A High Frequency Trader's Apology, Pt 2
chrisstucchio.com
chrisstucchio.com
But in fact that entire scenario he just described is perfectly legal. The industry calls them "flash orders", and due to industry pushback against an SEC initiative to ban them in 2009, they are still legal.
Nutshell description:
- you put in an order to buy AAPL
- your trusted exchange "flashes" the order to a company that has paid for the information, so they know you want to buy AAPL. They now have exactly .5 second (SEC rule) to act on that information before your order hits the wider market.
- said company goes and buys up the AAPL on the market, ahead of you
- your order to buy AAPL hits, executing at a slightly higher price than you expected
- said company sells AAPL at the new higher price (perhaps to you)
You've acquired AAPL at a slightly inflated price, and another company has pocketed the extra money you paid. They could buy low and sell high with a GUARANTEE of success because they knew your order was coming to the market. We used to call this front-running and call it illegal, but not in the modern U.S. stock market.
This is completely legal in the United States. It's happening now. It happened hundreds of times while you read this sentence.
This has always happened. Suppose you're in a room with your investor club and you want to buy 500 shares of AAPL. The evil "flash order" is akin to mentioning to your local group "hey guys I want to buy 500 shares of AAPL at $x, in case any of you want the other side of the transaction".
If none of your local group (with whom a trade would incur no transaction cost) wants it, then it goes through to the wider market where it may or may not find a counterparty.
The 0.5 second time limit is arbitrary, chosen to accomodate a range of latencies for the various local would-be-counterparties.
Nobody is stealing anything. It's just a way around some of the fees that one willingly pays for harder to fill orders. If an entity on your local ETN has a market making strategy where he/she is willing to fill some orders, then a flash order rule can improve efficiency by removing transaction cost.
If you think this is a bad thing then you have a profound misunderstanding of its mechanics. Ironically, like most of the anti-HFT claims, your position benefits the old school establishment exchanges. ETNs are the little guys trying to compete against the big guys by offering better technology and lower transaction costs. Flash orders are part of their service to fight against the competition-stifling fees charged by the big guys to route electronic orders.
And yet most market participants think banning them is a no-brainer.
The NYSE's members are the big market making firms who have the most to lose if there is more competition -- IE more small firms using market making strategies.
They prey on the fear and ignorance of their audience.
This gets back to the nature of the market price. There is a fallacy that the market price is the price of the last trade. At any price, there is some combination of supply and demand.
Similarly, at any latency, transaction fee, etc., supply and demand may converge slightly differently (without NBBO rules). Since there is always the risk of trades occurring too far away from the price on the larger market, any ETN trader is going to figure out what his/her risk aversion is to this phenomenon and design his/her strategy accordingly.
For an HFT trader on an ETN w/o NBBO rules, it might make sense to buy a data feed from the NYSE to be sure to be aware of the up-to-the-second prices there. The NBBO rules benefit both the large mega-exchanges like the NYSE and also benefit other market participants who would have to buy a separate data feed to reduce risk, by strapping the cost of that additional data onto the backs of all the other participants, even those who would have a greater appetite for risk or whose strategies don't depend as much upon the ultimate depth at a given price.
What makes this even sillier is that for the small investor, trades on something like eTrade cost $20 each. If you're simply moving the investment between two investments that means two trades. This makes a LOT of strategies utter failures and not worth trying. The world of ETNs is just one more step of automation, dollars, and sophistication away from a simple eTrade account... hence its highly disruptive nature and the many startup hedge funds that have sprung up.
The genius of ETN creators was the realization that there is lots of depth provided by smaller, more niche players... to the point where many trades can be filled directly on the ETN for no fee. This opens up the door to many strategies that would simply have been impossible before, and combines the capital of each of these smaller players... each of whom has some exposure and who combined offer non-trivial depth... enough to take away volume from the major exchange monopolists.
This was not lost on the major exchanges who used their clout with regulators and journalists to institute the NBBO rule, to paint flash orders in a negative light, and generally sure up their monopoly positions against any competition.
Yes I said it, the major exchanges should be subject to a major anti-trust investigation and broken up. Of course, everyone (regulators included) is so afraid of upsetting the market that this will never happen... and if some other country opened up a market allowing such things, the US would ban Americans from using it.
What is your definition of "ETN"? The only definition I know is "Exchange Traded Note" which doesn't make sense here. Do you mean "ECN"(Electronic Communications Network)?
- I believe you have to opt in to have your order flashed. You seem to be suggesting this is happening behind traders backs
- Only one market in the US still supports these types of orders. It may have happened hundreds of times while reading your sentence but that is a tiny fraction of executions occurring across all exchanges during that time-frame.
No, the SEC mandates that the limit be < 500ms, i.e. if it were > 500ms, then the exchange would be required to publish the flash order.
The limit was 30ms at Direct Edge, and I believe everywhere else as well.
There are also mandatory fill rates - i.e., if you don't fill at least 30% (or some such fraction) of the orders that are flashed to you, you get kicked out of the ELP program.
Your timeline is also wrong:
- said company goes and buys up the AAPL on the market, ahead of you - your order to buy AAPL hits, executing at a slightly higher price than you expected
Your limit order to buy AAPL at $610.00 hits the market and goes unfilled.
In fact, the ELP program was mainly about increasing order flow. It was a way for some HFT's to jump the queue - rather than being faster with public bids on ARCA, you can just flash people on Direct Edge.
The people being flashed chose to be flashed (note: there was a flag you could set to turn flash orders off) because it would reduce their routing fees.
A much more legitimate criticism of flash orders is that they benefit Goldman or Getco to the detriment of Joe's HFT Shop.
When I read about these things it reminds me of when I was working for a gambling site a few years back. Arbitrary rules, designed to maximize profit for the bank.
If your ELP members aren't filling orders, then customers pay more to have their orders routed and DE makes less money.
I'm going to go out on a limb and say that you, and those who upvoted you, have little or no experience in financial markets. Comparing flash orders to front running a trojan is a terrible analogy. For one, the mechanics are not as you described (other commenters have touched on this). But most importantly, where this is still practiced, the market participants have voluntarily decided to do so and the order flow is openly published, unlike a hacked computer where the user is oblivious.
And even if we assume flash orders are evil, I think it's disingenuos to mention all of the SEC drama surrounding them without mentioning that several leading exchanges have voluntarily stopped the practice, and that flash orders make up a tiny percentage of total trading volume.
Hint: the people doing the trades are not brokers.
>And even if we assume flash orders are evil, I think it's disingenuos to mention all of the SEC drama surrounding them without mentioning that several leading exchanges have voluntarily stopped the practice, and that flash orders make up a tiny percentage of total trading volume.
So you're going to argue, in the same post, both that flash orders are not front-running and that several exchanges, to avoid liability, have stopped doing them? Intriguing.
It honestly isn't clear to me at all what you're griping about. It's like you read an article on the Internet and suddenly you're an expert on flash orders. I am not an expert on flash orders (and someone correct me if I am wrong), but it's pretty clear to me you're very confused.
The practice was stopped because it became unpopular due to the dramatization of HFT in the media. In the presence of a large number of fragmented equity exchanges, flash orders facilitate lower transaction costs and lower latency.
Equating flash orders and front-running does not make sense as the HFT doesn't know of the order before it is placed vis-à-vis the retail trader. The HFT appears to have prevoyance vis-à-vis slower market makers, which is why they are the ones complaining about high frequency market makers. It's actually quite analogous to Hollywood complaining about Netflix, which is why it confuses me that HN participants find technology in the capital markets so anathema.
And also because they have paid to get the information 0.5 second sooner, so that they could trade on their insider information.
Further, I submit that any HFT will only place a buy/sell spread in the case where the volume is high enough that they can complete their purchases within seconds or at most minutes.
