> I would assume TXSE would want quick access to that order flow
Perhaps Texas could use a different trading model that doesn't require ultra high speed trading.
Matt Levine often mulls the idea of a system with a trading window that doesn't let the fastest connection to the order book win. Perhaps an order book that works at human speeds so humans can trade too (I can think of a few ways to do it - but would need modelling to try and figure what actually works). He points out that most trades are done in the last hour, so really trading only needs to occur once a day.
The issue is whether a market trading system can be designed with suitable restrictions that beats the current market design (for listed companies and for traders).
Designing markets is hard because you have to assume every player is selfish and only cooperates where it is to their benefit and will defect or cheat if the incentives of the market encourage that (Enron in the California energy markets).
Unlikely since SEC would need to approve of a different system of market trade incentives.
Edit: Personally I would like to see an exchange that was more international. I'm from New Zealand and our good businesses often list on the Australian exchange rather than the NZSX. The system of ADRs for other countries feels like a massive hack.
Reg NMS’s Order Protection Rule (Rule 611) says you can’t trade through protected NBBO quotes, outside a few narrow exceptions. That’s the letter of the law.
The practical effect isn’t just a bit of latency. It rewires incentives. With 611 in place, the question for latency-sensitive firms becomes: what HFT tactics can I run that are 611-compatible? Without 611, the question would be: what HFT tactics actually add value for my counterparties? That’s a very different optimization.
For firms on direct feeds (often building their own synthetic NBBO), 611 doesn’t add much information. The constraint is compliance, not discovery.
Because NBBO is size-agnostic and top-of-book, anchoring execution to it lets micro-lot quotes steer outcomes. You can influence the protected price with tiny displayed size. That’s great for gamesmanship, bad for displayed depth, size-sensitive pricing, and near-touch discovery.
Also: if two informed counterparties want to trade away from the protected price to reflect size or information, 611 mostly blocks that outside limited carve-outs. We lose mutually beneficial, size-aware prints to satisfy a benchmark that ignores size.
On settlement, the uniform benchmark helps in calm markets. But it’s naïve to think that holds through a real black swan. In stress, timestamp ambiguities and fragmented data make “what was executable” contestable, and disputes spike regardless of quote protection.
In a sound market structure, the clearer (CCP or clearing broker) should carry and underwrite that tail risk—margin, default funds, capital, and enforceable rulebooks. Instead, 611 shifts accountability onto quote-protection mechanics, insulating clearers from responsibility and, perversely, amplifying systemic risk when the system most needs well-capitalized risk absorbers.
There is no reason why shares should be bought on sold in time frames far too short for anything to have meaningfully changed about the companies or market conditions.
This is the free market speaking. If there was one exchange there would be ~no~ less latency arbitrage. But there are many... Which creates a competitive landscape and reduces fees for investors. The by product is you have many HFTs that come in to take advantage of mispricings, even if they are on sub millisecond scales. It doesn't harm the company or the investor. Its quite the opposite... Investors benefit from competition amongst exchanges and HFTs.
In addition you have redundancy in the markets system. Exchanges are important for national security... Having everything centralized would risk people's retirements, savings, and more
One other aspect of HFT that is good for the general investor is that HFT injects liquidity, making it easier for a general investor to liquidate their position, which is a desirable thing for human traders. HFT does not magically make human investors engage in more or less speculative behavior.
HFT is an easy thing to attack, but I've never encountered a lucid argument for why it's bad. "It's not fair that I'm not as fast" isn't really a reason unless you explain why removing liquidity (i.e., making it harder for you to find a buyer at your price point), paired with you moving up the "trading swiftness" rankings, is preferable.
the common argument against it is that it guarantees a technological arms race and by those conditions pushes the smaller groups out of the competition.
it's unfair in the same vein that the rich are always offered better loan rates than the poor. Yeah, it's obvious why that would be, but it's not fair either.
although imo pushing small-backer arbitrage out of the equation is a good thing.
What HFT arbitrage does is rapidly push markets with different prices into alignment with each other. If you banned HFT but kept multiple markets, the inevitable result would be markets with bigger differences in prices for longer.
The only kind of trading that really good HFT trading pushes out is other slower less efficient arbitrage traders, but why should we want more worse arbitrage traders if the result is markets being more out of sync?
What's important economically is that traders that trade based on fundamentals can do so efficiently across multiple markets. Efficient HFT arbitrage trading helps that, it doesn't hinder it.
> the common argument against it is that it guarantees a technological arms race and by those conditions pushes the smaller groups out of the competition.
But a cursory examination of history would reveal that the literal opposite has happened.
The issue isn't that there's a lot of change coming from inside the markets, it's that there's a lot of change coming from outside the market, and it's all interconnected.
The behavior of other participants is itself a first-class signal.
Rule 611 compresses that signal. By forcing everything to orbit a size-agnostic NBBO, it collapses a lot of the “behavioral bandwidth” (depth, imbalance, sweep patterns, replenishment, cancel/replace cadence) into a single top-of-book tick. Less resolution, less information.
High-resolution flow tells you who wants what, at what size, and how urgently. When we gate execution through protected quotes, we encourage tactics that flick the top-of-book with tiny size and discourage truthful size revelation. That’s signal destruction dressed up as protection.
