Powered by glolg
Display Preferences Most Recent Entries Chatterbox Blog Links Site Statistics Category Tags About Me, Myself and Gilbert XML RSS Feed
Saturday, Feb 28, 2015 - 21:06 SGT
Posted By: Gilbert

Second Half-Life

Reform means redistribution to the upper class;
Entitlement means redistribution to the middle class;
Welfare means redistribution to the lower class.

 - Lingo of the times, reddit.com


The past week's been awash with unmentionable rumours, and in keeping with the official stance on speculation for once, we shall not explore this topic further.

There were some interesting developments in the latest national Budget, with Our Most Successful Investment Firm finally deigning to dip into its pockets (N.B. I'm not sure if the quoted Barclays economist got their average returns right) [Errata 1 March 2015: Oops, he did. Got bamboozled by all that "14% over thirty years!" track record shilling]. Other than that, the powers that be are belatedly trying to promote non-degree career paths after the last unconvincing cheer squad hack job... with the ambitious target of placing one-third of the cohort in ten years, presumbly some in S$1300/month waiter roles. But okay, mai hiam lah.

On the tech front, automation is not only poised to overthrow the automators themselves (again), but even their pets - what is the world coming to? Just as well, perhaps, that we might soon be able to downsize by shedding unnecessary bits. This reduction of energy requirements will probably qualify for rebates in time to come, the way things are going. More reason not to be too obsessed over one's diet?

That's about as far as protecting parts of oneself that are private can go, though, as one realises that the NSA (and associated allies) are like a Pokemon trainer... in sharing the motto, "gotta collect 'em all!", having pre-hacked billions of SIM cards. Then again, given how many people I know willingly load their computer with spyware, I'm not sure I can even care too much about Lenovo doing it for them.

And for once, a dress made the headlines without any figure threatening to slip out of it, in the best traditions of Lopez's Versace number - a decent if fairly unadventurous piece (not that I'm any expert) made waves after it was realised that some saw it as white and gold, while others saw it as black and blue (which is apparently the correct colour combi)

An admission here: despite various explanations and tricks, I have been unable to perceive the dress as anything other than white and gold. Of course, as the checkershadow and similar illusions demonstrate, the eye is not particularly hard to fool, and nor is different people seeing different things that out of the pale, given the various forms of colour blindness (possibly also solved). What was remarkable was how evenly the opposing camps were split, which was probably what fueled the debate to such heights - it wouldn't be so enticing if no-one else were on hand to disagree strongly.

Finally, my discovery of the week - it turns out that I might have unwittingly acquired an Erdős number of 4 some time ago, that is if one accepts publications in computer science venues as valid for this purpose. That said, given how networks work, it is probably more remarkable to have a large (finite) Erdős number, as the median is by all accounts about five. Well, given how I just struggled with a relatively trivial trigo transform, I'm more shamed than anything.

Oh yes, and tried some Counterstrike at Bukit Timah Plaza. cs_iceworld, it never gets old!


And Life Continues

So where were we... ah yes, Bell Labs with Derman at Chapter Seven. This was his transition from high-minded academia to *shudder* working for filthy lucre (Bell Labs more than doubled his salary from Boulder), and he began by observing that he considered it the first time he actually commuted to work, because he never considered his previous doing of physics as "work" (a sentiment exemplified by Erdős)


Depending on what one is working on, I wouldn't either


Since he saw himself as beholden to the institution for his pay, and therefore also its myraid bureaucratic rules, he went as far as to call his boss's secretary when his car ran out of gas, to warn her that he would be half an hour late. She had to exasperatedly inform him, when he arrived, that nobody actually gave a shit.

The other cultural shock was that Bell Lab employees were not allowed to publicize their academic credentials (this made sense in that new hires would often have to report to managers with less-advanced degrees), with AT&T instead imposing their own absolute ranks upon their staff. Still, the Labs hierarchy was shifted up half a level from AT&T's main one, so Derman was "almost a supervisor" when he entered. Further, he notes that while pure talent was everything in physics, rank and politics took precedence in organizations.