At best an HFT will cut a few minutes (more often seconds) off of the time of a trade. Anyone who is worried about waiting a few minutes for a trade to complete is Doing It Wrong. Therefore, the "liquidity" HFTs provide is only in the cases where it's not truly needed because the volume is high enough to provide it anyway.
I've traded in a lot of stocks where there was very little liquidity, and fewer than a few dozen trades would happen per day. I'd sometimes wait hours for a trade to complete. Where are all the HFTs providing me "liquidity"? They weren't anywhere to be found, because in the cases where the volume is too low HFTs can't make a sure bet. And when the volume is high, they aren't needed. So what good are they again? At any point that an HFT is willing to buy stock, it's because there's a high likelihood that someone will show up to pay more after a few seconds or minutes.
IMO the popular media actually has it right. HFTs really are just stealing pennies on every transaction, because they only enter a stock and "add liquidity" if and when the stock doesn't need it.
The onus isn't on me to prove they aren't providing liquidity, but on the HFTs to prove they are, and the OA doesn't even come close. In every example OP gives, the people actually buying and/or selling stock would have been better off without the HFTs in the mix (at best their trades were a few minutes faster).
If we just taxed or put a fee on a "short term hold" of a stock, say less than one day, then high frequency trading would vanish. I see HFTs as a parasite on a broken system, nothing more, and I'm shocked that so many HNers idolize them.
Bid/ask spreads used to be 10 cents or more. Now they're generally a penny. That is evidence of a more liquid market.
In practice, this is because computers are now market makers instead of humans so they can do this job at a lower cost.
You clearly ignored one of my examples.
Consider the example of Fry and Zoidberg. Fry places his sell order at 12:01. Leela and Bender are absent, so Fry's order goes into the book. Zoidberg places his buy order at $9.50 at 12:05. The bid/ask spread is now $0.50 and Fry has not traded.
In this case Fry is better off with Leela in the market.
At any point that an HFT is willing to buy stock, it's because there's a high likelihood that someone will show up to pay more after a few seconds or minutes.
If you want to take that risk, place an ALO order at the bid or ask price. You won't cross the spread, though your order might go unfilled.
The fact that most people don't place ALO orders suggests they don't want to take the risk.
(Edit: sorry, this doesn't really relate particularly to this sub-topic.)
Now mulitply that by 60 holdings which are turned over 100% per year and you can see why pensions would want those mutual funds to save $6,000,000/year in liquidity costs.
(Obviously mutual funds don't just do program trades, they also negotiate directly with each other etc. but the example still stands).
I wonder if ETFs would have ever grown as large as they had if HFT never existed (the time periods of the rise of ETF parallels to the rise of HFT.) Anybody have any insight here?
ETFs control tracking error, how closely the ETF follows the index to which it is benchmarked, by letting authorised participants (APs) arbitrage the ETF against the underlying. For example, if you are a SPY AP, you can turn in a certain quantity, called the creation unit, of S&P 500 constituent stocks and "create" the correct number of ETF units. Alternatively, you could surrender a creation unit of SPYs and the ETF trust would "destroy" those ETF units and issue you the S&P 500 constituent stocks [1].
The risk for APs increases non-linearly vis-a-vis the liquidity of the underlying. This has been empirically witnessed as a "U-shaped relationship between fund premium and market liquidity, which suggests that more active trading does lead to lower mispricing but only after a certain level of liquidity is reached" [2].
[1] http://www.londonstockexchange.com/traders-and-brokers/secur...
[2] http://onlinelibrary.wiley.com/doi/10.1111/j.1468-0416.2008....
Edit: I am assuming here that we are talking about trades that actually change my long-term position. That is, I am selling or buying those million shares for good because I am re-weighting my portfolio or something like that.
While $6M in not large by percentage, there is no reason to want to give that up.
On the other hand, imagine a world of only portfolio managers and no HFT. That is, trades happen only between portfolio managers with no middle man.
Then yes, on one day you as a portfolio manager would have to give up $6M. But where do those $6M go? Logically, they must go to another portfolio manager. By symmetry of portfolio managers, it follows that on other days, you will be the one who gains $6M. That should net out to zero on average, at least assuming that all portfolio managers are equally sophisticated.
On the other hand, the HFTs earn money, otherwise they wouldn't be in the business. Where does that money come from, if not from the portfolio managers?
So it seems that as long as you're looking at a narrow micro perspective, the story makes sense. But once you add up everything to a macro perspective, the argument vanishes.
This does not necessarily apply against algorithmic trading in general. Algorithmic trading may well serve to always have some orders in the order book even on low volume markets, since humans just cannot trade on as many markets simultaneously as a computer can. But the arms race to ever lower latencies just seems useless from the point of view of society.
HFTs will not net positive out over the long term so they are not taking away from other investors. (this is based on what I said in another comment above).
Mutual funds trade huge volume and need liquidity to take advantage of market research. Without liquidity their size would require them to stay smaller and require more investment professionals for any dollar amount they have under management. Without liquidity you would see smaller and smaller mutual funds as their capacity got capped which would lead to higher expense structures for investors.
--- As far as the arms race, I agree with you.
But the arms race to ever lower latencies just seems useless from the point of view of society.
5 years ago it made sense and helped investors, now or at some point in the future the market will reach the point of dimishing returns from incremental small gains in hardware. At the same time HFTs will be investing more in hardware than they will be netting out of the system. Soon, if not already the HFT business will be mature and they will net negative and become consolidated. At that point (if it hasn't happened already) HFTs will have benefited the market with additional liquidity but any profitability long term will be negative.
I also know that HFTs are taking money out of the equation, not adding money into the equation. If they weren't, they wouldn't be doing it -- and if the only value they're adding is reducing the time that a trade takes place by minutes or seconds, then I still submit that the value added to the stock market isn't worth the ACTUAL dollar cost.
The OP pointed at the example with Fry and Zoidberg. In that case, Leela either makes $0.05/share on the transaction (which otherwise would have gone to Fry and/or Zoidberg), or takes a hit of $0.50/share. In the latter case Fry (say that's you're mutual fund) makes $0.50/share more than they would have otherwise, but that money doesn't come out of thin air -- it's lost by Leela.
But in order for HFTs to do well, they have to make more money than they lose, so for every case like the above where Fry does better than he would have otherwise there is more money lost by people who would have done better without the interference of HFTs.
So for every $100,000 saved by a mutual fund, much more is being siphoned off by HFTs, for no real added value.
This assumption is questionable. There are a vast number of market participants who do loose money.
The best start up example is the bias toward reporting companies who just got funding and not reporting all the companies that hit the dead pool.
We don't really have a good idea on the net of HFT strategies and it is just as safe to assume they net zero.
Lets just say HFT firms did net to zero, would you still have the same argument?
Mutual funds beat the bench?
Hegde funds?
Airlines? They spend a ton of money on fixed costs.
Some make money some do not.
An HFT investment likely involves someone deciding they want to create an HFT fund, going out to institutional investors and selling them on the idea that THEIR fund will be profitable.
That happens over and over and investors in these HFTs (as well as hedge funds, PE etc. all think they are picking winners who can make it work. Whether they can or not will depend on execution.
Investment in a sector, industry, or asset class does not mean there is net winning from that sector.
At the end of the day, HFT is still competitive since they are bidding against each other (hence the arms race for faster equipment). Even if the market is growing for HFT now, it does not mean it will be forever. At some point it will become mature and all that will be left is net negative HFT profitability with all the benefit going to investors in the form of liquidity.
I am not backing up HFT because I trade in HFT, I do not. I am backing it up because I find it agitating that smart people on HN look at HFT and assume because they are smart, and because they don’t understand it that it must be bad. If something in the market is happening you don’t fully understand it really pays to sit back for a second and think about it.