Letting informed counterparties print away from the protected price (to reflect size or information) increases informational content. You get cleaner read-through from actual willingness to trade, instead of a compliance-driven dance around a fragile benchmark.
So yes: other people’s actions are the best data feed. The more of that behavior we can see—in size, time, and venue—the better our discovery gets. 611 reduces that visibility by design.
If HFT was genuinely good for the entire market than absolute latency would be what matters but it is only relative latency between HFT firms that matter because they are all competing against each other using the same tactics where whoever is fastest wins.
All participants contribute to price discovery. A nanosecond order book helps with price discovery the deeper it is, regardless of whether orders clear.
> The better the computers hooked directly into the exchange get you mean.
I think you're trolling with this one. But you had an advantage typing your comment into a web browser compared to all the people who wrote theirs on paper and put it in an envelope with a stamp.
A nanosecond order book helps with price discovery the deeper it is, regardless of whether orders clear.
This is a claim, it is not being backed up by evidence.
I think you're trolling with this one. But you had an advantage typing your comment into a web browser compared to all the people who wrote theirs on paper and put it in an envelope with a stamp.
This analogy doesn't make any sense. Why would a person care about nanosecond price discovery? The only benefit is for whoever controls the computers that are able to do it and profit off of it.
If that's not true then why are these firms paying so much money to have nanosecond advantages?
Why do people doing normal trading want to avoid the exchanges that have HFT computers skimming money off their trades?
There is no mutual benefit here. If there was you would be able to explain it clearly and with evidence instead of just making claims about 'price discovery'.
The price is going to get discovered either way just as it has for hundreds of years, it happening a billion times per second does normal traders no good.
> Why do people doing normal trading want to avoid the exchanges that have HFT computers skimming money off their trades?
What's "normal trading"? A prop desk at an investment bank? A hedge fund? A pension fund? Someone who's just installed Robin Hood on their phone?
Most rational participants want lower trading costs and overheads, smaller spreads etc. HFT provides that - the evidence being "look at what the spreads are today, compared with what they were pre computerised trading".
> The price is going to get discovered either way just as it has for hundreds of years, it happening a billion times per second does normal traders no good.
Could you explain how a market participant who makes one trade a day is negatively impacted by high resolution price discovery?
> Could you explain how a market participant who makes one trade a day is negatively impacted by high resolution price discovery?
They’re negatively impacted because the SNR drops when the gain is turned up and causes algorithmic ringing to perturb the price that would be generated by high quality signal.
People who extract money by injecting algorithmic ringing into financial markets are a social parasite.
> The better the computers hooked directly into the exchange get you mean.
I need you to understand that HFT makes decisions based on human-defined parameters. It's not AI-driven. What's the difference between a human saying "these are my parameters, now CPU, go trade based off those" versus "these are my parameters, now underling, go trade based off those"
the only difference is speed, plus i suppose those underlings might suck at following their boss's directives compared to a computer
> The better the computers hooked directly into the exchange get you mean.
This implies a distinction between "our" (i.e., humans) and "the computers." Can you explain what distinction you meant if it wasn't AI? After all, everything computers do outside AI is done pursuant to knowable, describable human control, no different from pressing keys on a keyboard.
So if you didn't mean AI, I think it's worse for your argument, not better.
The difference is that I will actually explain what I mean instead of just making claims with no evidence.
Can you explain what distinction you meant
The distinction is that the computers making millions of trades per second are owned by few people and have a huge advantage. They don't lose money and they don't hold anything.
They aren't wanted by actual people making decisions, they are there in spite of what traders buying stocks to own them actually want. They are there because they make money and from the exchange and make money for the exchange.
Retail traders are the product even though they don't want to be.
It's pretty funny that there are people in this thread pretending like "price discovery" is a real thing that happens in markets based on information. We've all seen Dogecoin and BBBYQ. The emperor has no clothes.
We could make the distinction between price discovery, i.e. what price are people currently willing to buy and sell at (short-term) vs value discovery (long-term).
When I press the buy and sell button, I want the transaction to happen as quickly as possible. So does everyone else.
My millionth of a second is different than yours, and everyone else’s.
It is no different than buying or selling anything else. And there is no loss from the additional liquidity, you can easily set a limit at which you want to buy or sell.
The only way to implement this is to eliminate competition between exchanges.
There are two different things being talked about here.
Trading based on arbitrage between exchanges will happen in one way or another no matter what.
Trading millions of times per second automatically on the same exchange when some people have low latency computers at the exchange with huge amounts of extra information is not necessary.
Also, Wall Street would love this. The more of the order book you submitted, the more information you have about its composition.
The point isn't to make something 'wall street hates' it's to make something that doesn't get money eaten away by automated computers in the middle so that it's the best option for people making trading decisions on people time scales.
> The point isn't to make something 'wall street hates' it's to make something that doesn't get money eaten away by automated computers in the middle so that it's the best option for people making trading decisions on people time scales.
Why is this desirable? It seems like an argument designed only to serve the interests of a small class of person who insists on doing manual trades themselves.
