Forget adding speed bumps, frequent batch auctions are the way to go. Group orders every second and match trades (and maybe give priority based on volume). There is no reason any company or trader cannot commit to price for 1s and it completely eliminates high frequency shenanigans.
It shifts the problem so that every bot is waiting for the end of each second to gather as much information as possible from external sources before placing its orders. So when a human trader places a market order, they will have a disadvantage equal to the volatility over 1 second. By contrast, in a real-time market, the HFT bots make your market orders very close to the current price everywhere else.
It's not clear to me that batching actually solves all our problems. It might be better, but it doesn't seem 100% clear cut.
Yes, in the case where external pricing information is released a short time before the window closes, collocated HFT will be able to win. But I'm not sure it would be any worse than the current situation. Also with batch auctions there can be multiple collocation points in different cities.
But when the external information is released earlier in the window (which should be just as likely if random, and maybe even mandated for press releases to occur at the beginning of the window) the playing field between collocated HFT and smaller traders is much more level than the current situation.
But most information used in trading isn't from external sources, its from the order book itself. Collocated HFT with 1ms latency can read the book, calculate next move (let's say 20ms), and submit new order in 1 + 20 + 1 = 22ms. A trader with 100ms latency and the same system would take 220ms to respond.
This is where most of the HFT advantage comes from, they can see the order book and respond before it even arrives at your computer. With a batch auction everyone sees the previous results at the beginning of the window and has 1s to respond. Low latency is still desirable to allow for more processing, but its no longer overpowered.
edit: 1 sec was chosen completely arbitrarily, the best choice could be more or less, or even a random sized window.
There is currently a dark pool using a similar method:
"The London Stock Exchange’s Turquoise Midpoint Dark Pool has a service known as Uncross, which matches orders at the midpoint of the bid and offer with auctions held randomly every 5 to 10 seconds."
Which is kind of the point. If people think that rolling auctions are better than a CLOB, they are free to set up exchanges that use auctions and see if anybody trades on them.
Spoiler: No one will.
Or are they arguing that the government should step in and tell people how to trade?
> when a human trader places a market order, they will have a disadvantage equal to the volatility over 1 second.
No, there's no disadvantage for the human's 1-second-old stale order because batch auctions use uniform-price clearing (aka single-price clearing, all orders in a batch interval transact at the exact same clearing price).
I agree completely. Delayed 'ticks' are the way popular RTS games handle latency [1], and yet it doesn't hamper the high-level players from needing very quick reaction times and strategy. In the HFT world, it might require a rethinking of strategy, and honestly some 'thinking turns ahead', but I don't see that as a bad thing.
Relevant papers for frequent batch auctions. To me, seems like such a hurdle to get politicians to understand what frequent batch auctions are, and how they would help the HFT situation. Especially if the introduction paper is 84 pages long.
Both interesting reads though. We actually implemented batch auctions as a way to set initial prices for our prediction markets at Inkling Markets (http://home.inklingmarkets.com/) based off these papers.
Those are really good papers & I highly recommend them. A couple of things I've always wondered about them:
- their proposal seems to require either coordinated legal regimes & consistent global clocks or a monopoly exchange, both of which on the surface seem impossible.
- their matching algorithm both still has a time priority component as well as a random or pro rata component, which is problematic.
- finally, the main complaint seems to be with the profitability of the arb not decreasing, but isn't that a direct function of the legislated price floor imposed on the markets?
Global clocks aren't required, as batch intervals on different exchanges don't need to be synchronized. Although the batch interval is a window of cross-exchange risk, the uniform-price clearing means it is still more efficient (in "volume-space"). Continuous auctions have really short risk windows (more efficient in "time-space"), but the discriminatory pay-as-bid price clearing means that all that risk is shouldered by everyone but the fastest traders (who are thus able to extract arbitrage rents from the slower traders).
Batch auctions don't deal with the main area of "high frequency shenanigans" which is cross exchange arbitrage. There isn't "a market" there are lots of little ones, geographically dispersed and across jurisdictions.