(I’m guilty of not doing that with a variety of topics too and I do not fully understand HFT, but I understand enough to say that I cannot say with any certainty that it is bad and lean more to saying it is net good.)
Sorry, crazy premise. If a particular HFT netted less than $100k/year, I'd be surprised if they kept doing it, and I've certainly heard of HFTs who made in excess of $400k/year.
>There are a vast number of market participants who do loose money.
Irrelevant. It's not about whether people ever lose money, but about whether HFTs add any real value to the market. If no one can give me real evidence that they do, then I say that the rules should be changed to make such trading unprofitable; if it went away, then the people who are actually buying stocks as a medium to long-term investment would make MORE money (on average), or at least lose less.
[Citation Needed]
My thesis above is that there IS no value added. You're saying that they are being paid for value provided, but other than possibly causing the price to be updated more quickly (on the 100 millisecond scale mostly, but possibly as much as a few minutes sooner), I don't see how that provides any real value. And honestly I don't see how even helping the prices move a few minutes sooner adds any value.
In ALL the examples in the OA, if there were no HFTs and you simply WAITED a few minutes, then the people actually making trades would end up with better deals. Sure the market wouldn't be as volatile and instantaneous, but in what tangible way is that a bad thing?
So, again, how does moving prices a few seconds sooner provide value? The examples given all assume the market would be completely static without HFTs, which is obviously bogus. HFTs are only interested in stocks with minimum levels of volume, meaning there would likely already be a lot of liquidity, and they won't touch stocks with too little volume, where they'd actually be useful (OP or another HFT, correct me if I'm wrong).
I think extraordinary claims require extraordinary proof. It certainly seems to me and a lot of other people like they're stealing money from valid transactions by taking advantage of a corner case in the system; claiming that "price setting" adds value when in the very examples given they don't seem to add any net value (except when the HFTs are doing it wrong and losing money!) seems like an extraordinary claim. So where's the proof that the market wouldn't actually be better off without them?
Except for the example you choose to ignore, even after it was pointed out to you.
For a third time: If the only example that causes someone to benefit is the one where the HFT is losing money, then HFTs are motivated NOT to help people.
So no, I don't count an example where an HFT screws up as a positive. If the best you can say is that "some of my revenues are redistributed to people who otherwise wouldn't have been able to make a sale because there were no other buyers for a particular stock", well, sorry, but that pretty much proves my point.
You're claiming to be a Robin Hood, but without the whole moral justification -- you're robbing from random people and occasionally giving some of your proceeds to other random people. The latter doesn't justify the former.
Incidentally, if you feel HFTs add no value, why do people choose to pay them? In my example (and in the real markets), there is nothing stopping Fry from posting a sell order at $10.04 and not trading with Leela. So why does he?
A well-run insurance company spreads risk among all of its customers. There are NON-PROFIT insurance companies that do an excellent job of spreading that risk around (Kaiser Permanente spends 95% of its insurance dues actually paying for medical care, for example). When you buy insurance you're not EXPECTING to have a problem, but you want to be protected if you do. When someone DOES have a problem, the insurance company hasn't "screwed up" -- the entire point of insurance is to cover unexpected problems.
The only time that the insurance company could be said to have screwed up is if they're losing money overall. And in that case they'll eventually go bankrupt.
>why do people choose to pay them?
Ignorance? Given how many times I've seen advice to never (or at least only rarely) place a market order, I have no idea why anyone does.
Or possibly it's just the "greater fool" theory: They will sell to the HFT if and when they think they know better than you do.
You're claiming that the HFTs are providing insurance to the people who are buying or selling? I find that to be a stretch, since the entire point is that you look for sure bets and attempt to only sell this "insurance" to people who (in your opinion) don't need it.
And frankly most people trading in the stock markets don't know they're even "buying" this insurance. If they did, then yes, I could buy your argument. But they don't. And the ability to choose makes all the difference in this case.
I'd like to hear from someone trading stocks for a mutual fund, and whether they would "choose" to pay an HFT.
The reason that we don't have market makers standing around in pits any more -- that computers can do the trades directly -- also implies to me that we don't need those intermediaries at all. There's a lot of economic leeching that goes on in Wall Street, and I suspect that much of it is obsolete. HFTs are merely another symptom of the same problem.
The only time the HFT could be said to have screwed up is if they're losing money overall.
I find that to be a stretch, since the entire point is that you look for sure bets and attempt to only sell this "insurance" to people who (in your opinion) don't need it.
No, you look for bets with a 90% chance of collecting a $0.10 spread and a 10% chance of a $0.50 price drop (for a net profit of $0.04/trade across many trades). Actually, the odds are usually much worse than that, if your expected profit/share is greater than a penny, you are doing fantastically well. Any HFT who hunts for a sure thing isn't making any money - there are far more 51% gain, 49% loss opportunities than there are sure things.
There may be far more 51/49 opportunities. However, saying that HFTs hunting for sure things isn't making money is far from true.
I worked for a UHFT firm that had a winning percentage of more than 95% of the time. The 5% losers had VERY small losses too. All arbitrage trading. Not market making like everyone seems to be focused on in here.
Either this logic, that profit is evidence of value destruction, is flawed or capitalism doesn't work.
You assume that the capital markets are a zero sum game. But by promoting liquidity HFTs make the markets (a) more attractive to play in, and, (b) cheaper for companies to finance themselves from.
Strawman. That's not at all what I said.
In this particular case the only time the market really benefits is when the HFT isn't making money. And I didn't say that "because he's making money it's evidence of value destruction." Just that the only example OP was able to point out that was an advantage to the consumer was when the HFT was losing money.
Do you have an example of where sub-100 ms execution truly adds significant liquidity?
I'm open to listening, but have never heard a convincing case for why HFT below such a time threshold adds net value (i.e., enough to make up for the volatility it can cause).
Edit: Changed to "solely germane"
The studies I link to provide compelling evidence that traditional market makers provided much lower levels of liquidity.
Do you have an example of where sub-100 ms execution truly adds significant liquidity?
Please reread my post starting from the words "The Latency Arms Race".
The third link talks about reducing the bid/ask spread. Fair enough. However, what people seem to forget in these discussions is that if you are genuinely trading in the long run, then the bid/ask spread is mostly irrelevant anyway, because it is totally dominated by intraday fluctuations. Whether I buy at the offered price at 12am vs 1pm makes a much bigger difference than whether a HFT reduced the spread by some small amount. I have yet to see an argument that the sum of intraday fluctuations + bid/ask spread is helped at all by HFT.
So one study remains (this one: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1624329). This one might genuinely apply - I haven't had the time to go through it - but the abstract isn't exactly enthusiastic, and probably disregards the social costs that come from people being sucked into the field. So I am still skeptical. Confirmation bias applies as well, I guess ;)
In other words, the better an HFT gets, the less benefit they're providing. This isn't doing much to convince me.
Yes, in that case, it would suck to be Fry. But the problem at that point is that the price dropped before Fry got his order in, and that's a price of doing business in the stock market. Considering that an HFT would do all in their power NOT to be helping Fry in that case, it doesn't really support your position. You're not in this to try to help people like Fry; you're in this to AVOID helping people like Fry as much as possible.
>If you want to take that risk, place an ALO order at the bid or ask price.
ALO? How does that differ from a limit order? I only ever use limit orders when buying or selling stock -- and pretty much every bit of investment advice I've read recommends that you only ever use limit orders, though advice on where to place your limit varies. If "most people" use market orders, well, most people also lose money in the stock market.
Yes. This isn't all that uncommon - see insurance companies, for example.
Without Leela, the problem is not that the market moved before Fry got his order in. The problem is the market moved before his order was filled.
