This page has been archived and commenting is disabled.
JPM's Carl Carrie On Algorithmic Trading
When the former head of product development in the electronic client solutions group at none other than JP Morgan, Carl Carrie, was last quoted on Zero Hedge, he had some very nasty words for High Frequency Trading. Today, in a podcast transcript by algotradingpodcast Carl shares much more light in just why any reform movement against HFT and PT in general will be met by a huge pushback by exchanges, brokers, infrastructure providers, telcos, and all derivative market players:
Clearly, algorithmic trading is a huge factor. High-frequency trading for Arbitrage's, indexes, ADRs, pairs, ETFs, interlisted trading, as well as automation around auto-working, have all been factors contributing to the growth of algorithmic trading and trading on exchanges.
The exchanges themselves have also been contributing factors. They've invested heavily in capacity and throughput. And the allocations of assets to European equities has also been a minor factor.
Carl also touches on another, so far undiscussed issue - the industrial oligopoly and the economies of scale advantages to the select few:
In the electronic trading space, you're seeing the beginnings of a fallout, and you're seeing larger scale players, some of them become clear winners. Not that they can permanently sustain their competitive advantage, but for a period of time, there is an economic advantage in being the preeminent, top scale player, and probably the next two rungs below.
Hey Christine Varney - if you can look away from Google for longer than 10 seconds, maybe you can focus on where the next real fight for monopoly is ocurring, with materially greater consequences than Firefox being bundled in with Windows 7.
Most notably, Carl discusses the emerging risk types with this new technology. Not surprisingly as Joe Saluzzi would attest, and much to the chagrin of program trading "specialist" Irene Aldridge, the key risk is liquidity, and much more so to the downside, i.e., when it disappears.
There are new risk types. I think, it used to be about timing cost and market impact. Those were two twin pillars that most algorithmic trading has been based on.
And I think, if you look at what's happened recently in the credit markets, it hasn't opened our eyes to liquidity risk, but liquidity cost and liquidity risk is perhaps a different animal. It's not just about price volatility. It's about volume volatility. It's about timing of that volume volatility. It may be there today, and when you want to get out of your position, it may not be there tomorrow. And how do you reflect that into your own trading and into, not just your alpha generation, but on the risk side of the alpha generation? Most risk models don't really take into consideration the kinds of anomalies that we may see on a yearly basis.
It's not a Six Sigma event, typically, that happens when we have a liquidity crisis. And a liquidity crisis very easily moves across from one market, as a class, to another. So, you've got this contagion correlation effect that's massive. So, I think, it's important for all of us to develop new science and new tactics to really deal with that. And particularly, as you talk about emerging markets, there's no sphere that is as liquidity-sensitive as emerging markets is.
Curiously, when Carl left JPM his parting letter had this to say: "Yes, I love equities but I think the biggest transformation in the
market over the next couple of years will be in the OTC fixed income,
credit and commodity markets that are both begging for more liquidity
and transparency and are ripe for a major transformation. I want to be
there at the genesis of that transformation." We at Zero Hedge completely agree with this statement and will be presenting some of our extended ideas on this matter over the next several weeks.
- 5305 reads
- Printer-friendly version
- Send to friend
- advertisements -


seems like what everyone is alluding to and what Tyler is implying is that these automated systems are functional so long as prices rise, but they are quite untested on falling price trends.
it as though these sorts of automated, HFT came into robust fruition in 2009 and were not so much at work during the pernicious down-drafts of 2008. Surely they were there, but it seems like some sort of massive quantum leap has occurred just this year in their use and leverage, etc.
Quants literally had to turn off their systems in 2008. You should have seen them. They were sweating blood and did not know what hit them.
The first quant meltdown was in August of 2007.