Further, there are already pro-rata matching engines, and they also have "high frequency shenanigans". In fact, they actually can encourage more gaming as you have to put in more quantity than you want, so you have to do something else to manage your risk.
Batch auctions do defeat cross-exchange arbitrage. I went into some detail about this here: http://cdetr.io/smart-markets/ [excerpt from Section 2.1 Batch auctions and mechanical arbitrage]
> ... batch processing with uniform-price clearing eliminates mechanical arbitrage opportunities. To be clear, if only the local market utilizes batch processing with uniform-price clearing, and the external market utilizes conventional serial processing and discriminatory pricing, then there will still be arbitrage opportunities in the external market. That is, the fastest traders can still profit by propagating prices from the local market outward to the external market. But arbitrage opportunities in the other direction, from the external market into the local market, are eliminated.
I suppose it depends on what your definition of shenanigans is. In the context of IEX & their artificial delay that is purely to deal with cross exchange trading.
I really like this. A goal of a stockmarket should (among other things) be to create a level playing field for investor. This would really help this and offer no negative incentive to a true investor.
The EU is imposing a tax on financial transactions effective January 1, 2016.[1] The tax is tiny (0.1% on most transactions, 0.01% on derivatives) but enough to discourage high frequency trading. It's going to be interesting to see how that works out.
The EU is imposing a tax on financial transactions
The devil is in the details, but at first glance that looks like an absolute disaster in terms of stock trading. For example, if a mutual fund bought a stock twice in a year, and then sold that stock twice in a year, that's 0.4% being gifted to the government right there? From just that mutual fund. And another 0.4% from whoever is on the other side? Wow!
Am I reading that right?
Also it says "financial institutions" but looks like it will also apply to individuals. The Wiki page says "If acting on behalf of a client, e.g., when acting as a broker, it would be able to pass on the tax to the client."
I think that could dry up trading volumes by about 95% in the affected countries. For example, when I sell stock now, a market maker takes the other side. But he won't want to if he has to pay 0.1% for the privilege. So now I can't sell a stock to an intermediary? I need to wait for another investor or a mutual fund to take the other side of my trade?
What am I missing? That can't be right?
Edit: one immediate thing that sticks out is that "derivatives" will be created to replace stocks, since the tax savings is about 90% by doing that. For example, let's say a stock costs $10.00. Why not create an "American style" call option on that stock, strike price $0.01, expiration date 2099. The value of that option is $9.99. But the tax rate is 0.01% on the derivative, vs 0.1% on the stock itself. Someone hasn't thought things through here.
Yeah, that makes no sense; every sane plan I heard for taxing stock transactions involves a flat amount per transaction, the idea being to discourage large quantities of trading (or at least convert it into revenue) while leaving the calm, occasional trader/mutual fund mostly untouched.
> if a mutual fund bought a stock twice in a year, and then sold that stock twice in a year
I think the idea is that such moves contribute to the instability of the market as a whole, and hence should pay a small penalty for it.
> I think that could dry up trading volumes by about 95% in the affected countries.
Yes, it is not entirely clear that large trading volumes contribute anything positive to the economy as a whole. There are real people somewhere in the picture who are trying to buy/sell shares in regular and are getting shafted in terms of either buying price or trading fees by high-frequency bots. The tax should even out things a little bit for the average person.
> The tax is tiny (0.1% on most transactions, 0.01% on derivatives)
What's tiny about 0.1%? The amount of money that would suddenly be leaking out of a big mutual fund when it rebalances or reconciles at the end of the day could end up being huge.
The amount of buying and selling something like a Vanguard index fund has to do to track its index with precision is probably pretty impressive, but it's not the kind of weird HFT which annoys people. Introducing inefficiencies into that process is something I cannot believe EU regulators really intended... did they?
Everyday investors lose, the taxman wins, and to what end? It just seems kind of stupid.
The taxman is my representative. Insofar as Europe is fairly democratic, the taxman is hired by us (the public) to collect funds for public spending. The taxman's victory is my victory.