ALO? How does that differ from a limit order?
It's a limit order which Adds Liquidity Only. But it's not even necessary in this case - if Fry is willing to accept execution risk, he can place a sell order at $10.05 instead of $10.00:
Sell(Fry, $10.05, 100) Sell(Leela, $10.05, 100) ---------------- Buy(Leela, $10.00, 100) Buy(Bender, $10.00, 100)
If Zoidberg and Amy each buy 100 shares, Fry's order gets filled. If not, Fry loses money.
Fry could also have placed his sell order at $10.04 and jumped the queue:
Sell(Leela, $10.05, 100) Sell(Fry, $10.04, 100) ---------------- Buy(Leela, $10.00, 100) Buy(Bender, $10.00, 100)
In this case Fry gets filled before Leela, so he doesn't need Amy to show up.
>the market moved before his order was filled.
In what tangible way is this different from "the market moved before Fry got his order in"?
The market for a particular stock depends on there being people who are willing to buy it. Saying "the market moved" blurs the fact that there simply weren't any buyers at that price any more by the time he got his order in.
If we're talking about a stock with a lot of volume, and Fry "jumps the queue" by setting a limit order in the current spread (something I almost always do myself, FWIW, based on that same investing advice I mentioned), then he'll sell his stock with or without HFTs. What value does the HFT add in this case?
The argument is that HFTs add liquidity. Don't HFTs only work with stocks that have high volume? How is adding liquidity when there's already high volume a value?
How does insurance benefit someone who's car doesn't crash? It's a couple of hundred bucks a month for nothing.
What value does the HFT add in this case?
None. If Fry doesn't want to cross the spread and pay for liquidity, he doesn't have to.
Similarly, Arthur's Steakhouse doesn't benefit me because I'm a vegetarian and don't want to buy what they are selling.
Don't HFTs only work with stocks that have high volume? How is adding liquidity when there's already high volume a value?
First of all, many HFTs do work with the long tail of US securities. Second, adding liquidity via price improvement (i.e., narrowing the spread) is useful even if volume is high.
They haven't lost money until they sell at a lower price. The price could still go back up and the HFT would make money overall. But Fry is still better off in this case because he got to sell right away at a better price than he could have gotten if Leela wasn't around.
In this case, everyone benefited: Fry got to sell when he wanted at a better price than the non-HFT's were offering and Leela made money. Leela made money by taking on risk. Leela makes money overall only if she is smart enough to win more often than she loses.
Why do you get to decide that? I could think of several examples where I would be concerned about a trade happening quickly (hedging bets made elsewhere, breaking news affecting the underlying value of a product, etc.)
I can't help but wonder if this level of liquidity is only really useful to HFT, if it is something that HFT is both the primary provider and beneficiary of, and if they're sort of using their existence to justify their existence, so to speak.
Chris uses mom and pop trader examples to explain the concepts here, but the fact is that Algo trading now accounts for much of the volume on major exchanges, and HFT is probably a decent portion of that.
I'd love to see Chris or any other HFT trader's take on that. He sort of does near the end, but would be interesting to see a more in depth discussion:
>Although the latency competition provides little to no value to the end consumers, all HFTs must play the game. If they don’t, a faster HFT will beat them to market and their order flow will be reduced.
This is a very real social cost of HFT - many very smart people spend a lot of time and effort reducing the latency of trading systems. I’m no fool, but when I worked as an HFT I often felt like one. The field contains a lot of very smart people, and the world would probably be better off if the stock market had a little more latency and those smart people were building products for the world. For example, after leaving HFT I built a useful consumer product.
But with current market mechanics, the global optimum where trading is marginally slower and smart people build useful products is impossible to reach. The latency arms race has us stuck in a suboptimal Nash Equilibrium which is similar in character to signalling competitions (e.g., the education bubble).
Yes. It used to be that humans were the market makers. Humans are (generally) a lot more expensive than computers so providing liquidity came at a higher cost. Bid/ask spreads used to be much higher.
HFT it just an example of computers replacing humans at a lower cost. A phenomenon that can be observed in many industries.
I've just read through the entire thread and I don't think anyone has suggested that.
Well, officially some traders in US exchanges are privileged in both information and fees ("specialist" market makers) in exchange for guaranteeing liquidity. I was not following the markets 20 or 30 years ago, but those "specialists" always claimed that they were, in fact, helping the market rather than themselves (it was demonstrably false in the mid 2000s for almost all specialists; I don't know about the 80s).
European exchanges in which no such privileged traders exist were able to pick up a lot of international and american trading volume in the last 10 years thanks to perceived anonymity and fairness. The german exchanges went as far as to slow down data streams to better connected users until they were able to offer the better connection across the board.
Yes. huge problems. 20 years ago (1992) stocks were traded in 1/8ths or 1/4s (minimum spreads of (12.5 or 25 cents) often more than one tick wide. By the turn of the century stocks had all been decimalized (traded in pennies) but spreads were still wide.
This article: page 25 of this paper[1] shows the average stock spread of S&P500 stocks being squished from 8cents in 1997 to less than 2cents in 2010.
[1] www.kellogg.northwestern.edu/faculty/...d/murphy_kim_spread.pdf
The short summary is no one can come up with something better than price-time priority for matching orders.
True, the HFT folks would still try to provide liquidity by maintaining a bid/ask spread. But every trade that executes because someone wanted to trade while the price was somewhere between the bid and the ask would cut the HFT folks out of the loop to the benefit of everyone else. And scary anomalies like the flash crash would be impossible.
After careful thought I am sure that it would not eliminate HFT. But it would reduce their size and impact on the market.
The key point of what I suggested is that the price drifts slowly, and orders can only execute at the current price. In today's market the price can move instantaneously.
That's a pretty big difference.
Part of the problem that others are having understanding you (assuming your system is, in fact, better) is that you don't seem to be giving a full explanation that relates to the actual order book initial conditions and new incoming order flow. I'm not an expert by any means, but I can't understand, concretely, EXACTLY how your proposal is intended to work. It feels hand-wavy to me, which usually means that an idea isn't fully-formed. (I mean no disrespect here; just stating my perception.)
I WANT the price to move instantaneously, even though my only interface is via Etrade's retail and mobile site, and I probably execute 400 trades a year. I strongly prefer instant over a few pennies here and there, even though I'm at a (at least theoretical) disadvantage in terms of market access as compared to the HFTs.
See the first post in the series. This is exactly how the markets work. The best price always wins.
If the bid/ask is $10.00/10.05, and I offer to buy at $10.05, I will trade immediately. If I offer to buy at $10.03, I jump the queue and will be the first to trade (provided someone is willing to sell at $10.03 or lower).
Suppose that the bid/ask is $10.00/10.05. Suppose that I am willing to buy at $10.20.
In today's market I will immediately make the trade at $10.05.
In my suggested market there is another fact to consider, the price. Suppose that it is $10.03. Then I become an outstanding buy order, and for as long as I am outstanding, the price will drift up. If the price reaches $10.05 without finding a seller, then I will trade at $10.05 with the HFT folks. But if a seller who is willing to sell at $9.90 comes along before that happens, we will trade with each other at the current price and leave the HFT folks out of the loop.
My suspicion is that, if this were implemented, a large fraction of trades would actually execute in the middle ground between what the HFT traders are willing to offer as a bid/ask. Therefore this would be even better for actual speculators than the current system. HFT players would still be in that market, but they would not be as important.
In short the difference with my system is that trades may have a small delay before they execute, but should execute at the same or better price for me than the existing market mechanism.
Your idea is definitely interesting. It certainly slows down price discovery, though probably not enough to matter.
I'd need to think more carefully about whether it's a good idea or not, but it's definitely the best suggested tweak to market mechanics I've heard in this thread (or the last one).