Lo and Khandani have a working paper about this:
http://web.mit.edu/alo/www/Papers/august07_2.pdf
Abstract:
During the week of August 6, 2007, a number of quantitative long/short equity hedge funds experienced unprecedented losses. It has been hypothesized that a coordinated deleveraging of similarly constructed portfolios caused this temporary dislocation in the market. Using the simulated returns of long/short equity portfolios based on five specific valuation factors, we find evidence that the unwinding of these portfolios began in July 2007 and continued until the end of 2007. Using transactions data, we find that the simulated returns of a simple marketmaking strategy were significantly negative during the week of August 6, 2007, but positive before and after, suggesting that the Quant Meltdown of August 2007 was the combined effects of portfolio deleveraging throughout July and the first week of August, and a temporary withdrawal of marketmaking risk capital starting August 8th. Our simulations point to two unwinds---a mini-unwind on August 1st starting at 10:45am and ending at 11:30am, and a more sustained unwind starting at the open on August 6th and ending at 1:00pm---that began with stocks in the financial sector and long Book-to-Market and short Earnings Momentum. These conjectures have significant implications for the systemic risks posed by the hedge-fund industry.
But then the question becomes what happened from August 7, 2007 to early October 2007?
In other words what/who pushed the market to new highs?
I have heard that the market crash last fall was actually quite good for several high frequency trading outfits.
HFT is both a buyer and a seller and so does not prop up (or push down) markets. HFT algos don't hold stock long term so whatever they buy they will be selling soon. And if their initial position was selling short they will soon be a buyer.
HFT has been running during market uptrends and downtrends, including the most recent market crash. Interestingly enough, even during the market crash HFT added liquidity. The liquidity was not "fickle" as it did not disappear before, during or after the market meltdown.
Finally, the Aug 2007 events involved quant funds that did not employ high frequency algos. Lo's MIT paper even provides examples that show that the affected funds typically traded on the order of once per day. Further, this event was like a neutron bomb, decimating quant funds but leaving the rest of the market unscathed (down 0.1% in those 1st 10 days of Aug).
excellent read -- TD back on his game
i don't see how Saluzzi, etc can talk about this focused liquidity issue and not just come out and say... well the new risk model is what happens when a bot fucks your out plan?
well, we all remember musical chairs.. GS, Citadel just yanked two chairs away from the russians - and they'll control the game from here on out
what does this mean ? "GS, Citadel just yanked two chairs...."
Agreed, dpos.
Nice post TD.
"MARK IT ZERO, DUDE"
In listening to the podcast it sounds as if Carl was commenting on what was driving volume growth, etc. at exchanges and stated two separate items: algorithmic trading and HFT.
"Clearly, algorithmic trading is a huge factor. High-frequency trading for Arbitrage's, indexes, ADRs, pairs, ETFs, interlisted trading, as well as automation around auto-working, have all been factors contributing to the growth of algorithmic trading and trading on exchanges."
Both use computers. Both use math. Both trade stocks. Neither of them are program trading.
So they use a different name! Program trading,HFT, all the same!
Wow, your pathetic attempts to pick a fight are getting outright childish. If you really work in some aspect of computerized trading, i feel bad for your employer if your work skills are on par with your debating or diminishingly relevant responses.
Instead of adopting your argumentation which usually follows the pattern:
1) Sarcastic swipe at a TD statement;
2) I am right because,
3) I am right;
I will point out that this post focuses on algo and HFT as was pointed out.
But keep up trying to marginalize this topic. Your persistent trolling here simply means that the implications of this thing going mainstream would mean at least one less worthless job in the space.
I'm not sure how it works in the industrial oligopoly but most other places defined terms have meanings.
In this case, program trading is a defined term: 15 or more stocks traded together with a value equal to or greater than $1m. It is on the NYSE website and the link was posted elsewhere on this site.
From reading all the posts, HFT trades tiny orders, like 100 or 200 shares. Too few shares to be 15 individual stocks. Unlikely that the value of 100 or 200 shares would be greater than $1m.
In the the referenced podcast Carl Carrie talks about the automation of work performed by algorithms such as allowing a single trader (at the buy or sell side firms) to trade more orders at once.
If you do not like my style, so be it, but facts are facts and sometimes, they do not agree with preconceptions and biases which are on obvious display on this site.
