This page has been archived and commenting is disabled.
Black Box Trading: Why They All “Blow-Up”
Submitted by Dominique Dassault of Global Slant
"We Are All Doing The Same Thing”
I recently listened to a podcast with some all-star [there are awards for everything now] “Black Box” equity trader. It was quite a “telling” interview & I thank him for his insights but I’d heard it all before. His confidence was staggering considering the general unpredictability of the future and, of course, the equity markets. He explained how he had completely converted from a generally unsuccessful, discretionary technical trading style to a purely quantitative and scientific trading mode. He seemed to be so excited that his models, according to him, were pretty much “bullet proof”. Having had more than just some tangential experience with black box modeling and trading myself I thought…you know…some people will just never learn.
**************************************************************************
You see some years ago I was particularly focused on quantitative investing. Basically, “screw” the fundamentals and exclusively concentrate on price trends/charts and cross security/asset correlations [aka “Black Box” trading]. I was fascinated with the process and my results were initially stellar [high absolute returns with Sharpe Ratios > 2.0]. And, after looking at the regression data many others agreed. I was in high demand. So I “made the rounds” in Manhattan and Greenwich to a group of high profile hedge funds. It was a very exciting time for me as the interest level was significant.
As it turned out I had the good fortune of working with one of the world’s largest and best performing hedge funds. Their black box modeling team had been at it for years…back-testing every conceivable variable “inside and out”…twisted in every conceivable/measurable manner…truly dedicated to the idea that regression tested, quantitative trading models were the incremental/necessary “edge” to consistently generate alpha while maximizing risk-adjusted, absolutely positive returns. We worked together for some time and I became intimately involved with their quantitative modeling/trading team…truly populated with some of the best minds in the business.
While in Greenwich Ct. one afternoon I will never forget a conversation I had with a leading quantitative portfolio manager. He said to me that despite its obvious attributes “Black Box” trading was very tricky. The algorithms may work for a while [even a very long while] and then, inexplicably, they’ll just completely “BLOW-UP”. To him the most important component to quantitative trading was not the creation of a good model. To him, amazingly, that was a challenge but not especially difficult. The real challenge, for him, was to “sniff out” the degrading model prior to its inevitable “BLOW-UP”. And I quote his humble, resolute observation “because, you know, eventually they ALL blow-up“…as most did in August 2007.
It was a “who’s who” of legendary hedge fund firms that had assembled “crack” teams of “Black Box” modelers: Citadel, Renaissance, DE Shaw, Tudor, Atticus, Harbinger and so many Tiger “cubs” including Tontine [not all strictly quantitative but, at least, dedicated to the intellectual dogma]…all preceded by Amaranth in 2006 and the legendary Long Term Capital Management’s [“picking up pennies in front of a steam-roller“] demise one decade earlier.
Years of monthly returns with exceedingly low volatility were turned “inside out” in just 4-6 weeks as many funds suffered monthly losses > 20% which was previously considered highly improbable and almost technically impossible…and, voila…effectively, a sword was violently thrust through the heart of EVERY “Black Box” model. VaR and every other risk management tool fell victim to legitimate liquidity issues, margin calls and sheer human panic.
Many of these firms somehow survived but only by heavily gating their, previously lightly-gated, quarterly liquidity provisions. Basically, as an investor, you could not “get out” if you wanted to. These funds changed the liquidity rules to suit their own needs…to survive…though many did fail.
**************************************************************************
Anyway…to follow up on my dialogue with the esteemed portfolio manager…I asked “why do they all “BLOW-UP”? What are those common traits that seem to effect just about every quantitative model despite the intellectual and capital fire-power behind them? And if they all eventually “BLOW-UP” then why are we even doing this?”
He answered the second part of the question first…and I paraphrase…“We are all doing this because we can all make a lot of money BEFORE they “BLOW-UP”. And after they do “BLOW-UP” nobody can take the money back from us.” He then informed me why all these models actually “BLOW-UP”. “Because despite what we all want to believe about our own intellectual unique-ness, at its core, we are all doing the same thing. And when that occurs a lot of trades get too crowded…and when we all want to liquidate [these similar trades] at the same time…that’s when it gets very ugly.“. I was so naive. He was so right.
**************************************************************************
Exactly What Were We All Doing?
We all knew what the leaders wanted and, of course, we wanted to please them. Essentially they wanted to see a model able to generate 4-6% annual returns [seems low, I know, but I’ll address that later]…with exceptionally low volatility, slim draw-down profiles and winning months outweighing the losing months by about 2:1. They also wanted to see a model trading exceptionally liquid securities [usually equities].