So when you say a price drifting upwards, that means that both buyers and sellers meet at that price. A corollary is that there were buyers and sellers at that price. In addition, this also means that the market could have also moved away from that price without a trade being executed (ie after major announcements -- last trade 10.50, next bid/ask 8.75/8.80)
So given that, "If the price reaches $10.05 without finding a seller" this statement does not make much sense and/or is irrelevant.
>then I will trade at $10.05 with the HFT folks. Secondly, there is nothing to differentiate, nor should there be, a HFT trader and an individual trader. A trade is a trade. From a buyer's point of view, it doesn't matter if you buy XYZ from MomPop Co vs UltraHFT Co. A trade is a trade.
>But if a seller who is willing to sell at $9.90 comes along before that happens, we will trade with each other at the current price and leave the HFT folks out of the loop. This is currently true right now. If someone enters a sell order below the bid, the trade is executed at the bid price (or whatever is at the top of the order book). It will knock out all the orders up to the limit price that is set.
Also as I mentioned above, "current price" (ie last executed) has no meaning. In the case of of the major announcement, last executed could be far off from the current order books.
>My suspicion is that, if this were implemented, a large fraction of trades would actually execute in the middle ground between what the HFT traders are willing to offer as a bid/ask. Nothing you suggest is anything different from the current system.
Take that blinder off.
In the model that I suggest, the price is a number set by the exchange that trades are allowed to happen at, that moves in a predetermined fashion. Even if there are buyers and sellers who are willing to trade at a different price right now, the exchange won't allow that trade to complete until the price set by the exchange goes into a range where that trade can happen.
Under this model, which IS NOT how the market currently works, the people who have an opinion on which direction the stock will go will experience trading delays, but they have an improved opportunity of making trades in the middle of the range that market makers would be willing to offer.
Define: Last Trade: amount that the trade last executed. Order: Bid or Ask order on the order books.
So, let's use your example of bid/ask $10.00/$10.05 and last trade of $10.03 You place a buy order at $10.20.
What you are suggesting is that the market sees if there is a sell order at $10.03. If there is, it executes. If there is not, it moves to $10.04. Again, it checks to see if there are any available sell orders at $10.04. If not, it goes to $10.05 and sees if there are any sell orders at $10.05. Since $10.05 is at the top of the order book, the trade executes.
Am I understanding this correctly? We both agree that in this given state, the trade will never execute at any other price other than $10.05.
So, you added that a seller comes along and places an ask order at $9.90. We now have a crossed book (bid greater than ask). So what price does this execute at? You suggest $10.03 because that was the last trade.
I would argue that this would be incorrect.
Let's take the scenario of a low volume security. Let's say it's a far out of the money or far in the money option. If you follow the market, you'll see that often times, the volume of these trades are maybe 1 or 2 trades per day. You will often see that the Last Executed price is often either above the bid/ask spread or below the bid/ask spread.
Take for example, CAT Jan 14 $110 PUT. Last trade is $20.45. Current Bid/Ask Spread is $21.80/$22.15. The last trade is almost 7% discount of current market value.
So, let's use this CAT example. Current order book shows $21.80/$22.15. Last trade is $20.45. A buyer comes in with an order at $25 and a seller comes in with an order at $20. What price gets executed?
>Even if there are buyers and sellers who are willing to trade at a different price right now, the exchange won't allow that trade to complete until the price set by the exchange goes into a range where that trade can happen. Please explain this sentence further. It makes little sense. The only way that buyers and sellers can trade is if they cross the bid-ask spread. If you are saying that two other people are willing to make a trade, that means that they've cross the bid-ask spread. In theory, the "price" that you mention should be in the middle of the bid-ask spread.
Let's use a concrete example: Going back to your $10.00/10.05 - last trade: $10.03 Am I understanding correctly that if someone places a buy order at $10.05 (ie order book $10.05/$10.05; last trade: 10.03) that that trade will not get executed? What happens in that scenario? Does the price rise to $10.04? (ie order book: $10.05/$10.05; "price": $10.04) In this case, the trade does not execute neither, even though two people are willing to trade at $10.05. So the "price" rises to $10.05 and the trade executes. If this is the case, as a trader, I would front run the seller at $10.05 and put a sell order at $10.04. Anyone that sets a long dated limit order will be at an disadvantage; this will encourage speed.
In this type of market mechanism, you'll see more trades move to OTC. If I find another individual that is willing to make a trade, I will do it off the exchange rather than on the exchange.
Lastly, please explain how you would be able to distinguish between an HFT trader and a regular trader? Does it make a difference? Why is it better for individual investors that they only trade with other investors?
>but they have an improved opportunity of making trades in the middle of the range that market makers would be willing to offer. This currently happens as it is. By SEC regulation, brokers are required to find the National Best Bid/Offer (NBBO) when you make a trade. In addition to this, brokers can perform Price Improvement actions and execute trades at better than NBBO (ie buy lower than ask/sell greater than bid). For example, ETrade has an 85.5% rate of price improvement. (https://us.etrade.com/e/t/activetrading/apptemplate?gxml=sca...)
Incidentally, I wrote a post about this market mechanism here: http://www.fatwallet.com/forums/finance/1004723/
Please forgive me if I am not understanding what your are suggesting, but a lot of what you are suggesting seems like it will not work in the real world.
The critical thing is (I think) the concept of price takes time to move, making it possible for use the liquidity that exists without HFT, but still letting HFT cover the gaps.
For the example below, we suppose this new price can only move $0.01 a minute (obviously real thing would be more complex, but this makes things simple.)
Then, the bid/ask is $10.00/$10.05 and the current price is $10.03. By placing a buy order at $10.20, the price will start trending upwards. So, if no-one comes along to trade in the next two minutes, the price will reach $10.05 and the order will fulfill since there are sellers at $10.05. This will clear the buys and the "pressure" on the price will return in to between the bid/ask spread.
On the other hand, if a seller comes along after a minute and places an ask at $9.90, the trade will complete at $10.04, and both traders benefit.
I'm not sure what the implications are but I'm not sure if you understood the time-varying, averaged price concept btilly was proposing.
In the CAT example, I would expect the price to have trended to between the Bid/Ask spread from the last trade price (say it's $22.00) So that's where the order can execute if buyer and seller arrive at the same time.
Again, I have no idea if this is good idea and the mechanics seem over complicated, but thought it was worth pointing out what might be a good idea.
In warfare, it pays to react faster than your opponents. You don't have time to think slowly about how you're going to shoot. You'll just have to shoot accurately and utilize tactics that you already know.
There is no compelling social reason why we should reward smarts in this case.
At any rate, the current system is much better than rewarding the tallest guy with the ugliest jacket.
Now since speed is no longer a barrier to entry, the spoils will go to the people willing to provide the most liquidity.
By spoils, I mean the returns to market making. Right now market makers are largely rewarded for being the fastest and beating out other market makers rather than for providing the most liquidity to speculators/investors.
The markets do not do that automatically. Before automated trading, Alice could sell, but would probably have had to do so at significantly less than $10/shr, even if the market was indicating $10 was the true price of XYZ. Bob could buy, but would probably have to do so at significantly more than $10.05/shr. The primary reason for that: human beings skimming from Alice and Bob, because there was no effective way to compete.
The order book now looks like:
SELL BIG - 1000 $600
SELL SPEEDY - 200 $600
-----
Buy SPEEDY - 200 $599
BUY BIG - 1000 $599
A buy and sell order comes in for 200 shares at market, and SPEEDY makes $200. BIG makes nothing. The problem is that BIG is offering more liquidity to the market (1000 shares vs 200), but all the rewards go to SPEEDY. With a pro-rata, BIG would make $166 and SPEEDY would make $33.
