Putting the Brakes on High Frequency Trading | Hacker News Reader

Putting the Brakes on High Frequency Trading

nytimes.com

38 pointselfinlike58 comments

This article is long on inflammatory language and short on analysis or evidence.

There are plenty of real issues with HFT as currently practiced (front-running data feeds, gaming with flickering orders, dark pools, etc). However, the article ignores these real issues and suggests that liquidity itself is a bad thing (or something not worth having).

Liquidity means having the ability to transfer an asset quickly with minimal loss of value. Anyone who invests wants a liquid market -- illiquid securities are notoriously problematic.

The article is misleading. It is the absence of liquidity that is a disaster. Market crashes are what happens when liquidity dries up. Just ask anyone who was trying to sell their house last year.

As long as the rules are fair (no front-running, preferential trades, equal access, etc), every market participant adds value. Markets are auctions and auctions run best when there are active bidders.

I doubt that HFT traders really engage into frontrunning or market manipulation. It is illegal. Supporting HFT infrastructure R&D is expensive. And takes nontrivial effort. It is simply too risky to engage in illegal activities.

Edit: no point -> too risky

"There is simply no point engaging into illegal activities."

There are many major firms such as SAC capital that built a business around insider trading (which is also illegal).

Just yesterday: http://www.businessweek.com/news/2012-10-02/ex-sac-capital-m...

Unlike HFT, insider trading doesn't require investment into infrastructure and R&D.

Investment of someone elses money. To someone out there this is a rational risk to take.

And if you do it right it doesn't leave a paper trail. I've always suspected insider trading is the norm, and the careless people get caught.

I thought all these computer trading algorithms had failsafes so they abandon the market when something goes wrong? Liquidity that disappears like that doesn't seem so useful.

Failsafes are triggered irregularly and are thus prone to bugs. It is also very rare to go back and review data with a separate implementation of the failsafe to see if the normal implementation is functioning incorrectly in certain circumstances, so errors in the failsafe algorithm can be difficult to detect until a major problem occurs.

There's nothing preventing humans from abandoning the markets as well. This isn't unique to HFT. Human market makers can also withdraw their liquidity when things don't look good.

When one market participant breaks the market with his crappy program, that's a huge problem for the companies that use the public markets to raise money - the entire reason the markets exist. There was plenty of liquidity in the 90's and before to fulfill that function, even with the huge spreads back then. You are missing the forest for the trees.

Breaking the market with a crappy program costs the responsible firm tons of money, up to a complete insolvency.

This system is mostly self-balancing.

Mostly. Until it isn't. It only takes one really bad screw up to cause a boatload of damage.

To the firm who is responsible. If they are buying up securities at well below-market rates, the problem solves itself.

And anyone relying on that firm. What happens if a major bank pulls a Knight and blinks out of existence? Can this happen? I don't know. The answer to that question should be an emphatic "no."

There are already triggers in place now to stop trading when it gets out of reasonable bounds not right? Say like a x% shift throws an alarm that halts orders until a human looks at the situation, etc. I like the ticking exchanges (like the FTSE does not right?) where it "ticks" on the microsecond or some such so some of the clever hacks (locating near the datacenter) are eliminated.

The real question at stake here is do we want to hold back real algorithmic innovation that we actually want? Of course we need to have alarms to head-off flash crashes, but we want a lot of great people working on getting information into the market as fast as possible right? I mean the end goal as I see it is the replace of most traders to algorithms that bring prices closer to market with better/faster information.

I would disagree. We do not need innovation. We need stable functioning markets, which is a goal that is in direct conflict with innovative trading. The system, and the participants' confidence in that system is of far greater importance than the efficiency of the system, and infinitely more important than the opportunity for some smart computer guys to make a buck doing essentially nothing of value. "As fast as possible" should not be the goal. "As fast as necessary" should be, and we are well beyond that.

YOU are missing the forest for the trees. When the a stock is misvalued for about twenty minutes on the public markets, NO ONE raising money in the public markets (in the form of IPOs or secondary offerings) loses money. Not a single one.

Also, you're just wrong about the reason the market exists. The market is just as much about people trying to raise money as it is about people wanting to transfer risk.

> There was plenty of liquidity in the 90's and before to fulfill that function, even with the huge spreads back then

Are you saying things were great back then? why would anyone want to pay huge spreads?

Why do you assume small, temporary mispricings are the worst that can happen? It's all connected.

>> It's all connected.

is that really the best argument you have?