It is tiny. Your Vanguard fund doesn't do manual rebalancing like that. It rebalances automatically by deciding what to buy as new money is added to the fund and what to sell as money is withdrawn.
Charge an extra 0.1% when someone deposits and 0.1% when they withdraw, and that means 0.2% covers the entire cost.
Many low fee funds charge exactly 0.1% annually for the magement fee. They are frequently designed to be 'no load' funds, which, under regulatory requirements, may charge a maximum management fee of 0.25% per year. Additionally, they are prohibited from charging the load fees you described (entry/exit) if they are marketed as 'no load'. Investors benefit substantially from the availability of these funds, and charging 10 basis points (0.1%) per transaction makes them untenable.
You should also consider that rebalancing happens more frequently than when shares are created or redeemed. Suppose a fund has a mandate to track the S&P 500. If a symbol is replaced by another in the index, the fund will need to alter its holdings, incurring transaction fees.
I'm not really sure what point you're trying to make with your first paragraph. The regulations are going to change. The funds aren't prohibited from following the law! Investors will then benefit 0.2% less substantially from these funds, I guess.
As for the second point, how is that any different from rebalancing? Money added to the fund is used to buy the new shares. Money withdrawn from the fund is marked against sales of the old shares, no?
You mean investors, at a minimum, will pay at least 3 times what they used to pay to invest their money. That's far more substantial than you are making it out to be. SPY is an ETF tracking the S&P 500. It has ~180 billion in assets. 0.1% is a nontrivial $180 million.
It is rebalancing, but you seemed to state that rebalancing only occurs at creation and redemption. In the case of SPY, the holdings will change every time the underlying index changes, which has happened 20 times in 2015. Each change in holdings requires, at a minimum, 2 transactions, and involves 40 tax payments as a consequence.
Google finance says SPY is up about 0.7% today. 0.1% is fairly trivial.
I actually stated the opposite on rebalancing - that it happens continually - so I'm not sure where the confusion lies. Money is continually entering the fund and continually leaving the fund. Perhaps using some numbers will make it clearer what I mean. Lets say you decide to invest $20 in the fund. And I decide I need my $20 so I withdraw it from the fund. And the fund happens to need to sell MSFT to buy RHT today to rebalance. So it sells MSFT for $20, which you get $19.98 of (minus your tax), and I invest $20 which they use to buy RHT of which I get $19.98 (minus my tax).
Mind you, I don't actually know how this tax is applied for a fund. Is it applied to the fund purchasing shares, or to my deposit in the fund, or to both?! Maybe it's like VAT and it just gets passed down to the final buyer?
Anyway, I'm generally a fan of financial transaction tax. I think it should replace all other taxes. The major benefit being that it can't be avoided by the most wealthy or by corporations, like most other taxes can.
0.1% is far from trivial. If you purchased SPY at the open and sold at the close today, you would have paid 0.1% of the open and 0.1% of the close, or 38% (back of the napkin math) of your profit on the trade in tax. And, that's in addition to any capital gains taxes you need to pay on the profit already.
You appear to be confusing creation/redemption with rebalancing. Creation/Redemption occurs at end of day when people enter and exit the fund. In the trivial example, suppose you're paying for a full unit of the fund (so 1 share of MSFT, 1 share of RHT, etc.) Then the fund manager will need to go and purchase 1 share of everything for you. Now there is also rebalancing, where the index being tracked changes. Suppose we have SPY. There were 20 instances throughout 2015 where the symbols in the S&P 500 were changed. Even if no one created or redeemed SPY, the fund manager would be subject to 40 transaction taxes simply to track the index (sell the old shares being replaced and buy the new one in the index).
SPY, for the record, has an expense ratio of 0.09%. How can State Street afford to manage the ETF for you in the face of this tax? Well to start, you will pay higher fees, which will lower your returns.