Still, it probably wasn't that bad of an environment. Derman noted that such titans as Bardeen, Shannon and Dennis Ritchie had made big advances there, not to mention the many slightly-lesser lights, even if he was frustrated by his supervisors' cryptic decisions as to whether works could be published. Moreover, life was easy if you knew how to coast along, and Derman mentions pondering classic combinatorial optimization puzzles with his colleagues (the R-rated 2/2 formulation, later generalized to N couples; it's only gotten better from then, as this effort from Stanford demonstrates; and you thought academic life was profoundly unexciting...)

Oh, and he picked up programming, with C (yes, without the plus-plus) having just been introduced at the Labs by Ritchie. Derman spent several years implementing HEQS (Hierarchical EQuation Solver), a nonprocedural language for solving equations (while Derman gave an URL at which it could be purchased from Lucent [a descendant of Bell Labs], the passage of ten more years seems to have obsoleted it)

Despite all the perks, Derman remained unhappy at his perceived subservience, which even his three year-old son picked up on. Not only that, he felt demeaned by the "moneyness" [sic] of his work there. Obviously, the answer was to go for even more money on Wall Street, and so that was what Derman did.


Yes, No, I Dunno

"Economics is a profession of people who weren't good enough in math to become physicists who go on to spend their careers trying to prove that they were..."

- an MIT-trained math prof, whose comments were clearly not well-received by her economist counterparts


Now, the p*****s envy of economics and finance has long been documented (mostly by those who're actually trying to explain real-life observations [often unsuccessfully] from theory; see Machine Dreams: Economics Becomes a Cyborg Science for a discussion of how [some of] modern economics developed from computational theory), but it remains unclassy to press the issue in polite company (Feynman didn't care)

It's not that mathematicians and physicists can't be wrong, though - many heavyweights have held opinions (e.g. deities and dice, the non-existence of electrons, etc) that were later shown to be incorrect. Even pure mathematics, where errors can only come from logic and not incomplete information, has seen its fair share of refutations and retractions (N.B. I do not profess to understand the given examples)

Seeing as how specialized subfields are getting, one can easily understand how a new paper's referees might all miss some subtle quirk in dozens of pages of dense formulae, especially given how nearly a thousand Ph.D.s fell at something so comparatively straightforward as the Monty Hall problem (and even when the entire premise was fabricated... or when some cells in an Excel spreadsheet were accidentally missed)

And a last digression on Erdős here - one of his beliefs was in "The Book", in which the most elegant proofs of all math theorems are kept; one could, certainly, suspect that there is No Such Book, but that would be cruel. But if so, it would seem that The Book would have to be infinite, if there is no upper limit on the answer to an "interesting" math question... or even its length, i.e. we cannot even begin to conceive of almost all questions that exist! A humbling thought, indeed.

Unless you're a local primary school science teacher, in which case keywords rule all (yes, it's worse than previously reported). Thinking schools, learning nation, indeed. Well, at least they might introduce computer science as an official subject at secondary school level. About time. Oh, and the triple science stream might be making a comeback at JC level too - one of my teacher friends mentioned how chemistry had inadvertently become the common denominator, since it sits between physics and biology, and taking all three is (was?) not allowed. Well, none of my business.


We Resume Normal Programming

Ok, ok, back to the book. Derman did interviews with Wall Street firms, and was soon made known of the difference between equities (stock trading) and bond trading. In the former, there was nearly no math involved - guts and balls were everything. Whereas bonds, with their "coupons" and "interest rates" and "maturity dates", admitted calculus! Obviously, this was what he was going to be hired for.