Plus the model, itself, had to be completely scientific with programmable filtering and execution [initiation and liquidation] features so that it could be efficiently applied and, more importantly, stringently back-tested and stress-tested . Long or short did not really matter. Just make money within the parameters. Plus, the model had to be able to accommodate at least $100M [fully invested most of the time as cash was not an option] and, hopefully, much more capital. This is much easier said than done but, given the brainpower and financial resources, was certainly achievable.
This is what all the “brainpower” learned…eventually.
First of all, a large number of variables in the stock selection filter meaningfully narrowed the opportunity set…meaning, usually, not enough tickers were regularly generated [through the filter] to absorb enough capital to tilt the performance meter at most large hedge funds…as position size was very limited [1-2% maximum]. The leaders wanted the model to be the hero not just a handful of stocks. So the variables had to be reduced and optimized. Seemingly redundant indicators [for the filter] were re-tested and “tossed” and, as expected, the reduced variables increased the population set of tickers…but it also ramped the incremental volatility…which was considered very bad. In order to re-dampen the volatility capital limits on portfolio slant and sector concentration, were initiated. Sometimes market neutral but usually never more than net 30% exposure in one direction and most sectors could never comprise more than 5% of the entire portfolio. We used to joke that these portfolios were so neutered that it might be impossible for them to actually generate any meaningfully positive returns. At the time of “production” they actually did seem, at least as a model, “UN-BLOW-UP-ABLE” considering all the capital controls, counter correlations and redundancies.
Another common trait of these models that was that, in order to minimize volatility, the holding periods had to be much shorter than a lot of us had anticipated. So execution [both initiation and liquidation] became a critical factor. Back then a 2-3 day holding period was considered acceptable, but brief, although there were plenty of intra-day strategies…just not at my firm…at least not yet. With these seemingly high velocity trading models [at the time], price slippage and execution costs, became supreme enemies of forecasted returns. To this day the toughest element to back-testing is accessing tick data and accurately pinpointing execution prices. Given this unpredictability, liberal price slippage was built into every model…and the model’s returns continued to compress…not to mention the computer time charges assessed by the leadership [which always pissed me off].
So, given all of this, what types of annual returns could these portfolios actually generate? A very good model would generate a net 5-7%…but 3.5-5% was acceptable too. So how then could anybody make any real money especially after considering the labor costs to construct/manage/monitor these models…which…BTW…was substantial?
Of course there was only one real way…although it would incrementally cut into performance even more, in the near term, but ultimately pay off if the “live” model performed as tested. The answer = LEVERAGE…and I mean a lot of it…as long as the volatility was low enough.
Back-tested volatility was one thing and live model volatility was another thing so leverage was only, ever so slowly, applied…but as time passed and the model performed, the leverage applied would definitely increase. Before you knew it those 5% returns were suddenly 20-25% returns as the positive beauty of leverage [in this case 4-5x] was unleashed.
******************************************************************
Fast forward 10 years and the objectives of hedge funds are still the same. Generate positive absolute returns with low volatility…seeking the asymmetrical trade…sometimes discretionary but in many cases these “Black Box” models still proliferate. And BTW…they are all “doing the same thing“…as always…current iteration = levered “long” funds.
What has changed though is the increased dollars managed by these funds [now > $3.5T] and the concentration, of these dollars, at the twenty largest funds [top heavy for sure]. What has also considerably changed is the cost of money…aka leverage. It is just so much cheaper…and, of course, is still being liberally applied but, to reiterate, in fewer hands.
Are these “hands” any steadier than they were ten years ago? I suppose that is debate-able but my bet is that they are not. They are still relying on regression-ed and stress tested data from the past [albeit with faster computers & more data]. They may even argue that their models are stronger due to the high volatility markets of ’08/’09 that they were able to survive and subsequently measure, test and integrate into their current “Black Boxes”…further strengthening their convictions…which is the most dangerous aspect of all.
Because…Strong Conviction + Low Volatility + High Levels/Low Costs of Leverage [irrespective of Dodd-Frank] + More Absolute Capital at Risk + Increased Concentration of “At Risk” capital + “Doing the Same Thing”…adds up to a combustible market cocktail.
Still a catalyst is needed and, as always, the initial catalyst is liquidity [which typically results in a breakdown of historic correlations as the models begin to “knee-bend”…and the perceived safety of hedges is cast in doubt] followed by margin calls [the ugly side of leverage…not to mention a whole recent slew of ETF’s that are plainly levered to begin with that, with the use of borrowed money, morph into “super-levered” financial instruments] and concluding with the ever ugly human panic element [in this case the complete disregard for the “black box” models even after doubling/trebling capital applied on the way down because the “black box” instructed you to]. When the “box” eventually gets “kicked to the curb”…that is when the selling ends…but not after some REAL financial pain.
********************************************************
So can, and will, this really occur once again? It can absolutely occur but it is just impossible to know exactly when…although there are plenty of warning signs suggesting its likelihood…as there have been for some time.
