybobby 3y ago
    Even if you're right about the market, HFTs will make you a little less right for no reason other than that they've invested billions of dollars into being faster than you.
    - lotsofpulp 3y ago
      And they will give you a little more liquidity.
BurningFrog 3y ago
The clear negative effect is the many billions spent on shaving communication milli- and microseconds to win zero sum games.
- tomatotomato37 3y ago
  [dead]
MarkMarine 3y ago
Bernie has a great proposal to tax these transactions, and fund some wonderful social programs with the money.
- PoignardAzur 3y ago
  > and fund some wonderful social programs with the money. That's silly. The whole point is they're a zero-sum game with semiconductor-thin margins. Any meaningful taxation will make them evaporate. That doesn't mean they shouldn't be taxed! But it would be a pigovian tax, the kind that serves as a soft interdiction and not as government income.

by 542458 3y ago

stefan_ 3y ago

The point is that whatever function you think HFT serves can be done just as well at fixed tick rates. This whole latency optimization mini game is just a pure waste.

super256 3y ago
And in which order would orders be executed when they come in and get queued? A delay wouldn't fix anything because the queue would still be FIFO. The only way to remove HTF would be removing the trading session, and instead only offering the market auction. Btw, a "tick" in trading is the minimum up/down movement of an asset (e.g. 0.25 points on NQ, or 0.01 on TSLA).
- denotational 3y ago
  > Btw, a "tick" in trading is the minimum up/down movement of an asset (e.g. 0.25 points on NQ, or 0.01 on TSLA). This is true, but the term is overloaded: if I say "the exchange slid the price one tick to comply with Reg. NMS" then I'm using it in the sense you mean; if I say "the model recomputes on each tick" then I'm using it in the sense of the GP.
- elictronic 3y ago
  Randomize the queue.
  - growse 3y ago
    That'll make everything more expensive, as every trade now has to price in the risk that it won't actually happen. Increasing market friction is not a good idea.
    - cosmojg 3y ago
      Given that randomization need only apply to orders placed simultaneously at identical prices, I don't see how it would increase market frictions enough to significantly limit their utility in price discovery, long-term investing, and the raising of capital. Individuals and corporations aren't making tradable decisions at subsecond intervals.
      - growse 3y ago
        By simultaneously, I assume you mean "within one time interval". I suspect you'd be surprised how many orders at the same price can happen in a very very short time window. As for price discovery, a lot of information can happen in one second. Quantizing just adds risk, which adds cost, which harms everyone. > Individuals and corporations aren't making tradable decisions at subsecond intervals. No, but they're benefitting significantly from the fact that the cost of trading is now minuscule compared to what it used to be.
        cosmojg 3y ago
        > Quantizing just adds risk, which adds cost, which harms everyone. No, it adds risk at the scale of the quantum, adding cost for those trading at the scale of the quantum, harming a very small subset of traders who would normally benefit from trading at or below the scale of the quantum, like HFT firms and short-term market makers. Should such institutions cease to exist because the cost of trading at such scales becomes prohibitively expensive, I highly doubt that society will suffer.
        growse 3y ago
        We already have evidence from history (where trading latencies and intervals were much higher) that the spreads become much higher as trading frequency goes down. The spread represents the amount of uncertainty and risk associated with a price. Every bip away from the midpoint your bid/ask sits is pure cost to the participant, and that's true for all participants. Even if I only come to market once a year to splurge my bonus on GME, I'm worse off paying 20c over the midpoint than I am paying 1c.
- cosmojg 3y ago
  > And in which order would orders be executed when they come in and get queued? That's easy! Come time to match and fill for orders placed at identical prices, there's no reason a random number generator couldn't do the trick. Trading second-by-second need not be reliably profitable in the extremely short term to ensure an orderly and efficient market. The whole point of a static tick rate is to disincentivize such myopic trading in the first place! I agree, though, that the logical conclusion of all this would be to replace the trading session with a market auction. However, while I appreciate the superior allocative efficiency of a Vickrey-style sealed-bid second-price multi-unit double auction[1] as much as the next guy, I have more faith in iterative improvements coming to pass before revolutionary ones. > Btw, a "tick" in trading is the minimum up/down movement of an asset (e.g. 0.25 points on NQ, or 0.01 on TSLA). Right, I apologize if I confused anyone with overloaded terms, but I think it's appropriate to expect the average Hacker News reader to correctly parse "tick rate" in the context of telecommunications and networking rather than confuse it with the trading concept of "tick size." [1] https://www.cs.cmu.edu/~softagents/papers/Emarket.pdf https://www.cs.cmu.edu/~softagents/papers/Emarket.pdf
  - growse 3y ago
    > > And in which order would orders be executed when they come in and get queued? > That's easy! Come time to match and fill for orders placed at identical prices, there's no reason a random number generator couldn't do the trick. Trading second-by-second need not be reliably profitable in the extremely short term to ensure an orderly and efficient market. The whole point of a static tick rate is to disincentivize such myopic trading in the first place! Adding deliberate randomness to the market makes it neither more orderly nor more efficient. And you're confusing "myopic" trading with market making. Buying now and selling 10 minutes from now brings together two other participants who want to trade but not at the same time. That seller may have needed to have accepted a worse price in order to execute when they needed to. But if you just look at the market maker, they briefly held onto something they didn't need for 10 minutes. What's the point!
    - AlbertCory 3y ago
      You are intentionally exaggerating. He didn't say to add "deliberate randomness" and he didn't say "10 minutes." A one microsecond or one millisecond time window would serve all legitimate trading purposes.
- AlbertCory 3y ago
  Someone else here said "random" for choosing which of two or more identical trades to execute. As it is now, it's "first one to arrive, measured in nanosecond time." I think the SEC decreeing "random" for trades within some time window (is the window of millisecond size? microsecond size? not sure) would be totally legit, and it would redirect financial resources to better trades, not faster ones.
- Ekaros 3y ago
  Disallowing fill or kill(or deprioritising) could solve this. So any sell order is split between buy orders matching.
Enginerrrd 3y ago
I don't think this would really get rid of HFT. You'd still have price discovery and a fight to reduce latency happening on some HFT time scale before each tick.
- cosmojg 3y ago
  And that's fine. Most of that activity would be happening off-exchange, meaning less market noise, less price volatility, and less need to consume limited resources like radio spectrum and New York City real estate.