The NYSE's definition of PT exists for the purpose of applying things like their old curbs. Most of the flow on NYSE meets the definition of programs, since all manner of orders are automatically batched into programs. However, a good portion of everyone talking about the application of automated trading techniques called it program trading because it was one of the first applications of such technology.
It's not that big of a deal to see the term used inexactly, and it's not really appropriate in the industry to use the definition of one regulator/exchange to get all grammar nazi on folks. In reality, these definitions almost always exist to determine where a rule applies to something that actually occurs along a spectrum of similar activity. Also, the terms used by a lot of these places outright contradict.
Let me add that industry jargon evolves from innovation, and often ends up as a meme marketing such innovation to the broader culture (eg, HFT, predatory algos, etc). Often industry jargon are misnomers and misleading except to insiders--take "hedge fund" for example, a term wholly bastardized and one which the average layperson still doesn't have a clue. Industry jargon is not regulatively defined until after the fact, and then the regulatory definition is just an outline of a concept in which innovators once again place much effort to figure out how to skirt any restrictions related to such regulatory definition. Semantics is a big game, yet a profitable one.
Case in point:
The concept of fiduciary duty is at the heart of the relationship among hedge fund managers, hedge funds and hedge fund investors. But until recently, “fiduciary duty” was not defined by any bill or law. Rather, it was the creature of caselaw, and much of that caselaw dealt with whether and to whom the fiduciary duty is owed, rather than the content of the duty.
That has changed with the Obama administration’s proposal on July 10, 2009 of the Investor Protection Act of 2009 (IPA). While the IPA has received significant attention because it would impose a fiduciary duty on broker-dealers that provide investment advice (currently, broker-dealers are subject to a less stringent “suitability” standard), for the hedge fund community, the IPA is noteworthy as the first proposed codification of the substance of a fiduciary duty. In addition, the IPA delegates to the SEC rulemaking authority to define the “client” to whom a fiduciary duty is owed. This could empower the SEC to resolve an ambiguity that has existed since the D.C. Circuit’s 2006 decision in Goldstein v. SEC, 451 F.3d 873 (D.C. Cir. 2006), as to whether a hedge fund manager owes a fiduciary duty to the hedge fund itself, or its underlying investors. That is, the IPA may enable the SEC to provide by rule that a hedge fund manager owes a fiduciary duty to each investor in a hedge fund, and not just to the hedge fund itself. For practical purposes, if the IPA were to become law and if the SEC were to provide by rule that a hedge fund manager owes a fiduciary duty to hedge fund investors, it likely would become easier for hedge fund investors to sue managers based on a range of manager conduct. This is because such a law and rule would more explicitly confer standing on hedge fund investors to challenge various manager actions. In this article, we explain precisely what “fiduciary duty” means in the hedge fund context, and explore to whom the duty is owed (the answer is by no means straightforward). We also explore: the practical consequences of identifying either the hedge fund or its investors as the manager’s “client”; Investment Advisers Act Rule 206(4)-8, the anti-fraud rule with a negligence standard; whether fiduciary duty can be waived; the definition of “client” in the Private Fund Investment Advisers Registration Act of 2009; and the executive compensation provisions of the IPA.
15385 has has a perfectly valid point that HFT and Program Trading are different things.
Allow me to explain.
Program Trading is more accurately called "Portfolio Trading". When an investor wants to trade a whole list of stocks all at once, rather than just one. The business of trading single stocks is called "Cash Trading". As one example, consider what happens when an index such as the russell 2000 or s&p 500 is rebalanced. Managers that run index tracking funds have to migrate their enormous portfolio's from the old index composition to the new one. Generally this involves trading large orders for nearly every stock in the index. The index fund is in a tight spot because other market players know the index funds positions and the exact trades they will have to do. The index funds do not normally have the expertise or infrastructure to execute these large "programs" while avoiding the speculators that would eat their lunch. (think of a program for a musical, not a computer program) So what broker/dealers is buy the entire set of trades from the index fund for a fixed price, and different dealers compete to bid better prices to the fund. The broker/dealer is left with a whole stack of risky positions, and it is now their problem to exit ("trade out") of these positions at a profit. Of course they can also lose money. That is what program (portfolio) trading is.