I also looked into the details of the tax [1]. It has horrifying economic ramifications:
* Up to 90% reduction in derivative transactions -- Markets are meant to move capital to people and firms who need it and risk from those who don't want it to those who want it. Derivatives are exceptionally important and reducing their use is a bad thing. For example, Apple needs to buy phones in one currency (Yuan) and sell them in another (USD). They face exchange rate risk, as the value of the two currencies fluctuate relative to one another. Apple can enter into a derivative to exchange their potential upside if the exchange rate fluctuates in its favor for a guaranteed exchange rate. Without these contracts, global commerce would be severely restricted.
* A long-run (20-year) reduction in gross domestic product -- this reduces levels of consumption across society, reducing tax income from other sectors and reducing employment, increasing the demand for government services. In other words, you are reducing the ability for the government to pay for services that it now will need to provide more of.
* An effective curb on automated high-frequency trading -- Why is this a good thing? HFT ensures greater liquidity in the market and simultaneously ensures that the prices of assets more accurately reflect their true value.
* An increase in capital costs, which could be mitigated by excluding primary markets for bonds and shares from the tax -- Increasing capital costs reduces economic growth, job opportunity, and R&D. The mitigation proposed is not as effective as one might think, as any purchaser in the primary market understands that the asset is less desirable in the secondary market to a purchaser as a result of the transaction tax there.
This tax is bad economic policy and will hurt everyone in the long run. However, it is a popular sentiment to want to punish financial institutions because they were bailed out. However, it's less popular to acknowledge that the government has made a substantial profit on those bailouts [2].
Okay, so you don't know how index funds track an index. That's too bad, but put that aside for a moment.
The notion that long-term investors in anything should be punished so severely on the front end (if you understood compound interest you would know that losing 0.1% right away is a bigger hit to the long-term investor than it at first appears, which is why basis points matter a lot when a long-term investor is weighing the cost of a mutual fund or taking into account brokerage fees) as part of a tax designed to penalize HFT and derivatives trading is just completely strange. It's an indicator that the people who created the law don't know what they're doing or have (really dumb) ulterior motives.
The problem with a FTT is that it is a cascading tax. Whereas a VAT or GST is charged only on the difference between input and output. I buy some wood for $10, make a table, sell it for $15, GST/VAT is 10%, so I pay out $1 buying the wood, and collect $1.50 selling the table. I only have to forward the difference, 50c, to the government.
With an FTT, that 'tiny' tax is charged indiscriminately all the way through the chain. And passed through the chain to the end consumer. So the market maker pays it twice (in and out of a position) and their margins are razor thin. So spreads will increase by twice the tax for starters. Plus you as end consumer will pay it twice (on buying and selling). So at a bare minimum that 0.1% will be 0.4% for an end consumer. More if more middlemen are involved in the chain, since it just compounds. What are annual returns? 4-8%? So we're talking about the government taking 5-10% of your profits? Does it seem tiny still?
The Citadel letter mentioned in the article (http://www.sec.gov/comments/10-222/10222-16.pdf) is a very good read. I work in the industry and they hit all the main problems with IEX's approach.
> This damage to market quality would be further magnified by the “fast pass” that IEX proposes to give its affiliated routing broker-dealer (the “IEX Router”) and its pegged order types. It is ironic that IEX—a company supposedly founded to protect investors from various types of latency arbitrage—now proposes to offer pegged orders and IEX Router services that can and will be used by sophisticated trading firms to arbitrage the latency that IEX itself would create.
Could you explain this in more general terms? Especially,
1) Does IEX propose to exempt some traders from the delay? Isn't that obviously worse than no delay at all?
2) What are "pegged orders"?
Also, if you don't mind being spokesperson for an entire industry, is there a sense that when we are debating 350 microseconds, things are a bit absurd?
I haven't read the documentation, but I can respond to your questions given what you quoted.
1) It seems that way. Presumably, they will require affiliates to follow stricter rules. Even so, you would expect that these affiliates would try their hardest to take advantage of their speed advantage.
2) Pegged orders are essentially orders that the exchange manages for you. For example, if you peg an order at the inside bid, your order will follow the market as it moves around. Essentially, these orders will react with close to zero latency since the exchange itself is modifying them. It's unclear how these orders will be used to game the market, since their logic is so simple and predefined by the exchange.