The rest of what you've written just sounds like "I lost money because computers are better than me at the task." I'm not sympathetic to that concern. Computers are better than me at lots of things, so I just don't try to compete at those things. I pay people with access to the computers to do them for me, and then I focus on the things I'm good at instead. Division of labor and all that.
Anyone who isn't involved in HFT should be in favor of rules that slow down trades to human time scales. HFT currently favors a small class of rich people.
> Anyone who isn't involved in HFT should be in favor of rules that slow down trades to human time scales
What are you basing this on?
I’m a former algorithmic market maker. Every plan to “slow down trades to human time scales” I’ve seen were trivially gameable. They were always proposed by a group of concerned citizens, and then jumped on by my bosses, because if the market is slowed down to pre-HFT speeds, Wall Street can make pre-HFT profits on risk-free trading again.
Do you think the internet would work better if we forcibly increased latency? If we did, if the argument were this would flatten the market and better let small websites compete with CDNs, do you think that would actually happen? Google and Cloudflare would say “oh well,” and disassemble their servers?
Our markets have structural problems. They are mostly solvable. HFT is none of them, which is why you keep hearing about it from folks who don’t want reform.
> "How does HFT provide social benefit to the world at large?"
It's explained multiple times in this very discussion. It's not our fault if you refuse to read them. But the most straightforward and obvious way is that it injects liquidity into the market, making it easier to sell when you need to liquidate. (It also reduces volatility overall, another good thing.)
> > the exchanges weren't established for the abstract sake of money, they were established to provide benefit to people.
Snort. The exchanges were established by wealthy people, for wealthy people, to engage in business with other wealthy people.
The NYSE was established in 1792. Is it your contention that anyone except the elites were buying and selling stocks in 1792? Let alone all the exchanges that pre-date the NYSE. The Amsterdam Stock Exchange was set up in the 1600s specifically to facilitate the buying and selling of Dutch East India Company shares. Was Farmer Aardhuis buying shares? Or aristocrats and royalty?
I suspect the root argument is really against the efficacy of markets and capitalism as a useful system for humanity, in which case I say that is a fair debate. The benefits are hardly obvious today.
Retail investors railing against HFTs are sort of like those San Francisco types who protest new development to the benefit of their landlords.
It's nothing like that since there isn't a limited resource and everyone has access to the core purpose, which is to trade stocks.
What I notice with these discussions is that no one can actually explain why a retail investor or anyone would want computers trading underneath them millions of times a second.
At best they try to give hft credit for the automation that happened with computers anyway.
The only people that want it are the people doing it. That's not a business, that's a grift.
> not me. I don't mind if it takes like 20 seconds or so
Which is fine! You can probably find a broker who will give you fee-free trading with that preference. The price you execute at won’t be as good. But unless you’re trading millions, that’s probably fine.
Do you only want us to get our news once daily via a physically printed and distributed newspaper? That’s the timeframe that we used to use for news updates. If not (I.e., you’re in favor of keeping the Internet), how would you reconcile the asymmetry between trading and news updates? If there is a disparity between these timescales, you end up with markets gapping hugely every time they open. This just increases risk and volatility. Yes, markets do this overnight today, but 24-hour markets don’t do this as much and they allow a trader to set stop orders that are active overnight to protect positions. So, for instance, futures markets will gap over a weekend break, but because they trade 23 hours per day during the week, they are much more smooth than they would otherwise be. Compare gaps in ES SPX futures contracts vs. gaps in the SPX itself, for instance. In general, smoother is better for everyone.
Humans do still manually trade. The existence of HFTs doesn't prevent or interfere with that, in fact it makes it easier by ensuring that prices are more likely to be converged to a consensus market price at any given millisecond.
Thinks of it like this. When I put in a trade at human speeds based on business fundamentals, I'm not looking that the millisecond by millisecond prices, I just put in a price I am willing to accept and if the market reaches that price I get execution. HFT makes that easier and more efficient across markets by ensuring prices are converged rapidly.
How do you think it makes it harder or worse? If I put in an order to buy at $x on a particular market because I think the stock is worth more than that for business reasons, what is it about the existence of HFT that is a problem for me?
Disagree? You think milliseconds is “better” somehow?
Then by that logic microseconds are better still! (A straight-faced argument made by thousands of HFT people.)
Then, surely, nanoseconds matter. Again, some traders care deeply about shaving single digit “nanos” off their response times by using smart NICs that can respond before the incoming packet has even finished arriving! Bypassing the CPU entirely because ermahgerd that would waste precious nanos!
Okay, what about femtoseconds? Attoseconds? Low single digit Plank time units?
Clearly the extrapolation is nonsense.
The problem is that there’s always an advantage to some rent-seeker to be faster than everyone else, so there will never be consensus between them and the general public. Or each other.
It’s a classic tragedy of the commons.
This is why laws are required, to prevent that one greedy guy putting “just one more cow” onto the pasture than the other greedy guys.
This is the “speculation is bad” take with a side of naïveté about how markets work.
Open an order book. Prices and quantities aren’t decoration; they’re live telemetry for supply, demand, and how tight the crowd’s consensus is at each level. That’s information, full stop.
A human (or machine) trader forms a view of fair value against that tape. The book helps decide how to trade—size, urgency, venue—regardless of motive: arbitrage, hedge, speculation, investment, cash-out. Intent doesn’t change the math.