High Frequency Trading, on the other hand, is a loosely defined term, but I will go with the most intuitive definition which is simply trading done by computers at a very rapid pace, faster than humans can reasonably trade. In order to talk about HFT in a serious way you need to break it down into at least 2 categories. The first are "algorithmic execution strategies". These strategies, or "algos" are computer programs that monitor orders and executions in the market, and work a large order by breaking it down into to lots of little pieces that can be digested by the market one at a time. Algos are used to execute virtually all institutional trades these days. They are virtually mandatory due to regulatory changes such as decimalization and reg nms. But more than later, it needs a whole article.
The second category of HFT is prop trading which exists to meet the urgency requirements of the real order flow coming out of the execution algos I described above. Three distinct concepts. The dots do connect, but they are not the same thing. In order to have a well informed opinion on the matter you have to understand all three.
Thank you for posting that. :)
The problem I see in this blog is that issues get conflated easily by a loose application of terms that have already been defined by the industry. Gives me a sense that the principals are not subject matter experts and are learning as they go. The same applies to the commentators, most of which have an ideological trait that turns off the technical experts. It's sort of like mixing religion and science.
And the principals also exhibit that ideological emphasis. It's a good combination to create traffic to the site but that's about it.
Well said, well said.
I don't think anyone can be subject matter expert on all aspects of the financial industry so the wisdom of the crowd has to help overcome the limitations of the individual. Constructive, explorative discussion is the only way to achieve it. The principals know some things and have some suspicions. Constant drilling and joint effort is the only way to achieve progress.
Commentators are those who could provide additional insight and do quality control. (Especially trigger happy postings get their quality review real fast and this helps.) The flood of anon postings makes it difficult to differentiate between signal and noise. If the principals manage to contain noise the "that's about it" will not come true. On blog with a more limited scope this process is more easily to achieve than here. Also, many of the interesting insights are confidential. Not everything is fraud, but a lot of things are done by bending the rules. Fair or unfair is in the eye of the beholder.
Far more entertaining to see this site develop than to watch cnbc.
"It's a good combination to create traffic to the site but that's about it."
Best summary of this whole thread. The wanton disregard of those presenting counter arguements shows this site is following in the path of many others. It will eventually become a rant zone for a various factions and become useless.
What is frightening is that accredited publications and a sitting US Senator have been driven to action by what has occurred here. This site is fostering a "win goes to the loudest shouter" approach rather than the reasoned, full explored analysis of issues.
"What is frightening is that accredited publications and a sitting US Senator have been driven to action by what has occurred here."
Don't be so sure about that. Fact is that industry publications have been on the "Flash Trading/Order, whatever you want to call it" issue for a longer time than ZH. And lobbying groups have been on the case also on both sides of the issues. It's a pissing contest between exchanges, money talks louder.
What is frightening/odd to me is that I can no longer post is logged in. I have to log in to get the site to accept comments/replies.
BoeingSpaceliner797
"If I were a river, I'd run uphill.
And if you know me, then you know I will."
You make an excellent point.
The real issue here is flash. That fact is HFT trading existed long before flash and will exist after.
Flash should go. The concerns around a trading approach and model, HFT, should be separated from the flash debate.
Unless there are some other HFT strategies that are equally/more destructive that we don't know about. Keep that in mind.
I for one have found posting here to be frustrating due to Tyler's HFP (High Frequency Posting) practices. You try to write up some serious information but by the time your done there are 10 new threads and whatever you were trying to say is quickly lost in rush to comment on the new thread. The site could be improved if Tyler would slow down the posts to one or 2 per day give people time to read and respond to the comments. It would also help if the other side was represented by at least one principal blogger.
Not really a good description of what's been going on here.