In terms of time frame, it's just the evolution of technology. For example, processors are clocked in nanoseconds. When your limiting reagent is how quickly you can update based on changes in the world, you need lower latencies. That's one of the aspects of the system that IEX tries to solve, by adding a giant delay to everyone's orders, so that the jitter swallows any small advantage.
One of the concerns of changing the rules, however, is that once smart people start using the system, they will eventually find some new way to game it, landing us not far from where we started.
Pegged orders are an order type where you tell the exchange "keep me at the best bid/offer" or even sometimes "keep me n levels off the best bid/offer". Its basically an order that allows the exchange itself to change your price for you so that you don't incur messaging latency and inconsistency.
I just finished reading Flash Boys by Michael Lewis, it was an incredibly enlightening read. To answer your questions:
1) I think it's 350mcs across the board except in the case of firms which conduct suspect activity (canceling orders frequently for example). Adding the delay institutes a level playing field across anyone trading on the exchange. The basis of HFT is leveraging faster connection speeds to gain insight into other's trading strategies and exploiting those strategies, all before the other firm's trade reach the exchange. This is a HIGHLY simplistic explanation dealing with one form of arbitrage and given for brevity on the subject.
2) EDIT: the other guy explained this better!
To your point about 350mcs being absurd to argue over, HFT firms manage regular trading speeds in NANOSECONDS. Look at 350mcs in those terms (350,000 nanoseconds) and it's not such a small number anymore.
If you thought Flash Boys was enlightening then you didn't read carefully. He makes shit up.
Take the one example of a trader that claims the market jumps just by he entering a symbol and quantity- not submitting the order. That's huge! It means espionage in a major trader's office! But since Lewis knows it's just bullshit coincidence, he just says "oooo spooky HFTs".
Flash Boys: Not So Fast is a far better book. And even better, it's actually acurate.
Lewis's book is just an ad for IEX, hoping to scare ignorant people into worrying about perfectly fine stuff. The star, Brad, in his first scene, is shocked, just shocked, that his 50,000 order could cause price impact.
(Think: the customer paid 5 cents (2500$) to get Brad to sneakily sell the shares - obviously price impact is to be expected.)
There are plenty of cross markets, and nothing stops anyone from running a 5-second auction (with simple pro rata or complicated rules like POSIT). But IEXs investors just think coiling 38mi of fiber is neater. At least it makes a cool display for data center visits.)
> Flash Boys: Not So Fast is a far better book. And even better, it's actually accurate.
I just read it on a recommendation from HN. Entertained me on a longhaul flight. Boy, does the author take apart Flash Boys. Alas, approximately no one will ever read this deconstruction.
IEX only adds 350 usec in delay. I'd also add that if you just read Flash Boys you probably got a very incorrect view of what high frequency trading is all about and particular what cross exchange market making is about.
"Kovac then writes, “In other words, this research, cited by Lewis himself near the conclusion of his book, contradicts everything he has said about front-running in the prior two hundred pages.” Kovac suggests this is some kind of “Aha!” moment. See, he seems to say, there’s no front-running and Lewis’s own sources say so.
Your guess is as good as mine on this, but Kovac seems to be the one who apparently didn’t read the research. Go to the next paragraph in Clark-Joseph’s paper: “[T]he private information about price-impact generated by an HFT’s small aggressive orders enables that HFT to trade ahead of predictable demand [that is, front-run demand] at only those times when it is profitable to do so (i.e., when price-impact is large).”"
To be fair, Flash Boys is a book where the main beneficiaries have a vested interest in portraying HFT and especially cross market arbitrage as predatory.
Further, themis trading has a vested interest in portraying HFT and especially cross exchange market making as predatory. Both groups work for market participants that want to move large amounts of shares without impacting the price. That is they want to subvert supply and demand. Their particular quote does a disservice to every one because it equates pricing demand into the market as front-running, which is ludicrous.