Prints are messages. Every execution updates everyone else’s priors. More prints → more information → smoother discovery.
Make the book sparse—only a handful of trades per day—and watch confidence collapse. With weaker consensus and wider error bars, people step back. Liquidity thins, friction rises. That’s not morality; that’s microstructure.
Time horizon doesn’t invalidate the signal. A strategy that unfolds over days and one that resolves in milliseconds both add to the dataset. If it trades, it teaches. More resolution in others’ behavior means better prices and deeper books. That’s the game.
Millisecond trading strategies have zero relationship to information about companies or economic fundamentals and are therfore not economically productive. They're exploiting microstructure inefficiencies like latency arbitrage, order front-running, not price discovery. That's not 'teaching' the market it's a tax on everyone else's execution. The fact that it's legal doesn't make it structurally useful. It is just a result of the rules not being updated when trading became automated at superhuman speeds
> Millisecond trading strategies have zero relationship to information about companies or economic fundamentals and are therfore not economically productive
It is not clear to me that the only economically productive information is "information about companies or economic fundamentals."
If I know some idiot is willing to pay 100x what a company is actually worth, that is economically productive because it gives me, someone better with money, a ton of resources that formerly were controlled by someone who didn't know how to leverage the assets in an economically productive manner. IT's the same argument as allowing adverse possession: transfer of assets from non-productive owners to productive owners, benefiting society as a whole.
With this, I've established a third kind of data beyond "economic fundamentals" and "information about companies."
Right. “Worth” and “price” are relative notions that are only useful when paired with a point of view. What’s your house “worth?” You might have a number in your head, but unless you have a buyer willing to transact at that price, your house isn’t worth what you think it is. And this relates to economically productive information. Your house might also be worth more if one knows that there is gold or oil buried underneath. It might make more economic sense to bulldoze the house and start mining or drilling.
The boundary is "arbitrary", but clearly there has to be one, otherwise we're all nodding in agreement as trading speed heads inexorably towards Plank time units.
We have some convenient "lines in the sand" that we can use as a guide:
- To make trading fair globally, the round-trip time for light around the planet could be multiplied a couple of times. That's about a second.
- The fastest possible time a human can parse the meaning of a long headline (not the full news article!) is... about a second.
- Nobody in their right mind should be buying any significant volume of shares without double-checking their order. There's no way to do this even vaguely carefully in under... a second.
Etc...
Bots trading faster than a second are trading with each other, and the only signals they have are each other.
Humans are what markets are for, not bots.
This reminds me of a story from WWII where a bunch of generals took a holiday at the same time, leaving a junior general in charge. He was in a bit of a panic because he was expecting to be overloaded with work... but found it easy. The generals were making work for each other by requesting reports, organising meetings with each other, etc...
Bots make work for bots, they generate signals for bots ever faster, to be processed by faster bots still, etc...
It's just... nonsense. Zero real information is being generated, they're just "riffing" off of the much less frequent human-initiated trades, all of which take minutes to organise and execute with due diligence.
It's like being asked to write a 10-page essay on a three-line poem.
And you keep incessantly posting about the moral evils of speculation which is only tangentially related to the Texas Stock Exchange. I guess morality policing makes your comments better?
It's a tradeoff. Suppose you are relegated to trading once per minute and you need to allocate $x per trade to make a profit. If you can trade 100 times per minute you need $x/100 to make a profit.
Or, maybe your strategy doesn't make money when you can trade once per minute, but does at a higher frequency.
What is fair about saying someone has to have 100x the capital to participate? You are focusing only on one dimension of cost, when there are several.
The only non-arbitrary boundary is the one set by physics. Everything else is arbitrary. There are ways to greatly reduce front-running of orders, but that actually is an orthogonal issue to HFT and microstructure level trading.
hah someone finally said it. The financial market or whatever the f it's called has become a Monty Python sketch. Like those "pro" StarCraft gamers that keep randomly clicking their mouse for no reason except to keep their APM counter high.
It's exhausting to see good reasons offered and then y'all proudly failing to understand those reasons. Like, there are literally multiple comments here explaining it.
If I have a condition that is resolved by standing upside down on my head, and I invent a 100 ways to balance my body on my head and explain the reasons for each of them, it doesn't mean that standing on my head is a good or sensible thing to do. And none of them address the core problem that is the condition that requires me to stand on my head.
Look at why shares and the stock market were created in the place, and how many layers we laid on top of them and made the means the end and the purpose. If it takes you more than 2-3 sentences to explain something that's purely invented by humans it was probably silly to begin with.
Like the other day I was just watching a video about UTF-8. It spent a good chunk explaining all its various quirks and rules and workaround that all went back to how ASCII was dumb in some of its choices in the first place and now we're stuck with that forever.
heh I thought of that while writing the comment, but I think I covered it with "something that's purely invented by humans" which finance definitely is, unlike medicine, surgery, physics, and to a lesser degree, computing.
Except for the fact that people don’t know what the price is even with all available knowledge. HFTs serve a pretty useful function in converging to a discovered price the market can bear continuously. Prior to HFT markets were generally less stable on their pricing and liquidity was much worse. The fact you can basically trade at any time on anything instantly is made possible by HFT.
