If you've missed my previous posts on the Crash of May 6th, please go back and check them now - there are also many intelligent comments in the comment threads. The Best Ten Minutes You Will Spend This Weekend, Canceling Trades, Possible Triggers & HFT, and WTF Stocks?
I want to take one more post to explain what one of the problems was during the crash - and I think this will help with potential solutions, and with understanding what went wrong. First, we have a lot of different market centers now. The market is fragmented. That means that you can trade IBM on the NYSE, or alternative exchanges, ECNs (electronic networks) or dark pools. Reg NMS "protects" the NBBO - national best bid or offer. That means that regardless of what exchange (I'll use the term "exchange" instead of "market center" because it's easier, although these alternative market centers are not all technically "exchanges") you trade on, you have to trade at a price that is at least as good as the best bid or offer on all of the exchanges. This is a good thing.
The NYSE, however, has a provision called LRP - Liquidity Replenishment Points - which is designed to act as a sort of market speed bump during times of significant price movements. Many people are familiar with "trading curbs" we have for the major indices, which can result in altered or halted trading if the major indices move by a certain amount intraday. Well, the LRP is like a trading curb for an individual stock - it temporarily changes the stock from automated electronic matching to specialist controlled matching - the NYSE specialist steps in and tries to slow things down a tad while he sources liquidity and tries to pair off orders. In the NYSE's own words:
"A volatility control built into the Display Book to curb wide price movements resulting from automatic executions and sweeps over a short period of time. When triggered, LRPs automatically convert the market temporarily to “slow” or Auction Market only mode, allowing specialists, floor brokers and customers to supplement liquidity and respond to the stock’s volatility. "
This is also, potentially, a good thing. If people are worried about automated electronic executions - and it's pretty clear to me that people ARE worried - then the LRPs are a backup plan that take over in times of stress. Interestingly, the trigger thresholds are not very large - they depend on the price of the stock in question and its average daily volume, but for a stock like IBM the LRP would be triggered with a move of only $1 (that's not a $1 change on the day, it's a $1 change in a hurry).
There's one key issue, though: when the LRP is triggered, the NYSE's quote (behind the scenes - you wouldn't see it, but the market centers who "talk" to each other to make sure that you get the best price would see it) gets a special tag on it, essentially labeling it as "slow," and as a result it's no longer protected by Reg NMS as part of the NBBO. Essentially, all the other market centers and exchanges no longer have to honor the NYSE's prices, and can simply trade "through" the NYSE's bids.
Now, this is a problem. If we're going to have trading curbs in individual stocks, they need to be standardized across all trading venues! Furthermore, it's clear to me that since the NASDAQ wants to cancel trades that are out of line with prevailing prices, we do need individual stock curbs. Why on earth should the trades be allowed to happen at all if they are then going go be canceled? Be proactive - prevent the trades from happening in the first place!
What's funny is the back and forth between the exchanges in the aftermath. From the NY Times:
"The absence of a unified system to halt trading in individual stocks led to bitter accusations between exchanges on Friday. Robert Greifeld, chief executive of Nasdaq OMX, appeared on CNBC to criticize the New York Stock Exchange for halting trading for up to 90 seconds in half a dozen stocks on Thursday.
“Stopping for 90 seconds in time of crisis is exactly equivalent to not picking up the phone,” Mr. Greifeld said."
A few minutes later, Duncan L. Niederauer, chief executive of NYSE Euronext, responded in an interview on CNBC, blaming Nasdaq’s computers for continuing trading while the market was in free fall.
“These computers go out and just find the next bid they can find,” he said.
Mr. Niederauer acknowledged the need to introduce circuit-breakers along the lines of those already in place on the Big Board, and his views were echoed by some chief executives of the new exchanges.
I think it's likely that we'll see some sort of standardized individual stock trading curbs instituted as a result, which seems to be a pretty reasonable solution. I can only hope that curbs will be temporary (and by that, I mean that they will be in place for only a short time after they are triggered) and balanced - so that they aren't designed to put an extended halt to trading if stock prices fall, while continuing to allow trading if stock prices rise.