Important topics are ignored by the media - who would have a duty towards their readers to ask questions and investigate but choose not to for some reason- so others step in an ask these questions and investigate. If a Senator finally feels the need to act on the information (which he cannot get from traditional media) you claim foul, even though lobbyists have access to Senators all the time.
Besides issues are explored and explained very well and instead of just consuming what is published here, you can always go and investigate.
I don't see the problem you are having.
"It's sort of like mixing religion and science."
Mine is a purely philosophical observation and not an attack on your posts/thoughts/contributions to this thread or site. In western civilization, religion and science have already been mixed (i.e., science is the new religion), IMO.
As I said before, industry jargon evolves from innovation, and often ends up as a meme marketing such innovation to the broader culture (eg, HFT, predatory algos, etc). The only "subject matter experts" are the ones developing and/or adopting the innovation, and then they're making up terms to communicate the concept as they go along.
"Loose application of terms that have already been defined by the industry" is a bad habit by all... often it comes down to the "dialect" used in one organization vs. another.
Hence, I think conclusion that "principals are not subject matter experts and are learning as they go" is subject to debate. HFT at GS via SLP is going to be different then HFT at Citadel... why else would Citadel want GS' IP?
Where was is stated Citadel wanted GS IP? A former employee of Citadel started a firm and hired the individual accused of stealing IP from GS. Can you provide information to back up your statement that Citadel wanted GS IP?
I am not into the risk management so I could be a bit wrong on it, but I am very puzzled why systematic risk is given some absolute measure, let's say 1. It seems to be obvious that systematic risk changes from year to year, still the most measures/assumptions I ve seen assume the total market risk is constant. (let's say Beta, etc)
Time-stability of beta is an interesting assumption. Another breeding ground for black swans.
those assumptions are wrong; the market risk as a derivative fractal variable of a larger set of variables, and it's represented as a constant by comparison, not neutrality and its existence, and is a constant just like trees are a constant or an orange .... now that we know what kind of derivative fractal variable market risk is, we also know that its micro-symmetric with total systemic risk. so for example if systemic risk approaches 1 then also the market risk approaches one, but the opposite does not have to be true, and it almost never is .... in logical terms systemic risk is a higher expression of different variables, and market risk is just one of those variables, but the opposite is not true. so who ever said that the market risk is a constants if observed solely by its proximity to 1 ( absolute risk) or 0 ( no risk ) was BSing ...
So, what's the "New" money, gonna be?
It's not paper...then what?
Food? Hoarding?
"It's not paper...then what?"
our world is too large for anything else but paper to serve as a medium of exchange. It could be "electronic" paper, but using anything else would increase the cost of transactions, so highly doubt any change is possible.
What if there's a disconnect, like Gold and ETF Gold. Owning the paper and not the "electronic" paper. Increasng the transaction.
It's hard or almost impossible to use the gold as a transaction medium for 14TR Economy. Also if you an average investor, the transaction cost of owning physical gold will probably be always higher. (Fees, Commissions, Inconvenience, unless you believe in dissolution of the country. Well, while everything is possible, the probability of dissolution is minimal. And even if a few states do separate, would you trust a dollar backed by Federal government less than IOU backed by the state of CA, somehow I don' think so.)
TD, do you have any charts (or point me in a direction) that shows the growth of these strategies (HFT and algo trading) as a percentage of total volume traded on NYSE, from August 2007 - July 2009?
Thanks
And the hits just keep on rolling.... Have at it and remember to have some fun while you're at it.
Lets take 3 seemingly unrelated and on the surface sensible strategies. Firstly let us assume that investors are not stupid and can interpret what CEO’s are saying. Their strategy will most likely be to hold very liquid assets so they can take advantage of the ride up and get out quickly on the way down. Next let us assume that retailers got burned in the lead up to Christmas where they stocked up too quickly and had to sell off at a discount. It would make sense for them to buy stock as demand rises this Christmas which will result in less orders coming in for early autumn. This would result in the PMI figures dropping and could spark a market decline.