Yep. I believe that the stability of the markets is of vital importance to civilized society, and computers (among other things) can endanger that system when used recklessly. We have clear evidence that the people writing this code often enough do not give a shit. And that is a problem. I'm just saying that there ought to be some thought put into how much we need to push innovative techniques when the stakes are so high. We almost broke everything in 2008 with derivatives and leverage. We out to think about how similarly scoped problems can arise when mathematicians are allowed to get creative before it happens. It only takes once.

none of the things you said describe a mechanism for how hft can topple society / endanger markets / other non specific harms.

seriously, reread your comment. now apply it to online shopping, telephone systems, satellites, air traffic control, etc. it applies to literally everything involving technology. should it all be banned?

I never said HFT. I said out of control computer algorithms. The point is that the markets should not tolerate cowboy coders breaking things in a way that endagers the whole system. Knight failed catastrophically. There is your mechanism. No go apply that to a larger, more unstable part of the market. What if Knight was BofA? Bad things can happen, and if the best argument for this sort of thing is "liquidity", thanks, but no thanks. We have enough of that already. I'm all for letting people gamble when they are the ones who take the loss. But we know that is just not how Wall St works at present time. That needs to change.

Any electronic trading system is capable of experiencing severe errors that have nothing to do with HFT; some of the most common examples involve manually keying an incorrect price or size for an outgoing order. The sort of safety mechanisms implemented by the exchange (which shouldn't permit errors by member firms from damaging the market) to prevent these sort of problems are similar to the sort of mechanisms that can prevent crashes from faulty automatic trading programs.

It's fake liquidity IMO, since HF traders are only willing to take a position inbetween someone selling at one price and someone buying a bit high. The trade was going to go through without them in the middle. They are just skimming some money from the buyer, who would otherwise have a lower price at the cost of delaying a trade for half a second.

Can you explain that situation? As I read it, it goes like this. Someone is willing to buy at $1.50, and someone is willing to sell at $1.60. If an investor comes along that wants to buy right now, he has to buy from the seller at $1.60. But say a HFT quickly throws out a sell order at $1.59, so the investor is now able to buy at $1.59 from the HFT. The investor saved $0.01, and the HFT made a trade he must have wanted. Where is the money being skimmed?

Lets take your example and extended to two investors. One who wants to buy and one who wants to sell. If Bid is at $1.50 and Ask at $1.60 then in an efficient market you can expect the two investors to meet at the middle and make the trade at $1.55. That will not occur. Instead of trading with each other, both investors would trade with liquidity providers who will step in to buy from one investor at $1.51 and sell at $1.59. As a result the investors will get their orders filled at only $.01 better than the expected bid/ask and the HFT guy would make $.08 in pure profit.

In reality this does not occur. What happens is that another HFT guy would step in to offer $.02 to each investor in price improvement and take $.06 in spreads. The next HFT guy tries to shave off a little bit more until the spread narrows to $.01. HFT guys then try to take little bits of that remaining spread with exotic order types, rapidly putting and taking off orders, and other tricks and squeezing more money out becomes increasingly difficult.

Edit: To clarify as per dchichkov comment, because of the competition between HFTs guys, the investors would only see a spread of $.01 or $.02. Thus both would trade around somewhere around $1.55 though not necessarily with each other.

In reality spread would be $0.01 or 0.02 and these two investors would meet in the middle. Even on different exchanges. Thanks to HFT guys.

Yes, I edited that little bit in for clarity.

The part about squeezing the very last bit of the performance still stands though. Because of mid peg orders and darkpools there could also be fractional pennies which are almost always collected by HFTs. If you watch the trade tape, you may see trades happening with $.009 and $.001 sub penny amounts. These darkpool are trades where someone offers $.001 in price improvement and takes $.009 from the spread at another venue if they are lucky.

If HFT firms buy at 1.60 and sell at 1.59, where is their money coming from?

As a high-frequency trader, I can tell you right now that you are just completely incorrect. I'm staring at a screen right now where we take both sides on every order, not 'skimming some money' from people. It's obvious you have no idea what you're talking about.

I would apologize for harsh language, but I have low tolerance for people who freely speak lies of matters of which they know nothing.

If you aren't making money from having the majority of your trades inbetween two trades you are necessarily losing money. But HFT is profitable for you, which is a contradiction.

tumult: As a HFT do you guys (in the gender neutral sense) ever worry about rogue algorithms? By that I mean an algorithm that could be used to harm the market or essentially "blind" other algorithms.

Except that's just not how anyone trades.

Before computers, its not like people just waited until they got a price match to make a trade. Human market makers sat in the middle making money the exact same way computers do. Except they weren't as efficient, fast, or smart, so they had to charge higher spreads to compensate.

Are you a trader that likes paying low spreads? Then you probably like the market better now. Are you a trader that loves getting paid high spreads? Then you are probably writing articles about why HFT should be banned.