-KD
21 comments:
Seems like 95% of the issue was just an event that no one had really planned for or foreseen happening. In retrospect it seems really dumb that people could miss this, but it's not really surprising whatsoever. It happens all the time. Dumb things cause huge problems again and again.
I do agree that rules preventing trades that would ultimately be cancelled should be in place. I imagine the lack of these rules is more like a result of the exchanges having never even entertained the possibility that the events could have occurred.
Very insightful Kid - I just came over from Barry to read this and it makes sense and is the first explanation I've read that makes some sense - but here's the question: if the moves to go into slow mode are relatively small at the NYSE, why haven't we seen this before? Secondary question - I get the point about the national best bid, but why would all the other exchanges bid down so far? A stock like ACN was trading at 0.01 cents. I assume it means that there were no offers? If that's the case, then why do we even call the HFTs that trade on these exchanges liquidity providers? If they're not willing to make a market at any price, then what is the point?
Great stuff again! (I'm inthewoods on Barry's site)
hi damian - 1) why haven't we seen this before? because normally you don't have a massive selling wave, and the problem is quickly self remedying. 2) i think you meant that there were no bids in ACN - not "no offers" - and yes, that's precisely the problem. the order books were very thin - there weren't any bids on the book, so when a sell order comes in and slams through all the bids, it's bloody.
you can't FORCE people to provide liquidity. Our markets have evolved into a variety of competing market centers. market centers compete for liquidity - ie, - ORDERS - by offering rebates to liquidity providers.anyone who posts a bid or offer is a liquidity provider. anyone who HITS a bid or LIFTS an offer is a liquidity taker.
it's like saying "hey - if you provide liquidity on our exchange, we'll pay you a little bit!" it's incentive...
remember, these market centers/exchanges are nothing more than a collection of bids and offers. when the shit hits the fan, bids and offers disappear - liquidity dries up. you can't legislate against that.
Thanks for the follow-up - my understanding of the relationship was clearly wrong - I thought they agreed to provide market liquidity in exchange for having access. If they don't need to provide liquidity during tough times in the market, we can do without them altogether - these guys are basically allowed to print money on these operations. Obviously the counter-argument is they reduce costs by closing spread, but I'm not really buying that. Just my opinion!
i can tell you that in my opinion there has never been a better time for the individual investor to trade cheaper or easier than currently.
Damian> If that's the case, then
Damian> why do we even call the
Damian> HFTs that trade on these
Damian> exchanges liquidity
Damian> providers?
There are 2 different threads of control battling each other:
1) The HFT systems that would see a lack of bids on quality names and race to place them.
2) The trading accounts of people blindly selling with market orders or stop loss orders (which are market orders).
What I suspect happened (following on moments after KD's explanation ends) is that some meaningful trigger point on stop loss orders was exceeded. This could have been a small wave of selling from Bloomberg running the video of the crowd getting agitated in Greece (which was at about 2:40PM EST), but whatever the case - a wave of selling started. That in turn brought the price down, which triggered some stop loss orders, which in turn fueled more stop loss orders, along with any humans and machines that just sold on the steep drop.
However, given the heavy volume at the time, the HFT systems that would normally jump in (albeit at much lower bids) didn't even get to see accurate representations of the order books, because I was seeing at least a 100ms delay in quotes from ARCA (the only ECN I measured accurately).
So, at least with ARCA and probably the other exchanges as well, everyone was running with at least a 100ms delayed snapshot of the world. Given that I stopped calculating this delay when my own software shutdown at 2:41PM (4 minutes before the peak of chaos), this is probably understating matters somewhat.
If you can't see that the order book is missing bids because you are operating 100ms behind the actual trades taking place, then there is a meaningful window when the bids in the order book can all be taken out before anyone even knows that they should be placing bids!
Further, once you recognize that you are operating with stale information (and 100ms is quite stale if you are seeing the markets plunge the way they were), there is no way you are going to enter orders, since you don't have any clue where to place them, and if you do - you place them with much wider spreads than normal, which in conjunction with market sell orders brings the trading price down along with the bid/ask midpoint.