I look at them as providing a service like an energy exchange does in ensuring power distributes evenly and regularly across a region across providers and grids. They clip a fee in the middle but they provide a service in stable supply and prices.
This doesn’t mean automation isn’t without risk, and like an energy exchange, when things go badly they can go very badly. But by and large you never notice either the HFT or the energy exchange while you benefit from their existence.
> He points out that most trades are done in the last hour, so really trading only needs to occur once a day.
Presumably then the last trader has the most information, and so the game would be getting the info as late as possible and trading as late as possible, but not too late.
That's one thing that will make blockchain trading interesting as it's discreet block by block, such as the ex-dividend date who holds a stock when the dividends are paid, might make fees for that block very competitive. The SEC is really struggling now with good regulation but it's coming end of the year to draw the line in the sand.
> Perhaps Texas could use a different trading model that doesn't require ultra high speed trading.
What would that look like? Periodic auctions? Certainly it could be done, I'm just trying to understand what problem might be solved, and whether the solution would be effective.
For example, even with the opening and closing auctions we have today, there can be an advantage to getting your order accepted right before the deadline. Some participants do this, most don't really (depending on the exact definition of "right before"). But the fact that some do tells me that some participants would do the same thing with periodic auctions, and at least for them latency would still be important.
If, as seems likely, latency is fundamentally important to at least some styles of trading, how do you incentivize participants to not value it?
You take bids continuously but publish the bids and "resolve" the auction every X seconds, where X is between 5 and 10. Then there is no speed advantage as long as you can get your bid in within 5 seconds.
There is still a speed advantage. You can look at correlated markets for example and trade at the end of the time window with 5 seconds more info than anyone trading at the start of the time window.
Except with the randomness you won't know if the window is 5 seconds or 10 or anything in between. So sure you could send in your bit at 4.99 seconds but it won't matter if you're a few microseconds off.
I suspect that randomness of some sort is the only way. Without that, whatever the rules of the game, there is a way to somehow get a slim advantage. You can make the advantage small and the costs to try to gain it large, such that it isn’t cost effective to try, but as we’ve seen with HFT, it’s amazing how much people will spend to pick up pennies. Even a tiny gain, exploited frequently enough can be quite profitable.
A Poison process. Just specify the expected value. E.g., if the expected value is 5 seconds and have gone some hours without an event, then the expected time is still 5 seconds.
E.g., like the time to some radio active delay and whatever its expected value to decay is.
Or even a fixed delay. Imagine trading with one month delay - You'd have to bid what you think the stock is actually worth, not what you think everyone else will think it's worth in five milliseconds. That's extreme, of course.
Eric Ries first started talking about a Long-Term Stock Exchange, he suggested long (potentially multi-year) lock-up periods. The LTSE he actually implemented doesn't have that. I speculate that this was a compromised because they are allowing dual listings which helps them gain market share but also would undermine the entire concept of very long lock-ups.
You could force companies to provide standing orders to sell unlimited shares at a price selected when they go public, and one they're unlikely to go below.
This would happily also eliminate price speculation entirely. The price would just be whatever the price is and most returns would come from dividends. Would require a bunch of tax and regulatory changes
> Perhaps Texas could use a different trading model that doesn't require ultra high speed trading
Wall Street (as in the sell side) is strongly incentivised to stamp out high-speed trading. It undercuts their dealer model. They have tried and failed to come up with an auction model that eliminates HFT without tradeoffs that real investors find unacceptable.
I'd say IEX has done remarkably well - it's not likely to displace NASDAQ or NYSE but it has solidified its place as the #3 US exchange by any reasonable measure. If TXSE achieves comparable market share I'd call that a wild success.
You're not wrong to say that most participants don't care about what IEX offers, but enough do to make a meaningful dent in trading volume.
IEX is "3rd place" in their mind only because the other 12 exchanges are owned by two companies. IEX is 13/13 for volume, and the main reason they have any volume is because IEX sometimes has the NBBO so you have to trade there per reg NMS.
You're right, my view is way out of date. I didn't realize CBOE had grown so much in straight equities trading. IEX is the best of the rest, but it's NYSE, NASDAQ and (to my surprise) CBOE as the clear top 3.
That’s what always surprises me when folks bring this up. Nobody in the market cares. Institutional investor experience some of the lowest trading cost in history. These complaints are most often coming from retail traders where again I don’t follow the argument. Instead of some guy on the floor picking up dollars with have machines picking up pennies. This is a win for everyone.
Nobody in the market cares because everyone in the market are cheaters. There is a huge untapped potential of people outside the current market that do not participate because they know the market is rigged against them.
A fair market could be huge, but the trick is keeping it fair. It used to be more fair, and we used to have a healthier economy because of it.
It’s not like most of the unfairness in the current market couldn’t be dealt with, probably with laws already on the books. Most HFT strategies are not only blatantly dishonest but also clearly illegal. The government looks the other way though, because corruption.
Where is the citation for this 100% HFT? Lots of numbers floating around but the only time I have seen those types of numbers are for short lived periods.
My issue as always with these numbers is that back in the day you had some bloke on the floor vacuuming up this wide spreads and today we have some of the smallest spreads in history. It has a cost but that cost is generally less than what you were paying before.
