Physicist, Startup Founder, Blogger, Dad

Monday, November 25, 2013

Goldman v. Aleynikov

Michael Lewis on the Aleynikov-Goldman HFT matter. The article mentions that Goldman trailed other players like Citadel when Aleynikov was hired. The head of HFT at Citadel was (I believe) a contemporary of mine in grad school at Berkeley, who did his dissertation in string theory. Malyshev, the guy who hired Aleynikov from Goldman, had his own legal problems when he left Citadel.
Vanity Fair: ... Serge knew nothing about Wall Street. The headhunter sent him a bunch of books about writing software on Wall Street, plus a primer on how to make it through a Wall Street job interview, and told him he could make a lot more than the $220,000 a year he was making at the telecom. Serge felt flattered, and liked the headhunter, but he read the books and decided Wall Street wasn’t for him. He enjoyed the technical challenges at the giant telecom and didn’t really feel the need to earn more money. A year later the headhunter called him again. By 2007, IDT was in financial trouble. His wife, Elina, was carrying their third child, and they would need to buy a bigger house. Serge agreed to interview with the Wall Street firm that especially wanted to meet him: Goldman Sachs.

... And then Wall Street called. Goldman Sachs put Serge through a series of telephone interviews, then brought him in for a long day of face-to-face interviews. These he found extremely tense, even a bit weird. “I was not used to seeing people put so much energy into evaluating other people,” he said. One after another, a dozen Goldman employees tried to stump him with brainteasers, computer puzzles, math problems, and even some light physics. It must have become clear to Goldman (as it was to Serge) that he knew more about most of the things he was being asked than did his interviewers. At the end of the first day, Goldman invited him back for a second day. He went home and thought it over: he wasn’t all that sure he wanted to work at Goldman Sachs. “But the next morning I had a competitive feeling,” he says. “I should conclude it and try to pass it because it’s a big challenge.”

... He returned for another round of Goldman’s grilling, which ended in the office of one of the high-frequency traders, another Russian, named Alexander Davidovich. A managing director, he had just two final questions for Serge, both designed to test his ability to solve problems.

The first: Is 3,599 a prime number?

Serge quickly saw there was something strange about 3,599: it was very close to 3,600. He jotted down the following equations: 3599 = (3600 – 1) = (602 – 12) = (60 – 1) (60 + 1) = 59 times 61. Not a prime number.

The problem wasn’t that difficult, but, as he put it, “it was harder to solve the problem when you are anticipated to solve it quickly.” It might have taken him as long as two minutes to finish. The second question the Goldman managing director asked him was more involved—and involving. He described for Serge a room, a rectangular box, and gave him its three dimensions. “He says there is a spider on the floor and gives me its coordinates. There is also a fly on the ceiling, and he gives me its coordinates as well. Then he asked the question: Calculate the shortest distance the spider can take to reach the fly.” The spider can’t fly or swing; it can only walk on surfaces. The shortest path between two points was a straight line, and so, Serge figured, it was a matter of unfolding the box, turning a three-dimensional object into a one-dimensional surface, then using the Pythagorean theorem to calculate the distances. It took him several minutes to work it all out; when he was done, Davidovich offered him a job at Goldman Sachs. His starting salary plus bonus came to $270,000.

... One small example of the kind of problems Serge found: Goldman’s trading on the NASDAQ exchange. Goldman owned the lone (unmarked) building directly across the street from NASDAQ in Carteret, New Jersey. The building housed Goldman’s dark pool. When Serge arrived, 40,000 messages per second were flying back and forth between computers inside the two buildings. Proximity, he assumed, must offer Goldman Sachs some advantage—after all, why else buy the only building anywhere near the exchange? But when he looked into it he found that, to cross the street from Goldman to NASDAQ, a signal took five milliseconds, or nearly as much time as it took a signal to travel on the fastest network from Chicago to New York. “The theoretical limit [of sending a signal] from Chicago to New York is something like seven milliseconds,” says Serge. “Everything more than that is the friction caused by man.” The friction could be caused by physical distance—say, if the signal moving across the street in Carteret, New Jersey, traveled in something less direct than a straight line. It could be caused by computer hardware. (The top high-frequency-trading firms chuck out their old gear and buy new stuff every few months.) But it could also be caused by slow, clunky software—and that was Goldman’s problem. Their high-frequency-trading platform was designed, in typical Goldman style, as a centralized hub-and-spoke system. Every signal sent was required to pass through the mother ship in Manhattan before it went back out into the marketplace. “But the latency [the five milliseconds] wasn’t mainly due to the physical distance,” says Serge. “It was because the traffic was going through layers and layers of corporate switching equipment.” ...
In this last paragraph Aleynikov sounds more like Sakharov than a millionaire quant ;-)
“If the incarceration experience doesn’t break your spirit, it changes you in a way that you lose many fears. You begin to realize that your life is not ruled by your ego and ambition and that it can end any day at any time. So why worry? You learn that, just like on the street, there is life in prison, and random people get there based on the jeopardy of the system. The prisons are filled with people who crossed the law, as well as by those who were incidentally and circumstantially picked and crushed by somebody else’s agenda. On the other hand, as a vivid benefit, you become very much independent of material property and learn to appreciate very simple pleasures in life such as the sunlight and morning breeze.”