hhs 3y ago

> But on the flip side, nobody has ever really demonstrated strong negative effects from HFT and many studies show positive effects If interested, this is a paper published in January 2023 called "Does high-frequency trading actually improve market liquidity? A comparative study for selected models and measures" [0]. And the abstract notes: "By comparing a variety of regression models that associate various measures of market liquidity with measures of high-frequency activity on the same dataset, we find that for some models the increase in high-frequency activity improves market liquidity, but for others, we get the opposite effect. We indicate that this ambiguity does not depend only on the stock market or the data period, but also on the used HFT measure: the increase of high-frequency orders leads to lower market liquidity whereas the increase in high-frequency trades improves liquidity. We hypothesize that the observed decrease in market liquidity associated with an increasing level of high-frequency orders is caused by a rise in quote volatility." [0] [0]: https://www.sciencedirect.com/science/article/pii/S0275531922002586 https://www.sciencedirect.com/science/article/pii/S027553192...

cosmojg 3y ago

The stock market provides an efficient medium for the allocation of capital. This means that corporations can more easily and safely raise capital from the public and that the public can participate in the positive-sum growth of publicly traded companies and the economy at large. Moreover, the stock market, like any market efficient enough to be worth its salt, provides a uniquely effective information aggregation mechanism via price discovery, facilitating more accurate and more precise forecasting and prediction for nearly any process which interacts with the greater economy. Studies of high-frequency trading have yet to show benefits to these processes above and beyond those provided by a slower static tick rate, and in fact, more recent studies seem to suggest that HFT may even exist only to the detriment of these processes, all while wasting undue resources, both physical and intellectual, on an endless arms race over ever decreasing trading latency[1]. [1] https://en.wikipedia.org/wiki/High-frequency_trading https://en.wikipedia.org/wiki/High-frequency_trading