Pro-rata rewards volume, FIFO rewards speed. It is debatable which is better.
It also looks like "Speed" and "Bigness" are orthogonal issues; you're just irritated that "Speed" got to the book first. But "Speed" is getting compensated for soaking up more risk; because "Speed"'s order fills first, "Big" has more time to react to negative market changes.
Note that in any case, traders are penalized in a variety of other ways for laying large block orders onto the book.
SELL BIG - 1000 $600 ----- Buy SPEEDY - 200 $599 BUY BIG - 1000 $599
Another buy order at $600 comes in and BIG makes the trade.
Now the order book looks like this:
SELL BIG - 800 $600 ----- Buy SPEEDY - 200 $599 BUY BIG - 1000 $599
At this point, SPEEDY wants to get back in so he offers 200 shares @ $600. The Order book now looks like this:
Buy SPEEDY - 200 $600 SELL BIG - 800 $600 ----- Buy SPEEDY - 200 $599 BUY BIG - 1000 $599
So now SPEEDY is sitting behind BIG and has to wait until all 800 shares of his order soaks up the market.
So, where is the problem?
What if the exchange traded once every 5 seconds in a blind auction, then there would be no real advantage to speed. Each computer would have 4.5-4.95 seconds to parse the auction results and post new orders.
Would this exchange be more or less liquid? (I don't know, but it is interesting to think about.) The barrier to entry to be a market maker would be lower. The market making code would look more like Poker bot code, rather than what HFT does now.
See this discussion on the Quant Finance Stack Exchange:
Does any of this low-latency work make it back upstream into Linux or other parts of the software stack? As a latency-obsessed person I would love to know that the fruits of all this labor were available to me for my own low-latency systems. To me, that alone would be enough to feel that the latency-race is providing value to the world.
Also, one thing that was not clear to me is how electronic market makers make decisions. The article clearly states that they "have no opinion or information on whether Apple is a valuable company." So what do they have information about? The price history of the stock? The current contents of the book? And on what basis do they make decisions? I guess answering that question in too much detail would be giving up the "secret sauce," but I'd love to know even in broad strokes how an agent that knows so little can so consistently make money.
The article clearly states that they "have no opinion or information on whether Apple is a valuable company."
This is an exaggeration. Some HFT's incorporate speculative mechanisms into their strategy, supposedly a few people buy the twitter firehose and market news and feed that into their strategy. But in general, HFT's run purely technical strategies.
So what do they have information about? The price history of the stock? The current contents of the book? And on what basis do they make decisions?
All this sort of stuff. For example, if the book has 20 buy orders at $10.00, and only one sell order at $10.03, that's an indication the price might be going up shortly. But really, it's not all that vital to predict things.
The HFT is passively waiting with orders at $10.00 and $10.03 for someone else to fill them - they make money when they sell to someone at $10.03, then close their position by buying at $10.00. The trick is to avoid adverse selection - if the market is likely to move up, don't get caught with a sell order.
It was something that needed to be deployed across a large number of machines and the bank in question didn't want to take all of the machines out of the support contract (by using custom built kernels) so they decided an upstream fix was the best solution. The fact it was switched probably means most of the bank's competing firms don't even realize it's there.
The author claims speed is mostly important for order precedence, but I find this to be a naive view. The potential upside for a trader coming to a conclusion and executing a trade first is minute compared to the potential for frontrunning trades that have been executed but not yet communicated to the market. Yes, this kind of frontrunning requires (sometimes[1]) automated illegal eavesdropping on secure financial transactions, and is highly immoral and illegal. But you don't make $100M a day without compromising some morals[2].
Algo I'm okay with, but HFT is a fancy toy for messing with the markets.
[1 | http://zerohedge.blogspot.com/2009/07/is-goldman-legally-fro...]
[2 | http://www.bloomberg.com/apps/news?pid=newsarchive&sid=a... ]
I work at a HFT firm and any serious suggestion of frontrunning would probably get you fired. The truth is the majority of HFT is liquidity providing or some form of arbitrage.
That's a silly assertion. What if it were $100B a day? That's $35T/year, or half the GDP of the world. Do you think that a company could receive half of the productivity of the entire world morally?
I work at a HFT firm and any serious suggestion of frontrunning would probably get you fired.
Another silly statement. Of course people don't talk about it openly - who plans illicit activities openly? I would expect that anyone involved in such a system would have no knowledge of it being used immorally/illegally - there is too much risk in that.
HFT is an assault rifle. People who make them talk about how they keep the peace, but that isn't the use that caused them to be outlawed.
Basically, liquidity providers upload their market making algo to the matching engine. Every time an order comes in, there is a 20ms 'mini auction' in which market making algos compete on intelligence and not on speed.
economics is a bullshit field.
Looks to me like a cryptologist predicting what passwords people would choose based on the needed computing power to crack them. I wonder if 'password' made the list.
Sometimes the assumption that people are rational breaks down, like with sticky prices caused by what psychologists would call "loss aversion". In that case economists make the changes they need to to their models to account for how people differ from the simplest model. This is basically how you get the entire field of Macroeconomics.
Stock market is economically important only when company issues new stock. Apart from that it's just a huge casino where people like to play with their or other peoples money.
I believe that most stocks have far far more liquidity than is necessary for people to feel comfortable investing in them. More liquidity has very little value at this point.
It's a problem because thousands of bright minds are piped into the financial industry when they could be contributing much more to society.
HFT doesn't matter at all to long and medium term investors that actually provide the value in the stock market(which is providing capital to help grow companies).
Just a decade ago you needed to buy into a mutual fund and pay sales loads and expense ratios over 1%.
Now, you can own an ETF containing a piece of each of the Fortune 500 companies for an expense ratio < 0.1% because spreads in those 500 companies are narrow and trade constantly.
Medium term and long term investors love HFT because it saves them 0.9% in yearly expenses. If you've got 100k invested in SPY instead of some Fidelity fund, HFT is saving you $900 a year.
Would you rather I build a photo-sharing site? Chase tenure with esoteric publications?
Do you even know why I do HFT? I do it because it's intellectually stimulating and pays well, plus I work in a small firm of ten smart people and no corporate politics.
As for the value to society you don't believe I'm providing, my arbitraging makes it possible for products like ETFs to exist. Most retail investors would be best served with an index fund, and my work ultimately provides that service.
You're already making the judgement that the work he's doing isn't valuable, finish it up and judge what he should do.
What would a stock be worth the day of the IPO if there was no way to sell it the day after the IPO?
meanwhile, people buy based on their expectation of future price.
So, this is a casino where the dealers are trying to make the chips worth more. It doesn't matter why, this is just reality (generally). But the dealers don't just deal chips all day - they spend real money in the real world, and loads of it. On sales, manufacturing, advertising, research and development, infrastructure, you name it. All to (again, generalizing) increase the price their chips are being traded at. (For various reasons they have incentive to do this, or if not incentive then at least philosophy/approach.)
Say you meet the dealer named (CEO of x). Now if you can convince him that doing y action in the real world will increase the price of his tokens, he might just do it.
Even if it means spending millions or billions of his company's money.
This is the status quo.
Now, what are the results of this status quo? Does it mean that companies will do whatever will increase their share price? No. Does it mean companies will do whatever they expect will increase their share price? No. Does it mean share price realistically reflects a company's investment in doing something to become financially or objectively or in some way convincingly "better"? No.
Nothing follows from this, but there are some things that almost follow.
One of them is that in the real world real money is being spent trying to get real customers, to get real innovations, and so on.
In a way, the precisely discovered price of a casino token is like a carrot on a stick. It points somewhere, (to stretch the analogy it points up or down), and does so as a result of the rider moving it.