David Brahm said...

Frightening -- glad I don't work for GS! But are you claiming that Misha Malyshev was a Berkeley contemporary? He got his Ph.D. from Princeton in 1999.

steve hsu said...

No, Malyshev's boss at Citadel was someone you know/knew. A student of Stanley's. Look up the court docs ;-)

David Brahm said...

You're right!

Cornelius said...

“I was not used to seeing people put so much energy into evaluating other people,”

I wonder what portion of the value that firms get from this type of vetting process is the aura of prestige/elitism that it builds around the firm. Yes, quants have more hard skills than other finance types, but from what I know of quant interviews vs. the daily demands of real quant jobs, the interviews are overkill.

I saw this same thing in consulting. Our interview process was extremely noisy. The aura of elitism we built around the firm kept most unqualified applicants from even applying. After that, we picked the applicants that already had consulting experience, followed by those with other applicable business experience. If there were any interview slots left, we nitpicked little details until filled up our interview list. Then we conducted case interviews to figure who we would extend offers to. I doubt the interviewees who made it through case interviews were much better than those who didn't, but those tough interviews certainly helped us build a reputation of elitism.

Diogenes said...

was this guy a quant? or was his role purely "technical"?

i read a want ad for a quant which said, "for marketing purposes only graduates of elite schools will be considered." this guy wasn't that.

branding and complexity are both deliberate attempts to avoid competition. govt does have a role in forcing commoditization when the market hasn't. health insurance and insurance more generally is a great example. today p & c loss ratios are only 50% and insurance is one of the biggest advertisers.

steve hsu said...

I've been in the position of having to select teams to accomplish *real* tasks with objective standards of success -- not just impressing clients. HFT firms, most software startups, and most science and engineering teams are also in this position. They may be misguided as to the best techniques for filtering human capital (brain teasers are noisy indicators, but can give you some insight into how a candidate thinks, etc.), but their intentions are real. A good quant/developer/technologist is worth 10 mediocre ones ...

Cornelius said...

My argument is that given the high ability level of those who make it to the interview and the noisiness of the process, cost of conducting interviews > value gained by getting slightly better candidates by conducting those interviews.

If that is true then either firms are irrational or they are getting some other value from the vetting process.

David Coughlin said...

$GS casts a pretty wide net looking for people to fill the back office. You probably are not going to be in the 8-figures/year bracket in the back office, but I have heard that Aleynikov was making $1.7M salary+bonus. That is starting to be professional ballplayer money, and the naive you-don't-own-my-work play holds a lot less water.

David Coughlin said...

I think that you are misallocating noise to the process. The noise is in the candidates, and the noise is probably much higher in candidates who don't recognize the tricks.

dxie48 said...

"They may be misguided as to the best techniques for filtering human capital"

Unfair advantage for those familiar with the ancient holy vedic numerical techniques and short cuts,

"Nirvana By Numbers"

dxie48 said...

Citadel HFT in action,


dxie48 said...

"caused by slow, clunky software—and that was Goldman’s problem"

Compare to modern direct application specific FPGA based hareware trading platform right at the network switches. Market data are extracted even before the network data packet is completely decoded.


What is the value of outdated software ?

dxie48 said...

The vanity Fair article was biased, using words which have totally different meanings in common life and computing without further explanation, such as

"subversion' In computing means a newer version of the software, nothing sinister. The agent seemed to use this word many times and seemed to imply sinister actions.

"super user status". Anybody can be a super user by booting up some of those legitimate DVD from computer magazine with no installation. The question is super user to which computer. If he needs to patch operation software he will need super user status. Nothing that super about it.

Cornelius said...

I agree that there is much noise in the candidates but based partly on
the fact that the biggest component determining whether someone passes
an interview is first impression, I still believe that the noise in the
process dominates.

David Coughlin said...

That is conventional wisdom that bears worth making its relationship to the $GS hiring process explicit before you accept it in this case.

yulva said...

The most important take away from reading this article is to never talk to the police (FBI). You are under NO obligation to answer their questions or give them permission to search your person, car, house, whatever.

A law school professor and former criminal defense attorney explains why you should never agree to be interviewed by the police:


Thomas Drake, a former senior executive at the National Security Agency (NSA), relates his experience in speaking voluntarily regarding illegal activity by the CIA and was subsequently charged under ten separate felony counts, 5 of which were brought under the Espionage Act...


LondonYoung said...


Before you tell the FBI you smashed your hard drives to bits because you didn't want them to see your pornography collection, remember that they are fully prepared for busting Al Qeda's cells trying the exact same thing.

MrJones said...

This interview process sounds like it's designed to filter out the exceptional deception skills of high IQ sociopaths.

Blog Archive