Depending on the market and the instrument traded, it can be very rare for "natural liquidity" to generate trades (e.g. mutual fund trades with individual investor through their brokers) and most traded orders interact with some sort of liquidity provider. In these cases, the HFT firms tend to step in front of market makers who are less technically savvy and take the money from them. In this case, customers tend to get better prices on their orders.

One of the big misconceptions about "flash" orders was that customers were being disadvantaged; many customers actually wanted this feature so that they could continue sending volume to more technologically advanced markets such as BATS and Direct Edge and avoid being forced to send to e.g. NYSE (which was very slow) by "trade through" regulations. I've been in the position of wanting this service, and I suspect that most of the criticism came from older markets that were unwilling to invest in technology and change their rules to facilitate faster customer trading.

>However, the article ignores these real issues and suggests that liquidity itself is a bad thing (or something not worth having).

The author does not suggest that. Rather, he says that beyond a certain point, liquidity stops having any practical value.

> beyond a certain point, liquidity > stops having any practical value.

Hogwash. Price improvement is "practical value".

If you're selling an asset, there always practical value to have an immediate buyer willing to pay you a better price than the next guy. That's why auctions go to the highest bidder rather than the second higher bidder.

I would like to say that I was misquoted in this op-ed. I have never said that "high-speed technology was “a destructive force in the market” with “no social benefit.”" I said that there is a limit to the benefit of speed, and we're finding that the latency race to 0 has found that limit.

I believe this is the entire quote: "The new electronic marketplace has several structural inefficiencies. These are what have permitted HFT to become a destructive force in the market, rather than a passive liquidity providing mechanism. This should not be construed to say that all HFT is bad, but there are 2 important points to make – the structural inefficiencies present in the market have created a massive misallocation of resources into technology that provides no social benefit, and structural deficiencies in market structure have allowed for nefarious or accidental actions to disrupt the market". I agree there are some contextual problems with the citation in the article. Retrieved from http://banking.senate.gov/public/index.cfm?FuseAction=Hearin...

Is there really a problem here that's worth regulating?

They point out flash crashes as being a big problem--those flash crashes quickly correct themselves back to their "true" value (whatever that means), and the people most hurt are the people who were using bad algorithms and took risks they didn't understand, Knight Capital being a recent example.

They reference volume being much higher than it ever has been on account of hft, which is true. But what does that have to do with the bubble of the 90s? Higher volume != stock market ruin.

Then there's the small time investor getting hurt. But are they really? The whims of the market now occur because of algorithmic trading as opposed to before, when the whims of the market occurred because of--who knows? Small time investors have been winning and losing since the system was invented. Markets can and always will be fickle.

So again, what exactly is the problem here?

Knight almost blinked out of existence due to computers screwing up. Why is it a stretch to think that far worse catastrophes are also likely to happen at some point? The systematic risk is large, and the benefit is near zero.

You have 100 shares of FOO. It is trading at $50. You bought it at $40, so you have a stop-loss order to sell it at $35. Some "liquidity provider"'s algorithm goes haywire and the price drops to $30. The exchange's "circuit breakers" halt trading or revert the trades, and then the price quickly goes back up to $50.

Unfortunately for you, your brokerage's computer saw the price drop, executed your stop-loss order, and put your shares up for sale, which were then purchased by another 'liquidity provider' (or hell, maybe even the same one that caused the value to drop in the first place). You've now lost $1500 because of HFT. Your trade can't be reverted, because fuck you, that's why, but the holy "market makers" get their trades rolled back so that their rich friends don't suffer anything for their alogrithmic fail.

The real problem is that your "stop loss" is a standing order to sell, at any price, if the price of your asset falls below some threshold. That is fundamentally a bad idea, and it's just this sort of trading behavior that creates liquidity crises.

No, the real problem is that one person's program broke and caused real losses for other people who were acting rationally. If you must account for a non-zero chance that some geek's algorithm will go totally batshit and lose you a bunch of money in unpredictable ways, there is a huge problem at the very core of the marketplace.

Who cares if institutional traders can't figure out why the price is moving.

All traders do is arbitrage, computers are much better at spotting arbitrage opportunities than humans are. That institutional traders can't figure it out merely points to their irrelevance.

Now, on the other hand what we should be concerned about is if HFT is ruining the investment climate and since companies rarely offer their stock it almost certainly has no effect on the investment climate which looks at the macro conditions across years rather than the microclimate of the next few minutes as traders do.

The price of a stock, or the value of an index at any point in time has almost no relevance to the economy as a whole. What matters is that over long periods of time these companies deliver great value to their customers.

Lets worry about GS, et al, selling stock that doesn't exist (naked shorting) before we worry about HFT.