In this case the 'rider' is investor opinion. (Alternatively, though false, you can assume that all information in the world is factored into the price, that the price is somehow 'efficient' with respect to all the information in the world about that company that's public. While ridiculous and false, this is/was a dominant or important academic model for a long time.)
What I'm getting at is that it's a lot simpler than just being tokens. There is a complex system here that you can draw some generalizations from. I mean, to take nothing else, just remember that when a company's "token" price gets too low, its investors will vote for an acquisition in cash by a bigger company in the same industry at a premium over its "token price". That would have a very real effect in the real world. It's just so much more complicated than you give it credit for. Is it good? Should we have it?
Well, all the recent private equity activity says, you know, not always. But there are some interesting things that happen when you put your company into the 'token world.' It's most certainly not just tokens.
Because they own some stock and/or because that's what in their formal/informal job description.
When I said "casino" and "tokens" I had in mind randomness of the game and that value of the stock is virtual value that humans assign to it in context of the game they play.
Apart from that rules are completely different from any game played at the casinos.
I don't think that anyone who makes decisions at Google actually does anything to increase their share price. And in my opinion that's because they get that in stock market there's a "funding part" and "gambling part" and what they were interested in was the funding (it's basically free money plus some transparency that builds up the trust of your customers). Gambling does not concern them.
> One of them is that in the real world real money is being spent trying to get real customers, to get real innovations, and so on.
Yes. But that real money comes only from the moment when company introduced its tokens into the casino. What gamblers do between themselves after that should not concern company in any practical way. Unless they did something stupid like putting more than 51% of their stock on the market.
Can company even take part in gambling of their own stocks? Isn't that insider trading?
> I mean, to take nothing else, just remember that when a company's "token" price gets too low, its investors will vote for an acquisition in cash by a bigger company in the same industry at a premium over its "token price". That would have a very real effect in the real world.
Yes. There are some points of contact between "funding part" and "gambling part" but they are in place just to spice up the game. I don't think they are healthy for the companies because running company is about physics and increasing market price is about fooling people into thinking that you do more than you actually do.
(b) The primary market cannot exist without the secondary market because no one will buy into your IPO if they don't think they'll be able to turn around and sell the shares at some point. This is obviously the case for securities that generate no income (stocks that don't pay dividends, etc.), but it is also true of other securities for a slew reasons. This is the core of the the complex and amorphous notion of liquidity.
(c) By publicizing the value of the underlying assets, prices in the secondary market inform the wider world about the performance of the managers of those assets. Investors who believe those managers are doing a poor job can put up their money, throw the bums out, and bring in someone who will do the job properly.
All of these things support the efficient allocation of capital and ultimately shape the behavior of all economic actors. Capitalism requires that this mechanism works fairly well.
There are people that have money and people that need money and stock market is a great mechanism that tricks people who have money into parting with some of them temporarily or permanently and funding the people that need money.
Companies get money from selling their stock and use it to fund their actions.
People with money buy their stock and between themselves use their stock as casino tokens to legally gamble.
Without the gambling part there'd be much much less interest in giving money to the companies. People would have to believe that your company will grow. With stock market people just have to believe that there will be some suckers that will buy the stock for more money. (the thing you said in (b))
Most investors are in not to own a part of the great company but just for the gamble.
I just think that price of the token is mostly meaningless from the point of view of the companies that issued them. If you own 51% of the tokens you don't care how much people are pricing them because you own the actual company and you care mostly about the company itself. It's not like you could just dump your 51% on the market without making the sky fall.
Price has some importance if the company wants to get more money by issuing more tokens and it's also very important for various managers who got some of the tokens as their compensation. I'm not sure if that last one fact is a good thing. Managers should concentrate on their jobs of running the company not on making an impression that their company will do well in the future but you can't effectively ban them from playing. They would play anyway via proxies.
I believe that price of tokens does not say almost anything about companies condition. It only says things about random peoples opinion on the company future condition which I think correlates very weakly with actual current condition and the actual future condition.
Determining the price of token more precisely has no more value to people not involved in playing this game of buy/sell than precisely determining the value of WoW items.
Stock market gambling just switched from game played by people to game played also by bots. I think it's a nice thing that there's a place both for humans and bots in this game. I don't share the opinion that people that use bots are somehow cheaters. They just play they game the way they like and don't seem to be destroying it in the process.
1. Dividends.
2. Stock buybacks.
Didn't Fry sell 100 shares to an anonymous counterparty?
However, the last example may be more realistic than you suppose. DEF CON last year contained a presentation[1] on security considerations in HFT networks, which I attended. It's thought-provoking and something I hope gets a little more attention.
(I have a lot of opinions on this subject but I am mercifully restricted from sharing most of them owing to professional obligation; we do a lot of work in this field).
Again: strongly recommend _Trading & Exchanges_ by Harris. The TCP/IP Illustrated of markets. Supremely readable.
To be fair, a lot of exchanges mandate specific order validation pre-checks and specific firewall hardware.
Some relevant questions that are completely glossed over:
1. What compels an HFT to actually trade? Is there anything forcing them to keep supplying liquidity even if the market's moving against them? How is an HFT different than an actual market maker?
2. How does an HFT decide that it has a better-than-even shot at turning a profit on a trade? Most of the objections to HFTs revolve around the answers to this question (i.e. pseudo-front-running by trying to detect large buys/sells that get split over lots of orders) and their implications (i.e. 'real' investors leaving the exchanges).
3. The "market-maker strategy" HFTs you describe are indisputably compensated for providing liquidity and taking on risk, but is the return on HFTs actually equivalent to the return on other investments with equivalent risk? If not, and they earn a premium, why isn't that evidence that something's broken?
That's all it's intended to be.
Is there anything forcing them to keep supplying liquidity even if the market's moving against them?
No.
If not, and they earn a premium, why isn't that evidence that something's broken?
Wait for part 3. I do believe HFTs are capturing fractions of a penny in rent and I have an idea of how to fix this.
Anyone can start a market. If HFTers were stealing pennies someone would have come along and started a market that banned HFTers (or changed the rules to otherwise get rid of them). Then all of the speculative investors would use this market instead since they could be assured that no one was siphoning off pennies on every trade. Eventually markets with HFTers would shut down due to a lack of customers.
The fact that this hasn't happened has to be considered evidence that the HFTers are, in fact, providing value to speculative investors not the other way around.
This is correct. It would be extremely, unequivocally forbidden under Reg NMS to create a quote-disseminating market center with the proposed property.
There are market centers that prohibit high frequency traders. Reg NMS requires any center that disseminates quotes to prohibit trading through the NBBO (best price at any point in time across all exchanges). So these market centers must be dark pools, not exchanges nor ECNs.
These centers represent a small and shrinking amount of trading volumes. Nonetheless, they exist.
For example, I believe the BATS IPO that happened last month began trading around $15, and was below $1 in about 900ms before trading was halted. (Nanex.net has news postings about these kinds of things, and I wouldn't mind hearing other people's opinion about the site)
I guess my question is, is it possible for traders to make money in the short term (hours/minutes) or is that forever in the land of robots now?
I can't speak directly to the BATS IPO debacle but I can say that when the NBBO in a symbol is locked or crossed (market speak for the bid and ask being the same or being inverted) matching engines can ignore the NBBO. This results in trades being executing at prices all over the map and can explain how a stock can go from $15 to $1 almost immediately.
I would go past this and say that anyone who did come up with a "logical" argument would be ignoring time value of money.
I presume it was designed to ensure fairness, but is it any more 'fair' than random selection (true random, from a physical source that cannot be influenced) from all available matching bids.
Would anyone care to present an argument on how FIFO matching, after best price, is beneficial in the broader sense?
If, instead of employing first come first serve you use a random selection then both of us have an incentive to tell the market that we actually want to sell more than 100 shares to increase the odds that we'll actually get to sell what we want. Various bad/unpredictable/unstable things start to happen in this sort of situation.