However, I am quite confident of the following:
1. The Fed Is Clueless On All Of This
[just too many moving parts for them…they cannot even coordinate a simple “cover-up” of leaked monetary policy…so no way the academics can grasp any of this…just liked they missed LTCM. Nor it seems…do they have any interest in it as it is too tactical for a strategically inclined central banker].
More Importantly Though
2. “Eventually They ALL Blow-Up”
3. The Only Real Question Left To Answer = HOW BIG CAN THIS POSSIBLE “BLOW-UP” ACTUALLY BE?
I think…pretty BIG.
- 17084 reads
- Printer-friendly version
- Send to friend
- advertisements -


He explained how he had completely converted from a generally unsuccessful, discretionary technical trading style to a purely quantitative and scientific trading mode. He seemed to be so excited that his models, according to him, were pretty much “bullet proof”.
well that explains it......now what could possibly go wrong?
What can really go wrong is what that guy in Europe did which has the traders all in a tizzy:
Game other people's black boxes. Figure out how they are programed by studying their trading habits, and use this information to rape them.
So, in the long run... (They all blow up), they're all dead, eh?
Fascinating.
Chaos modeling, just look at weather forecasts, the farther out they predict, the more inaccurate they become. Everybody playing is now relying on chaotic models but sooner or later a tornado will hit that wasn't in the model.
The FED is at the tip of the Matrixberg. Sort of like an unstopable vs immovable situation. Good luck with that.
That's the reason I tend to use an avalanche metaphor......
What a collosal waste of human intelligence and labor.
What is the point of the stock market again?
Think of it as a vacuum cleaner. It sucks back all the wealth that you've earned through sweat, toil, and labor so that people that think they're better than you can enjoy cozy summer parties in the Hamptons. They'd thank you for your willing participation, but instead they'll just hire Hilsenrath to tell you to pick up the pace a little, you missed a spot.
Bring it on because only with a complete collapse of the system can we ever have any chance of rebuilding without Central banks and "Debt Money". Even then, the chances are slim given the totally broken and corrupt system operating now.
Yep, and it'll fix the problem...for awhile. But sooner or later the "smart people" will show up again with a new (old) super shiny idea that make everything "better"....
On a long enough timeline, the survival rate of all Black Boxes drops to ZERO.
The proper application of a big hammer could speed things up though... ;-)
I thought the only way to make money is BTFD...
lol This is what we should do to all the central banksters.
German man takes revenge on ex-wife by sawing everything they own in half, selling it on eBay | Fox News
Hedge fund buys and sells through one of the "boyz". They know your position. They also know how to fuck you when its time. That why all these models blow up.
Screw the fundamentals, just focus on figuring out what other machines are buying and try to frontrun their trades. Follow charts because if a stock is trading down, but about to hit some arbitrary level drawn, it will shoot back up. However the timing has to be perfect for maximum profit.
This kind of garbage goes on and on and on. It is meaningless to the greater economy and has no bearing on the day to day business of the corporations the stock tickers represent, but makes billions for "traders".
So there isn't any real point for it. Absolutely no difference from picking lottery numbers.
Too much of anything leads to grief. Nothing complicated about that.
The only thing that is not likely to reverse any time soon is the dumbing down of the world's population.
The Catch 22 of curve fitting to complex polynomials.
It is always the outliers that change things, not the statistically relevant confidence points.
Science and math are tools to bring understanding to us humans about what is already there, what already exists.
Einstein's math teacher was convinced he would never amount to anything and yet he helped create the atom bomb.
A vital component of our existence is chaos; without chaos there would be no order, and vice-versa.
Think of it as the farthest points in the opposing loops of the infinity loop; in order for time to fold on itself and make a complete loop there must be chaos for it to reach the nexus.
It is always about the outliers, not the reliability. When everyone is doing the same thing - it forces chaos to intervene.
an algorithm is only as good as the person who designed it. it is eminently possible to beat an algorithm, you just have to better than the fker behind it. and i dont believe algorithmic trading will help you frontrun goldman sachs and the new york fed in the gold and oil markets. in my opinion this requires a higher level of cognition than machines are capable of.
“Plus the model, itself, had to be completely scientific with programmable filtering and execution [initiation and liquidation] features so that it could be efficiently applied and, more importantly, stringently back-tested and stress-tested”
There is conceit behind the numbers, and monstrous assumption that humans engage in. This assumption is that price numbers are perfect, and they add algebraically.
But, if one examines “money as credit” those accounting numbers are popping into being and vanishing in a temporal way. In the course of their life cycle they also create and extinguish debt instruments. These debt instruments may also physically vector into paths, even concentrating into wall-street. In fact, the obvious conclusion to a centralized private reserve system such as Federal Reserve System was exactly that – to physically concentrate the debts. Dollar as world reserve currency also ties into the reserve system, centralized at New York Fed.