While I don't think this view will get much airtime in other forums, I think one of the major problems was not only the impedence mis-match between NYSE and the other venues, but also that the other exchanges (or at the very least ARCA) were too heavily bogged down by the order volume to send out updates to the order book in a timely enough fashion for the liquidity providers to react.
Or in other words, we need to make everything run faster with more headroom (additional bandwidth, many more matching engines at the exchanges, additional multicast quote distribtuion channels), so that the HFT systems that provide liquidity can have an accurate snapshot of where the order book is empty as quickly as stacked up market orders can drain it.
There will be all kinds of calls for things to be slowed down... and that is exactly the opposite of what needs to happen here. Either market orders need to be entirely removed from exchange matching engines, or else things need to get much faster during peak volume periods to avoid a repeat.
-PeterPeter
KD> i can tell you that in my
KD> opinion there has never been a
KD> better time for the individual
KD> investor to trade cheaper or
KD> easier than currently.
That is certainly the case. Retail investors who rail against HFT do not seem to appreciate that the 1 penny bid/ask spreads they pay on most issues is relatively new, and that not too many years ago, you would routinely pay 1/8 of a dollar on just the spread.
Additionally, the massive trade volume and competition spurred between the exchanges by Reg NMS have reduced costs at the exchange to the point where a retail investor with a small account can get commissions as low at 1/2 penny per share at places like Interactive Brokers.
I get why large institutions trying to move large block orders with ancient technology outgunned by HFTs don't like the current trading environment. I also get why old style brokerage houses where you call your broker on the phone to place a trade and he rips you off with a high commission (I'm thinking of Themis Trading) don't like the modern trading environment (since it takes away their lucrative skimming off of the top).
But for the retail investor who's orders can't be anticipated (because they are "one and done"), there has never in the history of securities trading been a time as frictionaless as today's US equity markets provide.
Take away portions of Reg NMS, HFT, require orders posted to rest on the books a minimum amount of time, tax cancellations or many other poorly thought out ideas... and what you will see is a rise in bid/ask spreads as well as an increase in commissions charged by the exchanges and passed on through the brokers.
-PeterPeter
Very interesting follow-up Peter - I wonder if the HFTs were seeing similar delays given that they are usually co-located with the exchange? I know they measure access to the exchange servers in 5-10ms increments - of course if the servers themselves were overwhelmed, they would have seen similar bids.
Kid - I somewhat agree with you on the cost of trading - prior to HFT dominating, my costs were roughly the same through a broker like Interactive Brokers and/or Tradestation (back to 2003/2004). Now you could argue, fairly I think, that they continue to bring down the costs of trading - I'm not sure that they change that game that much - trading costs on exchanges pretty much naturally go down. However they might be speeding it up!
Overall, however, I think Kid makes an excellent case against blaming the HFTs, when it seems to more clearly be a problem of different rules at the exchanges.
"Retail investors who rail against HFT do not seem to appreciate that the 1 penny bid/ask spreads they pay on most issues is relatively new, and that not too many years ago, you would routinely pay 1/8 of a dollar on just the spread."
I think you assume that all retail investors invest via individual stocks - this isn't the case - most invest through institutions via mutual funds. Therefore, it is possible that the retail investor is being hurt by these operations.
Damian,
My non-co-located server (it sits in my brokers server room) gets quotes from ARCA with about a constant 4ms of network latency round trip (~2ms in each direction).
Each quote generated has a timestamp from the perspective of the ARCA clock, so that the receiver can measure how old it is (you see a bunch of packets, find the smallest change in clocks, and take that as the skew between their clock and your own for future calculations).
So, if I was seeing things show up ~100ms behind, then anyone co-located was perhaps seeing things ~98ms behind.
Regardless of the exact numbers, things were very slow, which itself inhibits liquidity creation (eveyrone was acting blindly on very stale information given the rate of change, so just like wandering around your room with the lights off, you will move more cautiously and slowly).
Co-location just means noone is getting the quotes faster than you are, so you can remain competitive at the fastest strategies - but it does not mean that you can get data out of the slowed down machines at the exchange any faster than they are capable of spitting them out.
> I somewhat agree with you on the
> cost of trading - prior to HFT
> dominating, my costs were
> roughly the same through a
> broker like Interactive Brokers
> and/or Tradestation (back to
> 2003/2004).