FIFO matching leads to a stable market.
Silly example I know but FIFO matching is consistent with this idea of fairness. Without it it's possible that market participants could keep getting jumped by the random selection such that they never get an order execution.
Yes, it does mean that whoever had a sell order at $10.10 gets their order filled faster. But is that trade off worth it?
Absent the Trojan, Bender has no idea if Farnsworth will show up or not. If Farnsworth doesn't show up, this might happen:
BUY(bender, $10.10, 100, 12:00:00.000)
SELL(bender, $10.15, 100, 12.00:00.100)
...crickets chirping...
SELL(hermes, $10.05, 100, 12.00.10.000)
SELL(zapp, $10.00, 100, 12.00.15.000)
BUY(amy, $10.00, 100, 12.00.16.000) (Amy trades with Zapp)
...MomCorp continues it's downward slide.
I.e., without the trojan Bender is just a guy who thinks MomCorp is going up.
1. Productive is Destructive ... literally (that's why we're in the position now of having figure out a way to mine asteroids -- industrial productivity under the current model is environmentally unsustainable)
2. Markets are (viewed in aggregate) evolving to be something akin to a global autonomic nervous system ... money like transduction makes the world go 'round.
Here's what that incontrovertible reference, the Wikipedia, states:
Another elegant definition of liquidity is the probability that the next trade is executed at a price equal to the last one.
The Wikipedia also "confuses" liquidity with volume: The liquidity of a product can be measured as how often it is bought and sold; this is known as volume. Source: http://en.wikipedia.org/wiki/Market_liquidity
To illustrate with arbitrary numbers, if HFT’s currently focus on 50% strategy (i.e., price competition and liquidity improvements) and 50% latency, a more optimal scenario would be 50% fewer HFT’s focusing 100% of their efforts on strategy.
http://www.chrisstucchio.com/blog/2012/hft_apology2.html#com...
We ran into a situation where the market was already volatile, and a bad trade exacerbated the issue by causing a number of HFTs to take unexpected losses and withdraw from their markets, consuming further liquidity while driving prices down, which created more losses for the remaining market-makers, who had to close their positions, consuming further liquidity, driving prices further down; all of those also negatively affecting long-term investors.
This also seemed to me an example of the opportunism of HFT, where the HFT shaves the spread by a penny or two during calm markets, but withdraws (and exacerbates issues) during volatile and troubled markets, which seems to me the point in time at which liquidity provision is most valuable.
I'm not suggesting that HFT should be outlawed, nor that HFT firms should be forced to register, act, and be regulated as official market-makers, with the associated duties.
But I do note the benefits seem to come with costs.
Long-term, I doubt it matters. It seems inevitable that the provision of liquidity will become commoditized, and that the days of concerns about flash crashes will eventually disappear into the past along with $50 retail trades, and $0.50 bid/ask spreads.
http://online.wsj.com/article/SB1000142405274870395760457527...
Also, the main reason many HFTs pulled out of the market is the risk of broken trades (regulatory risk [1]). Staying in the market would have been a big moneymaker absent that risk - spreads were often huge.
But broken trades were dangerous. If you buy accenture at $1.00, and sell at $30.00, you've helped fix the flash crash. You also just lost $10.00 - your $1.00 trade was broken, and you now have a short position you bought at $30.00 (Accenture recovered to $40.00, so you lost $10.00/share).
[1] Regulation by the exchanges, not the SEC. Maybe there is an SEC regulation mandating they do this, but I have no knowledge on that point.
Of course you can make a huge profit when a crash occurs, whether that crash is due to a vicious circle of algorithms, or a vicious circle of human psychology. There are algorithms out there that look to do just that, trying to profit from exploitable market anomalies, and it's great that those people have found a way to get paid for fixing some problems. But they're not HFT market-makers, they're a different group of quant/algo traders.
Personally I care little about HFT. The flaws in the technology will get ironed out; the competition for the low-hanging fruit will continue to intensify, and eventually many of the functions will become commoditized as they mature.
In the meantime, I think it makes sense for HFT market participants to be sensitive to the fact that many individual market participants have trouble identifying the value they've received because of HFT participation, but can clearly remember fears that have been induced by HFT driven events.
The HFT could have placed a passive buy order at $1.00 and a sell at $30.00 (or at $2.00, which he revised upwards as the price corrected).
The SEC/CFTC report on the 2010 Flash Crash describes how it happened. Quote:
One key lesson is that under stressed market conditions, the automated execution of a large sell order can trigger extreme price movements, especially if the automated execution algorithm does not take prices into account. Moreover, the interaction between automated execution programs and algorithmic trading strategies can quickly erode liquidity and result in disorderly markets.
http://www.sec.gov/news/studies/2010/marketevents-report.pdf
Yes, some people living on the edge of the market, picking pennies in front of steam rollers, probably had a few days shaved off their life expectancies due to adrenalin spikes, but I could really care less.
The most substantial costs were probably borne by unlucky individuals who had stop-losses that executed solely due to the downward spike, and afterwards found themselves facing a loss, and possibly a tax bills, but I have no idea how common that situation was. I do know that one of my close friends booked his career best day that day, so surely there were also some people who took substantial losses because they didn't expect the problem that occurred.
HFTs thrive in volatile markets. not calm markets. Those HFT firms that stayed in the market during the flash crash made a killing.
Is the problem that the nbbo is changing too quickly?
http://news.ycombinator.com/item?id=1564445
Note that your article provides no explanation as to who is being defrauded ("markets" isn't a person) or how, so I don't know what you want me to explain.
Or algos that are actively working other algos.
The only good reason to have marketplaces is to get efficient price discovery. Explain to me how HF algos stuffing quotes helps to do so. Explain how this isn't just a way to salami slice the cumulative market orders for any given stock?
Explain to me how the heat generated by Amazon's data centers helps bring packages to my door.
The answer is that it doesn't. It's a harmless side effect of the process that does.
The problem could be solved if there was a latency buffer on quotes being placed and canceled on the market. If HF algos had to wait 100ms before cancelling a buy or sell then quote stuffing would disappear. Tell me why having to wait a tenth of a second before cancelling a quote is horrible.
Neither you nor zerohedge has offered any evidence it has any effect at all, harmful or not.
Stuffing quotes causes the bid to oscillate between 10.00 and 10.01 rapidly for a fraction of a second. So what? If it scares you, wait 3 seconds for things to slow down and then place your order. Or if you want to be daring, place your sell order at $10.01 and hope you get filled at that price. Worst case you gasp don't get filled at $10.01 and have to change your order to $10.00 (omfg, no!!!).
Tell me why having to wait a tenth of a second before cancelling a quote is horrible.
Why do you believe I think this would be horrible?
I have no opinion on minimum quote durations - it would widen the spread somewhat (by raising the risk and therefore the costs of HFT), but I don't think it would matter very much.
To go back to my analogy, it would be like demanding that Amazon put marginally better air conditioning in their data centers. It would have a small effect on the evil hellish heat they create (don't you know that satan loves heat!), while possibly raising the price of goods they deliver by a tiny amount.
It may be legal to something, but unethical. It may be ethical to do something, but illegal. There are many examples of both. The reason for this article is apparently to spread understanding amongst people who think it's unethical to do HFT, but may not understand what HFT does and how it works. By writing this article, he is hoping to convince - via understanding - people that HFT is also ethical.
"Apology" doesn't necessarily imply remorse, just explication.
1) Apologia does not meet "I'm sorry". Read the first few sentences of the article for an explanation.
2) I shudder to think of a world where everyone felt the only criteria for whether it was all right to do something was legality. We should all be grateful that most people don't go through life with that mindset.