Why would you create arbitrage opportunities for no reason? That’s the only thing that would happen I can see from an exchange that can’t keep up with the NBBO price, which you are obligated by law to quote regardless.
People in the finance industry will arb between digital and human markets and net a profit from it. It seems pointless to me, but perhaps I’m not fully grasping what that would do.
Also it just changes the nature of the game. There's no incentive to interact with the batch until the absolute last microsecond. It will still be dominated by latency-sensitive participants, just in a manner where the difference between visible liquidity and latent liquidity is even more diverged from reality (on average).
Bit of an aside, but I really do not understand the concerns with trading speed.
I can trade at human speed now: when I want to make a trade, I put in the order and it gets executed. Speed elsewhere in the market makes it easier, not harder, for me to trade when I want to. And I don’t care who my counterparty is; that’s a fundamental feature of a stock exchange. If A is always faster than B because A is 2 racks closer to my broker in the data center… so what? How does that hurt me? Good for A.
A computer-powered trading strategy can react faster than me to news—true. But that’s fine because I don’t have to follow a breaking-news investing strategy. There are tons of others, many of which have proven to work very well.
Because lit orders get front run. Every sophisticated participant/algo is exceptionally efficient at extracting money from less sophisticated participants.
As someone who trades decent volume but doesn't have a fully institutional grade workflow, I have the fortune of dealing with this...
Simple lit orders (posting an order directly to an exchange) will be taking advantage of by both market makers, by HFTs, and by smarter execution algorithms. The algorithms running the bids and asks will widen spreads. Sell orders will peg to one cent below your ask, and if flows start to reverse, they will pull their liquidity and the slower participants get their liquidity swept through (adverse selection).
The next step up is to use something like a midpoint algorithm or hidden order, but hidden orders will be pinged with one share from the robots and you will get sniffed out and positioned against. If they detect size in a midpoint algorithm, the liquidity in the opposite direction will evaporate, and they will "walk" the dumb midpoint algorithm down, take the liquidity, and then reset the mid back to where it was. The list goes on. It's generally an awful environment for "regular" participants.
Moving on from simple improvements available to the more advanced retail space like midpoint algorithms and VWAP algorithms, you have algo routes that are explicitly designed to take advantage of the "lesser" order types. If they are in a position to get a fair fill, they will rest the order in case they see a situation they can take advantage of, and only take mid fill if the outlook deteriorates (this is all millisecond time frame stuff, but the orders will be worked in an automated fashion throughout the day - time frame is configurable).
On the more developed institutional side, liquidity is sourced in dark venues designed to ward off HFTs and front-running, or sourced in fair flash-auctions which are again designed to ward off hfts and information leakage from the auction spawner.
So the argument would be that perhaps the modern developments like batched flash auctions should just be the new baseline, and designed so that all of the participants feeding into them get an equivalent quality of fill.
These "phenomena" are fairly significant. Let's say you have a 100k position in a smaller cap stock. You may move the stock down a few percent if you start walking down your order and it becomes clear that you are looking to take liquidity. Vs 100k in one of the more advanced order routes where you're basically going to get filled near mid. And of course it goes without saying that 100k won't even move the needle in the institutional routes.
For a while I got so sick of it that if I was looking to buy back my short options (the same things happen in the option space, but with more slippage), I would stuff a basic midpoint algorithm on the underlying, it would be sniffed out and liquidity would evaporate, price would fall, and I'd slam the ask to buy/cover my short calls on the price drop. At least I could get a fair fill when I played two different areas of the market complex against each other... It's just a pain. To the average participant, they will find that liquidity is there when it suits the counterparty, yet not there when they need it.
NBBO/best bid offer itself can be illusory. There are many situations where if you sweep the bid, you will get a fair fill, but I'd you just hit the bid price, you will essentially take off the very small front order of an iceberg order, they will run their calculations, and the liquidity pegs a cent below you if it suits them. That's how it works.
This goes for all areas of the financial market, including the bond market itself, and it contributes to systemic fragility in addition to harvesting retail money.
Granted, almost no retail participant is actually shipping orders directly to exchanges like I laid out. They are going to payment for order flow routes. These are actually fairly efficient, but again, remember that if you are posting a bid or ask, exchanges pay you (yes, you actually net money, albeit small) to post these orders, and anyone feeding into PFOF routes is getting this income taken from them. The frontrunning risk in the payment for order flow routes is also much more severe, since your order is getting blasted out in all directions before it is posted. So when those sorts of routes go wrong for retail traders (ex making the mistake of posting a large order during a major market event), they're could catastrophically get screwed.
It's also worth noting that retail does have access to a relatively Fair auction system though. Open and closing auctions are probably the best ways to fill orders. Just be careful not to ship too much size into them since a large enough net imbalance (say in a small cap stock) in a closing auction will have the same "walk down the price" effect that happens with midpoint orders.
Personally I think that the institutional flash auctions are pretty neat. For my understanding this sort of liquidity sourcing is growing. I would think that this sort of functionality could be regulated and integrated into the base level market venues.