Just because the media has picked up on HFT (a new term frankly) in the last year or so does not mean that it is new at all.
In 2003/2004, there were plenty of high frequency funds trading.
There were funds doing statistical arbitrage measured in the milliseconds back in the 90s.
There are more funds now and things continually get faster and faster, but the odds that your orders in 2003/2004 were being executed with a contra-side party that would today be called a HFT are quite high (even if noone then was making a big deal about it).
DE Shaw, Getco, GS, MS and many others have all been at this for a long time. The fact that they now have more competition is good for everyone (except the early entrants).
-PeterPeter
Damian wrote:
"I think you assume that all retail investors invest via individual stocks - this isn't the case - most invest through institutions via mutual funds. Therefore, it is possible that the retail investor is being hurt by these operations."
Well, then the additional costs of trading that the funds pay due to HFT is the least of retails problems. Overwhelmingly, mutual funds under-perform SPY, and the added costs of HFT order anticipation as passed through to retail investors pales in comparison to the fees being paid to the (on the whole) lousy managers, and their marketing fees (google 12b-1 fees).
Retail gets screwed in many ways, but none greater than shoddy investment managers taking a piece of the pie, when they are underperforming (again, as a whole) the unmanaged indices.
It is like worrying about a paper-cut on your finger when you have slit your wrists.
-PeterPeter
PeterPEter - tremendous feedback and insight on ARCA latency. thanks. one thing I would note, however, is that BATS (I think it was BATS) put out a press release saying that they had no systems capacity issues at all.
"Just because the media has picked up on HFT (a new term frankly) in the last year or so does not mean that it is new at all.
Great info on the situation first of all - clearly a server problem at the exchanges.
"In 2003/2004, there were plenty of high frequency funds trading."
I've known about HFT for a long time - while HFT has been around for a long time, it has never dominated the volume like it does today. In the 90s it might have made up 10% of total volume as opposed to the 70% it makes up now.
"
Well, then the additional costs of trading that the funds pay due to HFT is the least of retails problems. Overwhelmingly, mutual funds under-perform SPY, and the added costs of HFT order anticipation as passed through to retail investors pales in comparison to the fees being paid to the (on the whole) lousy managers, and their marketing fees (google 12b-1 fees)."
Sorry but I'm going to call you on a "moving the goalpost" argument - the argument you made was that HFTs didn't hurt the retail investor, not the extent of the pain. While I completely agree with you about mutual funds (and ETFs are doing damage to them for the reasons you outlined), that doesn't mean that HFTs don't hurt them. In fact, I would argue that the evidence is very, very strong that it does hurt the retail investor. Goldman is able to win 70% of the time on trading operations and pull billions from the market using HFT (obviously not all their trading is HFT) - and they are just one player. These billions don't come for free - they have to come from somewhere, and I would argue they come from the retail client invested in mutual funds, either via 401ks or pensions.
Either way, thanks for the info - great to hear from someone clearly on the frontlines!
Damian - put it this way: i'll take $9.99 Etrade commissions and 1c bid/ask spreads over $39.99 commission and 25c bid ask spreads that we saw 10 or 15 years ago. that is HELPING the retail investor, any way you slice it.
if your mutual fund manager trades like an idiot and gets scalped by other market players, fire him.
KD> one thing I would note,
KD> however, is that BATS (I think
KD> it was BATS) put out a press
KD> release saying that they had no
KD> systems capacity issues at all.
Well, I'm not sure that ARCA would say they had a capacity issue either... The volume of trading was self limiting by the maximum speed of the matching engines, and their ability to spit out quotes.
On any given day, ARCA slows down at some points and delays quotes by maybe 20ms (9:30AM and 3:40PM-4PM are frequent). For just a subset of securities during normal periods of volatility, that's easy enough to program around and build extra margin of safety into the model.
But, 100ms during such wild market girations is obviously quite different.
Regardless of whether BATS was having issues or not (sadly, I only have accurate timing information for ARCA), you can't really trade with accurate information if ARCA is slow, regardless of what BATS is doing.