Full disclosure, I have worked in the HFT industry (though I don't now). Regardless of whether HFT is predatory or not, Flash Boys inaccurately portrays how the technical details of exchanges work. I do not, and have met no person who is aware of the technical details of the markets who finds it credible.
Granted, Flash Boys is positive publicity for Brad Katsuyama and IEX and much of what Brad and his team uncover is through hypothesizing, testing, and inference, no one in HFT comes out to confirm if these ideas are correct or not.
While the technical details of the system might be up for debate, Flash Boys paints a broader picture of a broken global financial marketplace, run by institutions struggling to keep up with new technology and unable to craft good regulatory policy. A system gamed by investment firms (or more accurately, investment intermediaries like HFTs) whose advantage comes from exploiting these poor regulations and new technologies, creating market imbalances that affect the entire marketplace and generate incredible profits at the detriment of the investors to whom these firms have a fiduciary responsibility to.
The broader picture of a screwed up financial system, how it got this way, and who it serves was my biggest take away from Flash Boys.
If you read carefully you will notice noone in HFT is ever offered a chance to speak about the issues. That is extremely telling.
There are 2 major forces in the markets, on one side are market neutral participants that are not taking a position long term. For the most part they are market makers selling liquidity.
On the other are large block traders, largely hedge funds who want to take advantage of that liquidity at the cheapest price.
Those 2 groups are naturally hostile to each other. Flash Boys presents, potentially in bad faith, only 1 side of that equation.
For instance, that you think the profits in HFT rival those generated by the hedge funds that back IEX is very telling. They are orders of magnitude different & are dominated by the hedge funds. That money also comes out of the pockets of investors.
That you don't mind it being inaccurate in its central thesis is problematic for those it paints negatively, but exactly what those on the other side want.
> Those companies have all said that IEX will introduce one more complex wrinkle that will end up benefiting one set of market participants over another, rather than leveling the playing field.
Is there a notion of fairness that is both canonical and pins down one market structure as fairest?
I'm not suggesting that the proposal is more fair. But in the name of fairness an unfair system must be changed in ways that benefit some system participants over other system participants. Likewise I am not suggesting that the existing system is unfair.
Only that the argument that some participants will benefit from changes more than others does not imply that the new conditions are unfair.
I don't think there's any such notion of fairness. But if there is...
There's a saying to the effect that "anything is difficult to explain to someone whose income depends on not understanding it". The hypothetical perfect notion of fairness will face many people giving fine-sounding but self-serving explanations of why it is not actually fair.
I agree. There's also that fairness can be applied to the social structure that generates the rules for the market. Suppose that there is a fairest market structure. We may even call it the "free market". It may be the case that you can enforce the structure of the free market on everyone else, but a lot of people would be rather angry with you if you did it like a violent dictator.
So it may be the case that a system that tries to enforce free market conditions must be allowed to deviate from that ideal in order to be itself fair.
In case anyone initially overlooks it, as I did: The proposed delay is 350 micro-seconds, or millionths of a second, not the milli-seconds that ping measures for me.
If the question is whether a 350 microsecond delay is acceptable and whether it's fair to small investors, we're asking the wrong questions. For example, should HN introduce a 350 microsecond delay to posting comments? Would that be fair?
(In the correction at the bottom, you can see the NY Times made the same mistake I did.)
If I'm not mistaken, this is all people who will make less money because of this change claiming that "some other people" will make less money because of this change. It would be considered "FUD" if it was a Silicon Valley company saying this.
I can't believe so much is written about 350 usec in delay. I mean, that must be the least controversial change you would ever make.
If someone wants to make an exchange with static added latency, that's not the same as making everyone's latency the same, which the article conflates. The change is nearly pointless, and they argue it to death.
There are much more interesting things going on with exchanges than this.
What if you just accept algorithmic trading, and allow companies to upload an "agent" that handles trades for them?
Each agent, running on a standardized VM, can be given the same amount of CPU cycles, the same standardized data feed, and logically run in the same time slice. Companies can add real time data to the feed, but it is also available to everyone else. Private data is delayed by a second or so, as are new agent uploads.
