27 February 2015

Quantifiably Nuts

There's something of an open secret in the financial quant world: it just doesn't work any better than passive investing. Every now and again yet more data appears to bolster that truth. Today's comes in James B. Stewart's column. A mother lode of quotes; enough to last a year.

But, here's the punchline:
Large investors are still pouring money into hedge funds. They added $1 billion during January and more than $88 billion in 2014, according to data compiled by the investment consultancy eVestment. Total hedge fund assets are now over $3 trillion.

Where's the punch?
Even as their high fees have minted scores of new billionaires, hedge funds have now substantially underperformed a simple blend of index funds -- 60 percent stocks and 40 percent bonds -- for three-, five- and 10-year periods. And the 10-year numbers cover the period of the financial crisis and the sharp decline in stocks -- the very calamity that hedge funds are supposed to protect against.

That last bit is the loony bit.

For those too bludgeoned by the remorseless drum beating from these here parts to go off and read the whole thing, just some of the high points.
"There's no money in being against hedge funds," [Simon Lack, founder of the financial consultancy SL Advisors and author of "The Hedge Fund Mirage"] said. "If you're a consultant and say, 'Put it in Vanguard,' they won't need you anymore."

Kind of the same sort of notion of web sites paid for by adverts: in the web site case, the client is not the user of the site but the various advert pushers; in the hedge fund case, it's the funds who are the real client.
Others I spoke to cited data suggesting that hedge funds have a lower standard deviation, a measure of risk, and higher risk-adjusted returns than stocks. But Mr. Lack warned that such data is necessarily backward-looking and does not account for the larger amounts of capital now invested in hedge funds. When he wrote his book in 2012, "I said hedge funds were overcapitalized and returns would go down, which is exactly what has happened," he said. "The best funds know this, which is why they're closed to new investors."
(my emphasis)

Once again, with feeling and three-part harmony: there's more moolah chasing non-existent real investment opportunities.

As always, the only way to avoid significant loss is to exit before that last straw lands on the camel's back. How to recognize when the straw is coming in for a landing? (Re)-read your Samuelson. Tidal shifts in moolah movements are what matters to the hedge funds and the pension funds and the .1% wealthy who seek them. (Why did they miss the silly acceleration in house prices, one might ask?) When there's tens or hundreds of billions of dollars in need of a resting place, buying a few thousand shares of a soon-to-skyrocket biotech isn't sufficient.

Simon Lack:
"There's an enormous amount of research that shows hedge fund returns aren't persistent. They revert to the mean. Of all the hedge funds I looked at, only 7 percent were consistently in the top 40 percent. What chance do you have of picking them?"

26 February 2015

Deming Down Analysis

W. Edwards Deming is reputed to have said:
In God we trust; all others bring data.

The latest place I've seen it is in the preface to The Elements of Statistical Learning, with the caveat that there's no proof Deming actually said it. The problem with Deming is that he was grounded in manufacturing quality control, Shewart control charts and such, where processes obey God's laws. In the human world, data isn't really the issue. Changing behavioral incentives are the issue. We humans, some of us, change the rules/incentives as suit some of us.

If we consider what caused The Great Recession, it's easy to blame The Quants. I've sipped that Flavor Aid some times, alas. Still, one cannot forget The London Whale and his mess. That, beyond question, was a quant going off the reservation. But, what exactly, motivated The Great Recession?

First, and mostly, was the massive increase in availability of funds into US, and Europe lesserly, housing. It's reasonable to ask, why? Did the quants, somehow, fiddle it. The answer is no. After the bursting of the DotCom Bubble, moolah began to pile up, yet again. Corporations' real investment demands just couldn't keep up, at a level of risk holders of the moolah were willing to accept. As usual, they wanted money for nothing (and chicks for free).

The subprime to liar loan spectrum was created by mortgage companies, not commercial banks, built into mortgage backed securities for the purpose of sopping up a tsunami of risk-averse holders of moolah. We see exactly the same motivation with the "Greek Crisis" (Spain is similar): holders of moolah want above market rate of interest at lesser risk. Housing, up until then, had been largely risk free; same with sovereign debt. But, because all the lemmings were headed in the same direction, they all went off a cliff. If they hadn't dragged the rest of the global economy with them, it would have been a minor event. The instruments came not from quants, per se, but from bankers, loosely defined.

Once the instruments came to exist in quantity, then the quants started to move the levers of arbitrage to make their money. Anticipating when the money flows are zigging before anyone else is lucrative. Such an exercise is closer to God's laws, in that the lemmings are blindly following the butt in front of them. With sufficient, accurate, data the quants could see where the plot was going.

Not to mention that folks like Paulson made smart bets, so it seemed, by putting a heavy thumb on the scale.