> lit orders get front run. Every sophisticated participant/algo is exceptionally efficient at extracting money from less sophisticated participants
Anyone executing via lit orders is either forced to do so or an idiot. That’s why most of the market doesn’t execute via lit orders. Which is fine. The trade is still reported ex post facto, and the inefficiencies this creates are always less than the convoluted auction formats one must use to make low latency non-advantageous.
On net I'd agree, with the caveat that the hyper speed liquidity comes at the cost of fragility. Liquidity that can disappear in a microsecond can and does amplify market shocks.
As for retail execution, while on net there's a standard and strong argument that a less regulated market is the most efficient, retail execution is most definitely at the bottom of the totem pole, with an order shopped around to parties that can pick and choose the profitable orders before it is sent to wider liquidity pools. I think that this side of the debate is more about evening out the playing field. On net the market may get less efficient, while slower speed participants have an improved experience. These aren't contradictory.
> while slower speed participants have an improved experience
Wall Street lobbies to ban HFT every few years, hoping people who don’t like how it looks will back them up.
Slower participants in equities are block traders. Folks moving hundreds of millions if not billions at a time. Large institutional investors. Dealers. Retail, on the other hand, would get hosed, though it would also become much more profitable to execute, so maybe that brings some service perks for larger retail traders.
I gotta say.. I'm way out of my depth on that one. As a guy who's mostly put some 401K $$ into index funds, and has the odd RSU/ESPP stock to sell - will any of these be an issue when I sell some of them later? I've only sold ESPP stock via "at this price" when I had a large enough number, and "at market" when not - but it's been in the hundreds of shares at most. May have a few thousand shares of my current employer's stock to sell next year - will this affect me?
No, it will not affect you. The above post is mostly correct, but it's misleading because it's from the perspective of a trader trying to make short term profits. But most of us are considered investors, we periodically buy/sell low volume of liquid ETF/funds/stocks and hold our positions for years. Market makers do collect a tiny premium for every trade, but it's irrelevant for the time horizon that investors are concerned with.
Since this is hackernews, I'd imagine there are a decent amount of people with low float tech stocks. Those things can be impressively squirrely. I've moved a billion dollar company 5% with a $100,000 order by being a klutz.
Granted, these situations are usually in times of market stress, but these are times when for better or for worse people do need to raise personal cash on occasion.
It depends on how thinly traded the stock is. Unless it's very thinly traded or you trade it after hours you'll almost certainly be completely fine. Like if you're talking about something heavily traded like a FAANG, you could probably dump several thousand shares pretty much any time during regular trading hours and have little or no effect on the price. Certainly not enough to be worth caring about for a one time transaction. OTOH if you're running some trading algo that does that kind of transaction thousands of times a day (or more) every day, then that would be a very different story.
Yes, it may! Assuming you are talking about at least a mid five figure position in a non MAG7 class sized stock, if you post the order directly to an exchange, that's enough to shuffle around the level 2 order book (obviously not in your favor).
By doing so you have completely identified yourself as non- informed, slow human flow. Ex: if you are looking to sell, it's blindingly obvious that the next likely move from you will be to lower the asking price. Even human traders will be a able to take advantage of that situation as a bread and butter trade.
One important aspect is that a lot of this is in terms of opportunity cost and risk. If you are posting the order at a "bad" time (let's say market makers are not long inventory and looking for liquidity), that's when one should expect front-running style action, as they want liquidity ahead of you. Likewise, if you're hanging out there and a market blip in the sector or in the depths of the market complex moves against you, you will get filled and "miss out" on the higher price that the price will settle at. And while you may think they don't care about a 50k order, these robots are hyper optimized and will have had PhDs and 9 figure plus data and infrastructure costs explicitly designed to capture every cent. That's why it's so obnoxious... It's not just market makers either. I know of a prop shop trade that involves harvesting rebates on trending stocks (stuffing the ask and amplifying the trend while receiving credits... if you've looked at stock charts you may have seen a seesaw pattern of liquidity exploration), and if you step into an active trade like that no doubt there will be at least some basic conditional logic to take advantage of stale liquidity.
I walk my very non-tech mother through manual executions on occasion. She finds it very funny that I can see her order, and without prompting has commented about how annoying the little game is.
Numbers - Let's say it's a 50 dollar stock that doesn't get a ton of volume. Most stocks are surprisingly illiquid. Which makes sense because of course nobody wants to deal with HFTs. The lit order book is almost a reference price for the actual trading that happens behind the scenes (midfills at dark venues, etc). Wouldn't surprise me at all to see 10 cents of additional slippage. That's $100. Also wouldn't shock me to see more if it's a smaller stock. Of course it's also fairly common for there to be midpoint liquidity right there for you to take. It just depends on the positioning of each of the participants, and a retail trader is at a distinct information disadvantage.
That said, it's highly unlikely that your 401k is at a DMA (direct market access) broker. Your order is probably first going to go to an internalizer (crossed with other customers), and then flashed to prop firms who will have the ability to take your order (and if they do it probably technically means that you've missed some money somewhere, although it may be in any number of obscure areas), and finally you're going to get sent to the market and pools of liquidity via a decent execution engine. On net, these routes don't work out that badly for retail participants.
Also, if you're talking a highly liquid ETF like S&P or Qs, don't worry about it. Just hit the bid.