For every trade you will incur some transaction cost. You may think that even though there is a Bid and an Ask price, you may on average get somewhere in the middle, but you don't. Your buy trades are likely to be filled on the Ask price and your sell trades are likely to be filled on the Bid price because some intermediary will get in between and pocket the spread. This has been the case, this is the case, and I haven't seen viable solutions to this what so ever.

What has changed over the past decade is who that intermediary is and how much of the spread they can steal. In 2001 New York Stock Exchange reduced the minimum listed price increments from $1/8 to $.01 . This spread used to be collected by exchange "specialists" with inherited seats on the floor for providing liquidity. The rise of alternative exchanges and computers allowed third parties to cut in to provide liquidity with smaller spreads and cutting out the specialists. These third parties are high frequency traders.

Investors are still loosing the spread on each trade they make, but it is so much less than it used to be. There are issues with the market microstructures that could be addressed, but trying to push computer generated should not be it.

I for one, never want to go back to the dark days of specialists and their $.125 and $.25 spreads.

PS. This was a discussion of the minimum transaction cost you can have. Additionally you may have to pay fees to your broker ( especially if you are retail and don't trade a lot ). If you are trading lots of shares at a clip you will also encounter costs from the market impact of your trades.

Or, implement Mark Cuban's 10 US cents per trade cost idea. http://news.ycombinator.com/item?id=2862003

No more problem. And it could be harder to game than taxing profits after the fact...

It's not really his idea, but it's probably a good one. I remembered it as being called a Tobin tax, but that really only applies to currency exchange. Instead, it's just a securities transaction tax: http://en.wikipedia.org/wiki/Financial_transaction_tax#Secur... .

The European Union is seriously considering adding a 0.1% tax against trades in stocks and bonds (and an 0.01% tax on derivatives). It would definitely curb HFT, and there's talk of it also generating needed revenue. The main criticism is that a transaction tax distorts the market in hard to predict ways, and that big players may work around it. The English Wikipedia article has a decent summary of the proposal status: http://en.wikipedia.org/wiki/European_Union_financial_transa...

Unintended consequences alert: read up on the UK stamp duty on UK equity trades. Goldman Sachs creates a derivative security matching the return of the underlying equity that you can trade with GS and avoid paying the stamp duty. Net result: GS makes money, less liquidity for the average investor. Rock on transaction taxes.

Goldman Sachs creates a derivative security matching the return of the underlying equity

So Goldman Sachs is using a loophole, how surprising.

What is your point, though? Should we not make laws that can potentially have loopholes?

As with most things, this addresses symptoms and not the actual underlying problems. HFT have exposed all kinds of problems which existed since the inception of the various markets but only appeared recently.

For example, Regulation NMS is an important rule that was introduced to solve a major problem (ensuring price fairness in a decentralized environment) but it didn't make sense back then. However, given the physical constraints, the problems weren't apparent. Now HFT has highlighted the broken nature of the regulation. The right solution is to force a re-centralization of the exchanges, but instead of that people are pushing to stop HFT.

I'm happy that there seems to be some discontent here on the substance of this article. Indeed, we perceive market 'crashes' to be negative, but in fact they (necessarily) benefit as many people as they hurt. Being able to obtain stock in a company for fewer US$ can be a good thing. And who's to say that banning algorithmic trading will prevent rapid swings in the market? They happened before algorithmic trading was here, and they will happen afterwards.

All of this rhetoric seems to be a knee-jerk reaction to something the author is incapable of grokking (no doubt due to a limited human capacity to do so).

HFT crashes do not benefit or hurt anyone, since they are rolled back.

I'm almost certain this statement is incorrect.

1) Just because a price reverts to a previous value does not mean that no-one benefited or was hurt from the fact that it fell or increased in between.

2) The 'rolling back' of an HFT-prompted crash (the causal link being debatable anyway) is not guaranteed at all.

Want less emphasis on speed? Let tick marks be less than $0.01. Market-makers are clearly willing to provide spreads of less than $0.01. Any time the price of something is artificially moved out of market equilibrium, there will be queueing (either to buy or sell the good). In this case, there is queuing to provide liquidity, and market-makers have every incentive to try and jump the queue. Instead of competing on price (they can't), they compete on speed.

I don't see the point of worrying about volatility. The price moves and people don't know why. So what? I don't care if traders get heartburn - that's part of the job.

The real problem with HFT is its basic unfairness. Firms that aren't physically located near the exchanges are at a disadvantage, as well as individuals located anywhere. I like Glenn Reynolds' idea of adding a randomized delay of up to one second to every trade.

Volatility is reflected in value, which I suspect you care very much about.

I'm not sure what you mean by that. What I care about is how my asset does over the long run. If it goes up 10% today and down 10% tomorrow, I really don't care.

Reply on news.ycombinator.com