At Bretton Wood, the idea was to have “world bank” loans denominated as dollars, to then concentrate those debts into Wall Street. This gave world money power to Wall Street, and by familial relation, also to the City of London (Bank of England).
Price numbers are money numbers, and today’s money is private bank credit. Private bank credit has its “debts” and additionally it taxes the future. It is the nature of private bank credit to tax the future. Upon hypothecation, new debtor is gambling that his new credit will be available in future. But, hidden in the private credit mechanism are underlying FACTS: 1) the credit may not be available in future. 2) Usury on this credit means it will drain toward its debt instruments (wall street typically) thus vectoring away from original hypothecator. 3) Debt money system will act like a cancer and direct its credit poorly, thus nourishing wrong parts of the economy; which negatively impacts the future 4) future is unknown hence credit will want to be insured, making this money type even more of an economic drag. 5) Credit will want to cross link in counter-party arrangements, adding systemic collapse risk; where chains of counterparties break down in domino fashion. 6) Usury will make debt instrument grow exponentially and thus demand much more credit back than what was initially created.
“How can you model prices when the numbers themselves are a conceit of private credit?” You can’t. It is impossible. It is hubris and ignorance that pervades the economic profession, the accounting profession, and all those who think they know how money works. The bankers know how it works, as they observe their ledgers. Adding numbers together implies algebraic linear function, but the numbers are NOT LINEAR.
Even something as simple as creating mortgage backed securities hypothecates a new type of instrument that acts as additional drain on credit money supply, thus changing the ratio of more debt demands to money.
Hubris, ignorance, conceit, greed – few economists will acknowledge that private credit as money has these flawed characteristics coded into its structure. The money system is NOT scientific; it was built up as fraud ad-hoc style over centuries. The main architects were Gold-Men, primarily Jews, and their goal was not then, and not now, the betterment of humanity and civilization. The goal of gold-men, now metastasized into Wall Street, is to TAKE RENTS. That is, to steal and act as a parasite; The control levers of a credit system, have as its core function - an ability to maneuver prices, to centralize debts, to control debts attendant creation/destruction; to then take usurious gains, to bribe/sabotage the polity. This private credit system is an enormous rent scheme enabling mankind’s worst elements. Honest producers and laborers, who trade their life energy for money, are in competition with those who create money as a con game; con artists who maneuver prices with various schemes to take rents. Some con artists are not even aware of their participation in the con; they are just maneuvering their algos and their pricing models. This credit system then pits man against man, where one is the master and the other a slave.
A private credit money system is not redeemable; it has too many internal contradictions and conceits. It needs to be thrown onto the trash heap of history. A proper money system has money type that relates to Wealth, and is volume controlled. A proper money system does not tax the future and demand from the future, but stands in the present. Debt instruments should only circulate locally, where they were created, so debtor can pay back creditor. This also creates a local economy, and does not feed the “international” monster. And make no mistake about it; the international is operating worldwide pyramid credit IOU scheme with a hybrid religious philosophy in alignment with Kabala and Occulted methods. They view the laboring producers of the world as animals to be harvested and killed as necessary.
Humanity needs new money, or higher civilization is impossible. Creditors should be able to loan their honest currency wealth type money, and debtors should be able to pay it back in a non- usurious way. www.sovereignmoney.eu
"Anyway…to follow up on my dialogue with the esteemed portfolio manager…I asked “why do they all “BLOW-UP”?"
Because a cycle has an up phase and a down phase.
The Nobel laureates at LTCM thought they had created a perpetual money making machine- then the market phase turned down and they were suddenly racking up exponential losses.
In 2008, the investment banks crashed.
Now, "everyone" thinks the market can't go down again, because they think the FED is the be all and end all. They will be sadly mistaken. Cycles always turn down after the up phase.
Great article and well written by all means. It's like anything else when the game gets crowded and all are using the same strategies the returns are not as big as they were initially and yes it will break, when and how we don't know but it will when the arrays of data can no longer support the complex query world of trading by the algos.
If you have never watched the Quant Documentary, which was released 3 days before the flash crash, it explains all of this pretty well and in a fashion to where the layman can understand as well. I like following the storyline of the guy who wants to be a Quant...his explanations are good describing the difference between Quants and the IT folks at financial organizations. It's I think about the 4th video and titles are they at the collection of Killer Algorithms videos. The other videos are good too but this one pretty much reinstates exactly what has been said here from Quants that got out of the madness. You have to love the part where Quant Paul Wilmott tells you how useless Economists are too.
http://www.ducknet.net/attack-of-the-killer-algorithms/
and not to mention the technqiques they use that work so well in physcial sciences dont always work in the markets