That said, I would recommend upgrading if you can. Use a midpoint order type, split your order into chunks, spread it out time-wise a bit. Market On Open and Market On Close order types are also widely available at better retail brokers. I think that these are the most fair fills you can get. Split it 50/50 between open and closing auction. It's just a drop-down order type selector, and you can queue for the auction when you set up the order (say, early morning before the day starts) and walk away for the day and come back to filled orders.
Don't use market orders outside of the huge indexes and megacaps. You're guaranteed a bad fill, and then you also run the tiny risk of a truly awful fill (if something happens machine speed before you can blink... been there done that).
Market On Close / Market On Open orders are really easy to use. Brokers like Schwab and Fidelity and interactive brokers will support them. More people should use them. You'll be getting fair fills side by side with smart money.
Thanks. It sounds you are talking about mostly market-rate sell orders? I tend to use limit type orders to sell at a price I'm happy with. Usually right near what the stock is currently hovering around. I.e. if it's hovering between $50 and $51, I'll put in a limit order for $50.85 and hopefully get someone to take it. Sure, I might have been able to get $50.95, but I might have also gotten $50.25 with a market-rate order. Am I doing it wrong?
Theoretically, when the market offers me an order book and I take offers on one or the other side that should be totally fair? I think until execution/fill the information should be totally between me and the exchange and no one else, right? I get that if I send a limit order that can not be filled, that that affects the market because new information is introduced (before the trade) but in the previously described case all the information going out should be after the trade already happened, right?
Sure, if you want to cross the spread you can usually get a clean fill in exchange for a bit more cost. That said, a fair price is fairly synonymous with a midpoint fill, and if you have a proper execution route you can get the ask (smart algo peg orders for example).
There is a caveat though, which is that top-of-book liquidity is increasingly thin every year. It doesn't take that much size to hit the bid, take out the first thin onion layer of liquidity, and have the spread widen away from you. If you look at the live order book depth you will see that the top of book is often thin and flittering. The deeper liquidity will react to the top levels getting cleared before you can blink. (That's why if you have a non-small order and want the bid price, sweep the bid and go a few cents under, you will get a much more reliable fill and won't be left hanging with the liquidity instantly repositioned a sub-penny below you).
I'm generally of the view that the more freedom in trading the better, but there is a concern on these ultra-high-speed trades that it ultimately prioritizes only certain blessed traders, those with expensive equipment, geographical proximity, and approved partnerships.
I've thought that one fairly neutral fix would be to add a random delay to the execution time of each trade. It could be very small... like between 0 to 1 seconds. Just enough to negate the 'all or nothing' prioritizing of a slightly faster connection.
I suppose you could but the problem is that the liquidity would be either shit or more charitably very different to other exchanges that already do what they're supposed to do.
> Just make all positions irrevocable for at least 10 seconds after posting
Sounds great for Wall Street. Spreads would necessarily widen as people buffer out. Meanwhile, you’ve turned every lit order into a 10-second option for market participants. Which means there is still a latency advantage to lifting or hitting a standing order first.
only if you make those bids visible. the bids are held in a buffer and at the end of the buffer, it either completes a trade or appears on the order book. iirc some market tried to do this with a literal loop with a physical latency.
> only if you make those bids visible. the bids are held in a buffer and at the end of the buffer
Everyone else's bids are invisible. The ones you submit--for your self and for your customers--are not. The larger a brokerage operation you run, the more of an edge you get under this system.
We created the unified tape and passed Rule NMS specifically to remove these incentives.
One interesting approach to this is the gas auction system in DeFi where (on Ethereum) traders bid to have their trades included first in a block, and that additional payment is burned / accretive to ETH holders. Though that turns "fastest connection" into "highest bidder" advantage.
Another approach that Aztec and some others are taking is to shield all transactions with zkSNARKs such that the intent of a transaction isn't known until it's completed. Combined with deterministic block times you could force random ordering of transactions in batches, effectively mitigating the fastest connection OR highest bidder advantage.
The real question is whether we even need stock exchange organizations if we can do it all on chain without them. I think the only thing you really need is someone to handle the stock ownership credentials in the event that legal action require involuntary transfers, that sort of thing. That could be a much smaller footprint organization, I think.
That's the promise of tokenized securities! Securitize + BlackRock are trying to make that happen. This should remove a lot of middlemen to the trading and settlement process.
Perhaps Texas could use a different trading model that doesn't require ultra high speed trading.
Matt Levine often mulls the idea of a system with a trading window that doesn't let the fastest connection to the order book win. Perhaps an order book that works at human speeds so humans can trade too (I can think of a few ways to do it - but would need modelling to try and figure what actually works). He points out that most trades are done in the last hour, so really trading only needs to occur once a day.
The issue is whether a market trading system can be designed with suitable restrictions that beats the current market design (for listed companies and for traders).
Designing markets is hard because you have to assume every player is selfish and only cooperates where it is to their benefit and will defect or cheat if the incentives of the market encourage that (Enron in the California energy markets).
Unlikely since SEC would need to approve of a different system of market trade incentives.
Edit: Personally I would like to see an exchange that was more international. I'm from New Zealand and our good businesses often list on the Australian exchange rather than the NZSX. The system of ADRs for other countries feels like a massive hack.