Now let us look at what happens with HFT and specifically on the prop trading side. The banks will still be making money but the markets will be in accelerated decline as a result. The big risk and perhaps Carl is hinting at it is that most trade will be hedged and this is our third sensible strategy. The problem is that all the markets then become interlinked and as the equity market drops 20 percent, then perhaps oil goes up 200 percent.
Bank risk models are about protecting the bank and what they fail to recognise is that if you damage the consumer, other players and the economy by managing risk in a narrow view the banks run the risk of the economy biting back.
"I’m shocked, shocked to find that gambling is going on in here!" — Captain Renault
Pay no attention to these smarty pants bloggers.
Just listen to our economy’s Baghdad Bob - Larry Kudlow.
Or if that hurts too much, you can just turn the volume down and admire the money honeys.
Viva CNBC!
The problem is that when the market volume is dominated by a few giants like GS, Credit Suisse, JPM etc, it is really a war between them and the remaining hedge funds, institutions, and retail investors. Guess which side will win? GS and co have deep pockets, and can easily determine when to run up the market and when to run it down. They have the cash, and the resources (inside information) to outlast others. It will always be at the point of maximum pain to the rest. If everyone thinks the market is going to have a pullback next week, I can guarantee you that will not happen. GS will have taken up positions accordingly. Only when everyone is bullish, and after GS has sold its shares to others, will it borrow those shares back, and unload them at a furious clip. This will trigger stops which will lead to further selling pressure, and then GS will buy those shares back after a major down leg and book a huge profit. Welcome to the new Las Vegas, where the house always wins. Trading is pure gambling at this point. Better to find other vocations if possible.
Right now, a lot of people in the blogosphere seem to be thinking that the SP500 will run upto 1007-1015 (38.2% Fibonacci retrace) and then chop down for a correction. I think EW theorists are backing this idea. I think GS will ensure that there is a melt-up by buying at the margin so that it just busts right through that resistance level without breaking a sweat all the way to 1250. All those who went short at SPX 975 are going to be hammered into the ground, and will feed into the meltup on more short covering. The pension funds and mediocre hedge funds are going to panic and buy, buy, buy. Their furious clients are probably calling them now asking why they missed the rally, and why are they not in the market. Pity, that they are going to end up as the bag-holders in the next few weeks. People are emotional, and to be in this war, you have to have nerves of steel.
http://www.nytimes.com/2009/07/26/science/26robot.html?em
A mixed metaphor. The cheats use robots in this case, the robots don't use cheats.
I keep hearing this stat that 70% of all trading volume is HFT. But I have not been able to find any primary source through google. I think it was a TABB group paper. Does anyone have a link?
It's hard to believe this Alice-in-Wonderland interview took place after the meltdown. It's a mad tea party for sure, no matter when it happened, but the eagerness of all participants to ignore the obvious (that ALL the old paradigms are now in question) and forge ahead as though nothing of much importance had happened to shake their world is tremendously amusing. I was going to say disturbing, but words like that belong to the old paradigm, I'm afraid. :-)
http://amoleintheground.blogspot.com/
"JPMorgan don't expect to make any money from high frequency algo trading. But they have to offer it to their clients so they have no choice but to pour millions of dollars into systems development to keep up with the other banks. It's an arms race - and those rarely end well for anyone."
- a former guy from equity trading IT at JPM
100% unadulterated crooks
JPMorgan and Madoff
http://www.nytimes.com/2009/04/25/business/economy/25madoff.html
the problem as i see it is that in terms of trading securities, the stock market is a retail market and relatively illiquid. most hedge funds are glorified retail investment vehicles who are all running the same highly levered arbitrage strategies. in the bond market you can put on a $1b arb and get out without moving the market a tick. in stocks you are talking about a 5% swing. have 5,000 HFs trying at the same time and you have a fat tail. the irony is that these computer generated trades are not value added for the investors who are paying massive fees and carried interest. there is no alpha from a computer and therefore at some point the market will make it unprofitable which drive money elsewhere. this is just a typical Wall Street fad that starts by making unreasonable profits only to see the law of diminishing returns kick in.. just give it some time.