Thursday, August 17, 2017

"The Next Quant Meltdown"

The quant quake was so scary Goldman's CFO started babbling stupid stuff. More after the jump.
From Institutional Investor:

Fund managers who survived the quant bloodbath of August 2007 say the strategy is safer ten years later. But with its popularity soaring to new heights, some wonder whether crowded trades — and rising leverage — will lead to another downfall.
Ten years ago, during the second week of a sultry New York August, Michael Mendelson was getting a sandwich at the Subway shop near the Greenwich, Connecticut, office of AQR Capital Management when all hell broke loose. As Mendelson glanced at his BlackBerry, out of nowhere, it seemed, AQR’s funds were in the red.

Very quickly, losses started snowballing, with AQR’s flagship fund down 13 percent during the first ten days of August before recovering all of it by month’s end. “You didn’t know how far it was going to go,” recalls Mendelson, a principal at AQR.

AQR was not alone. Some of the biggest and smartest hedge funds in the world suddenly tumbled in tandem, as their algorithmic trading systems went haywire.

As the events unfolded, other hedge fund managers scurried from their idyllic Hamptons retreats to contain the damage and lashed out at the quants, who turned out to be massively overleveraged and highly correlated — triggering an unwinding that was a preview of a collapse that would befall the financial world a year later. “‘These damn quants. They couldn’t get a date in high school, and now they f——ked my month,’” was the joke Peter Muller, who was running a secretive internal quant fund at Morgan Stanley at the time, liked to tell afterward, paraphrasing the frustrated fundamental managers.

But to everyone making high-stakes computerized investment bets, it wasn’t funny — it was shocking. “We had never seen anything like this before. . . . We were trained in statistics. For all intents and purposes this shouldn’t have happened,” says a manager of a major quant hedge fund who lived through the event, sleeping in his Wall Street office to try to contain the damage.

After a four-day quant rout, the Dow fell almost 400 points, then began to recover, with little overall effect on the market. But several funds ultimately did not survive the chaos that started that summer. Goldman Sachs Asset Management, which in 2006 ran the largest hedge fund empire in the U.S., eventually ended up shuttering two of its most prestigious funds, Global Alpha and Global Equity Opportunities, after losing billions of dollars. Tykhe Capital, whose funds had lost between 17 percent and 31 percent by August 9, also eventually shut down....MORE

For more on the quantpocalypse here's MIT's uberquant Andrew Lo via the New York Fed:
"What happened to the quants in August 2007? Evidence from factors and transactions data∗"
(77 page PDF)

From 2012's "David Viniar retiring as Goldman CFO (GS)":

.....*In August 2007 the market and in particular the funds suffered what came to be called the "Quant Quake".
As the Financial Times reported on Aug. 13, Mr. Viniar said one of the dumbest things of his life:

“We were seeing things that were 25-standard deviation moves, several days in a row” 
In March 2009 we posted "David Viniar, CFO of Goldman Sachs Blows Smoke at Journalists on AIG" which had this bit regarding the 25 Sigma comment:
Several folks, when they finally quit laughing, pointed out how blatently Mr. V was spinning.
Most however underestimated how infrequent 25SD events are, the most common guess being once in 100,000 years. Tee hee.

In a snappy little eight page paper "How Unlucky is 25 Sigma" we see that at 7 Sigma the odds are:
...The reader will note that as k gets bigger the probabilities of a k-sigma event fall
extremely rapidly:
• a 3-sigma event is to be expected about every 741 days or about 1 trading day
in every three years;

• a 4-sigma event is to be expected about every 31,560 days or about 1 trading
day in 126 years (!);

• a 5-sigma event is to be expected every 3,483,046 days or about 1 day every
13,932 years(!!)

• a 6-sigma event is to be expected every 1,009,976,678 days or about 1 day
every 4,039,906 years;

• a 7-sigma event is to be expected every 7.76e+11 days – the number of zero
digits is so large that Excel now reports the number of days using scientific
notation, and this number is to be interpreted as 7.76 days with decimal point
pushed back 11 places. This frequency corresponds to 1 day in 3,105,395,365
The authors go on to describe the problems involved in computing numbers on the cosmological scales required for 25 standard deviations. A good read, both for the statistically challenged and for pros like Viniar, a very highly paid PR guy, in addition to his CFO duties.
I was so tickled by that little paper that we posted on it a second time, the day MIT added it to their Physics arXive:
When Goldman Sachs was Really, Really Unlucky (GS)

And a third time in 2011:
"How Unlucky is 25-Sigma?" (and a huge apology) GS; C; BST; UBS; MER
We have more on Mr. V. but I'll always think of him as frozen in 2007.