j-pb 3y ago
> The stock market provides an efficient medium for the allocation of capital. Only if P=NP.[1] Besides Efficiency: the ratio of the useful work performed by a machine or in a process to the total energy expended or heat taken in. This also begs the question: "To efficiently allocate capital (energy) for what goal (work)?" And it doesn't look like the answer to that is "the betterment of humanity", but to push the bell curve of wealth as far to the left as possible with the longest tail possible, externalities be damned. The stock market isn't a usefull tool for society like kickstarter, but a wealth extraction mechanism for the wealthiest. So I'd argue that HFT isn't any more immoral than T. 1: https://arxiv.org/abs/1002.2284 https://arxiv.org/abs/1002.2284
- Dylan16807 3y ago
  > Efficiency: the ratio of the useful work And then you merely link a proof that the efficiency is not 100%. That's effectively meaningless.
  - consilient 3y ago

shw1n 3y ago

I thought they did find negative effects? I’m not necessarily for or against but didn’t the creation of dark pools lead to front-running and worse prices for customers? Michael Lewis covered this beautifully in Flashboys

denotational 3y ago
Can you explain how "dark pools lead to front-running and worse prices for customers"? I would agree that the price discovery works best when liquidity is lit up on a single CLOB, but claims about HFT leading to front-running or worse prices for customers (I assume you mean retail investors?) need justification. What do you mean when you say "front-running"? It has a fairly specific definition, but it is often used inaccurately in discussions around HFT. If by front-running you mean "HFT firms accept instructions as agent and then deal as principle prior to executing their client's instructions so as to benefit from the price move their client's order may induce" then you would be using the term correctly, and the SEC would be very interested, but this isn't what HFT firms do.
- AlbertCory 3y ago
  That's my (naive) understanding of front-running. But I'm not familiar with the argument that HFT facilitates it. The best argument is that HFT serves no legitimate purpose, and is just a pointless arms race. Money should flow to better trades, not faster ones.
  - denotational 3y ago
    Usually when people claim that HFT strategies “front-run” they are talking about aggressor/taker strategies that are trying to shave off every nanosecond of latency to increase the chance of a fill at the best possible price. Because these strategies are actively racing other participants for a fill, some people (inaccurately) call this “front-running”, presumably because they are “running in front” of other participants, but this isn’t what is usually meant by front-running. As another commented has pointed out, front-running (in the sense I described it in my previous comment) is explicitly prohibited by regulation in the US, so to misuse the term in this way is to accuse someone of acting illegally without actually understanding what the terms mean.

tmn 3y ago

It’s my naive understanding that HFT makes money at the direct expense of other market participants. Is this not a downside? I’m genuinely curious to the arguments against this.

qeternity 3y ago
You can’t state it’s at the expense without stating how.
- Ekaros 3y ago
  They make money/"wealth". That must come from either the seller or buyer or both. Unless existence of HFT directly leads to higher appreciation of assets traded.
  - qeternity 3y ago
    All trading is zero sum. If I own shares in a company for 20 years, that means someone else who sold them to me doesn’t own those shares and doesn’t earn those returns.
  - rcxdude 3y ago
    In the absence of a market maker (human or algorithmic), a much larger fraction of the value of a transaction is effectively randomly distributed between the buyer and seller. And it turns out buyers and sellers don't really like this much, and are happy to pay a small fraction of the transaction for a more predictable outcome.
- bobbylarrybobby 3y ago
  Even if you're right about the market, HFTs will make you a little less right for no reason other than that they've invested billions of dollars into being faster than you.
  - lotsofpulp 3y ago
    And they will give you a little more liquidity.

BurningFrog 3y ago

The clear negative effect is the many billions spent on shaving communication milli- and microseconds to win zero sum games.

tomatotomato37 3y ago
[dead]

MarkMarine 3y ago

Bernie has a great proposal to tax these transactions, and fund some wonderful social programs with the money.

PoignardAzur 3y ago
> and fund some wonderful social programs with the money. That's silly. The whole point is they're a zero-sum game with semiconductor-thin margins. Any meaningful taxation will make them evaporate. That doesn't mean they shouldn't be taxed! But it would be a pigovian tax, the kind that serves as a soft interdiction and not as government income.