If you are following the raw exchange feeds rather than NBBO as put out by SIAC (which I'm sure was next to useless for trading off of at the peak of chaos, since SIAC is always behind the raw feeds by a fair bit), you need to build a consolidated view of the trading books across all exchanges, and if ARCA is delayed materially, then you are still left with materially deficient information even if Nasdaq, BATS, Nasdaq OMX BX, EDGA, EDGX, NSX, Lava etc... are all performing better.
This problem always exists to some extent, as there is always an exchange that is spitting out more stale information than the others, or one where you have a faster cross-connect, but as the latency of one exchange (and a major one at that) increases and the volatility increases, there is very little one can do to trade safely and provide liquidity unless you use huge spreads.
And noone can ignore ARCA, especially once NYSE has stopped electronic trading.
-PeterPeter
"Damian - put it this way: i'll take $9.99 Etrade commissions and 1c bid/ask spreads over $39.99 commission and 25c bid ask spreads that we saw 10 or 15 years ago. that is HELPING the retail investor, any way you slice it."
Two final points:
1. You're assuming causation - that is that HFT caused bid/ask to compress - when likely there are multiple factors. I'm not saying it's not a contributor, and perhaps a large contributor, but it isn't the only thing going on in the markets - electronic trading, it seems to me, is probably the biggest contributor.
2. You're not addressing who is getting taken by HFTs. 401ks and pension funds (still the largest retail investing group in America) are the most likely victims. These retail traders are significantly restricted in what they are allowed to do. To begin with, most are limited to mutual funds by their companies. Secondly, most never pay transaction costs, so they don't care about bid/ask spreads. They trade only on EOD.
Look, for me (an active trader), the compression of bid/ask is great, so I'm not an enemy of HFT - but let's not close our eyes and say that they only add value to the market.
Put another way, if the money opportunity wasn't so huge, we wouldn't have hundreds of firms chasing the billions in dollars it generates. That money doesn't magically appear from somewhere without being taken from someone else.
ok Damian - but you just described yourself as an active trader. every dollar you make trading is at the expense of someone else. every dollar ANYONE makes trading is at the expense of someone else.
should we ban you from trading? where do we draw the line? How can you be taking advantage of all those widows and orphans like that? the NERVE of you! you should be ashamed! (note: sarcasm!)
i worry that this will quickly deteriorate into a tangential diversion that I've already written several thousand words about. Please go back and read my posts from last year about high frequency trading - they are on the right and sidebar under "finance must reads"
"should we ban you from trading? where do we draw the line? How can you be taking advantage of all those widows and orphans like that? the NERVE of you! you should be ashamed! (note: sarcasm!)"
A fair point - I think the difference is the access given. If it is me trading against other people or machines on fair playing field, then I'm down with it. If the HFTs are given an unfair advantage, ala flash orders and other techniques, then I think that creates an unfair playing field. So I draw the line to where everyone has the same access to the market. If we have that currently, and I don't believe we do right now (hence the SEC review of flash orders), then I'm fine with it. But I will definitely look at your other articles to review the arguments. Thanks for the conversation!
damian - yeah - i addressed both the topics you just mentioned at length. please read my other old posts. to summarize: ANYONE can colocate and anyone can run a HFT platform - if you're good enough, talented enough, and can make the investment. in other words, it's not just a secret club that only certain guys can join to get privileges.
flash orders: probably the biggest non-event but most overhyped term that the media has absolutely ZERO clue about. read my post about it - should be clear.
best,
KD
please see my comment on seekingalpha.
The question I have about the offer price going to $999,999.99 seems to be one you could answer. Please have a go at it.
Thanks,
TallD
Much has been said about HFT systems and the so called benefits of the emergence of dark pool "liquidity providers" to the transaction costs of retail investors. Spreads have narrowed commissions have been reduced on an individual investor single transaction level however has it on a market wide level. The specialist system while a lisence to steal was a highly regulated den of thieves. The brave new world of turbo charged HFT trading engines with its "arms race" of MIT grads designing systems which focus on latency periods in nano seconds and can roll the mkt 500 billion in ten minutes are doing the "retail investor " no favors. Un- regulated free market capitalism is never a pretty sight at the extremes.
Post a Comment