Derman did have something else going for him - this was 1983, and the whole concept of a "quantitative analyst" was still very much up in the air. "Financial engineering" wasn't even a recognized discipline yet, and firms were impressed by prospective hires who knew enough math to calculate bond yields, and knew enough coding to cobble those equations into software for the traders. As it was, Derman notes that CS Ph.D.'s were often not so good with continuous math, while math and finance ones didn't know how to program, which left the physics and engineering guys.


Yes, that's twelve sheets taped together.
(Source: quora.com)


So, by now forty years old, Derman embarked on his second life at Goldman Sachs, beginning with work on bond valuation models and rewriting old FORTRAN code in C. Interestingly, Derman loved his time at Goldman, despite by all accounts being pressed harder than at Bell Labs. Talented colleagues aside, perhaps his salary tripling had a bit to do with that.

One of those colleagues was Fischer Black, who co-invented the Black-Scholes model, and whom Derman suspects was denied a Nobel due to the committee being slanted against industry people (moreover, Fischer wasn't much for fashionable lemma-filled papers). They soon wound up developing the Black-Derman-Toy model together. Being a new hand, Derman was dreaming that it would be considered a grand unified theory of interest rates. The more worldly-wise Fischer in constrast saw it for what it was - a useful tool.

Despite all the happiness and newfound zest, Derman soon realised that research-types were very unlikely to make partner, since their contributions were not easily quantified. And, despite money not being a motivator for him not all that long ago, it took just a few years for him to feel dissatisfied watching young traders riffling through their stack of US$100000 bonus envelopes. Using the old jacket-over-chair trick, Derman began doing the interview rounds again, and doubled his (already considerable) Goldman salary upon moving to Salomon Brothers.

Long story short, it was a disaster. Derman reported an especially selfish, every-man-for-himself culture at Salomon (charitably described as "tough"), even by Wall Street standards. Quants refused to share even the simplest code snippets; employees were preemptively fired for interviewing elsewhere; author names were removed from reports once they had left. Unlike Goldman, which was at least "long-term greedy", Salomon was all about the short-term.

The best he could say was that they were good at using quantitative research to generate business, and by the time his second year there rolled along, he was just sticking around for the bonus. He was let go during a round of layoffs in late 1989, and while not explicitly mentioned, one can sense relief.


The Circle Complete

Derman's old mentor Fischer Black was nice enough not only to hire him back, but further recommended him for his old post as co-head of Quantitative Strategies, Equities Division. This brought with it the (now-cancelled) privilege of free lunches (long before Google), which he however disliked, as it led to overeating without fail.

He got back into the game with exotic options, most notably "guaranteed exchange rate" Kingdom of Denmark puts, which threw the Danes behind customized put warrants on the Japanese stock exchange. Can't get much more exotic than that. From what I could make out, arbitrage seems to be the soul of finance.

Derman describes much of financial modelling as following what physics would term an "inverse scattering problem" - whereas most physics models (after the relevant laws are found) involves applying those laws, the process generally goes in the other direction in finance: given the observations, guess the laws, which was what he did with implied binomial trees during his second stint at Goldman. Still, his conclusion, not as a fresh-faced newbie but a grizzled vet of twenty years, was daunting - in finance, there is no true model.

And he was clearly no mug by then, describing Value at Risk as "an unsatisfactory risk metric that has somehow become an industry standard", four years before it got blamed for precipitating the 2008 crisis. The enormity of risk versus uncertainty dawned on him, and towards the end of his financial career, he began to regard option valuation as analogy. Adapting Hillel's Golden Rule, Derman was left with one commandment: "If you want to know the value of a security, use the price of another security that's as similar to it as possible. All the rest is modelling. Go and build."

Sounds like sound advice to me.



comments (0) - email - share - print - direct link
trackbacks (0) - trackback url


Next: Topic Mashup


Related Posts:
A Life Quantified
Clear Is The New Black
Excuse Me, Are You A Model?
You Bet Your Life
Light It Up
Back to top

Copyright © 2006-2025 GLYS. All Rights Reserved.