As some say, you can only sell a bad instrument to someone who is looking the other way. That happened in Florida, the Spanish coast, Ireland, and Greece. There remains trillions of dollars of "savings" looking for a place to sit. A lot of it sits on corporate balance sheets. Ever more of it moved to buy back stock, thus moving moolah from the pockets of 1%ers to pockets of other 1%ers. The right wing bleating about inflation is misdirection. Not all of them are idiots, and know full well that inflation, caused by too much moolah floating around rather than output shortage, happens only when Joe Sixpack has the moolah to spend on stuff. The QE monies, ours and Europe and Japan, have been wholly sent to the financial sector; thus the stock markets have shown impressive capital gain. D'oh!!! But this also means that the following must be true: corporations have about run out of productive ways to turn fiduciary capital into physical capital, at least relative to the supply of fiduciary capital. In other words, there is, by definition, a savings glut. Which deadly sin is gluttony?

Shiller, in a recent piece, had the advice that if one continued to live as a student, then the moolah would pile up and a better life to follow. He ought to know better. We already have, globally, more savings than can be productively used; that's why so much went into non-productive housing. Adding billions or trillions more from whatever letter generation is au courant will only drive down the interest rate more. So, yes by not consuming, your money will pile up. (And by not consuming en masse, we get a Depression, by definition. The 1%, with their hoard of moolah, will be very happy.) But it is only your deferred consumption, not growing by some fat rate of compound interest. That latter is among the ripest chestnuts in the econ/finance lexicon. Up through the beginning of the 20th century, when new bits of science were discovered daily and thus new forms of physical investment could be made to exploit such, the permanency of compound interest in the economy was obvious. But, as we've filled out our knowledge of math and physics and chemistry and biology, we're like that old paradox: "how long does it take to get to that wall if you step half-way each step?" The answer, of course, is forever. But the paradox also means that each step of progress is of diminishing size. Don't assume that the second, or millionth, step is as great as the first.

25 February 2015

HomeR Simpson?

If one considers release 1.0 of a software as home plate, then RStudio is sliding in. And more vi(m) goodies! Yum.

23 February 2015

Proven Right, Again

Thanks be to LinkedIn, the following posting out of IBM have come my way:
the most recent R blurb
the linked from the previous

Now, this isn't DB2 and R, but Netezza (aka, tweaked Postgres), but still. IBM appears to get it. Too bad they didn't ship me much moolah for guiding them to the obvious. Oh well.
PureData System for Analytics introduced the next level of analytics by running analytical functions inside the database, which simplified the overall development process and allowed more people to use the analytical functions and benefit from it.

20 February 2015

Your Money Or Your Life

One of the differences, if not the motivating difference, between micro-quants and macro-quants is the maximization function. That last bit is econo-quant speak for "what's the source, amount, and purpose of profit"? For the micro folks, it's just as much as the laws allow for my client company. For the macro folks, it's a bit murkier. Generally, macro refers to the sovereign, but could also be a region, continent or planet. The max function for the macro folks is "benefit to everyone" in the sphere. So, in many places water supply is held communally through local or regional government, and price is set to balance household, farm, and industrial/commercial demands. We see the conflict going on in the western states in these days of drought.

What with ACA and baby boomers and the poor banging on the door of health care, there's a particularly nasty example of the conflict betwixt the micro and macro quants. The guy's name is Martin Shkreli. If you plug that name into your favorite search vehicle, you'll get a long history of his activities. The short version: a one-time hedge analyst who creates "pharma" companies with the sole purpose of buying up old, little used drugs (generally from the sole provider) and raising the price a whopping amount. Doesn't do anything else with the drug.

He got kicked out of the first one he started, and with his current one, attempted to take an old Bayer drug and move it from $100 to $100,000, according to reports. Today, comes another piece on the effort. Bayer, seeing that they'd been ignoring free money, bumped the price themselves, although not so much.

If there were a "free market" in pharma, with bans on patents, what would happen? R&D would fall to governments entirely; as things stand now, governments and academia do far more R&D than the general public is aware. In fact, there is growing controversy over the practice of universities claiming patent incomes on the R&D they do, especially that which is funded in part or whole by the Damn Gummint.

What Shkreli is attempting is simple: if my company makes a life saving drug, patients owe my company the rest of that patient's life earnings. After all, without that drug there would be no life. Ah the twin myths of democracy and civilization; enough of us remain Cro Magnon. Ayn would be so proud.

19 February 2015

To Your Health

As mentioned here more than once (I'll leave gentle reader to dig out the articles), the notion that Apple Watch would be some sort of groundbreaking health-o-meter have gone up the pipe. This is just one of many precis` of a WSJ story that's behind a paywall. Nevertheless
Apple's original concept for the Apple Watch was one that would track blood pressure, heart activity, stress levels and other functions, the Wall Street Journal reported Monday, citing sources familiar with the device. But none of those features will make it into the final product due to a variety of issues, according to the Journal. Some features didn't work properly. Others proved too complex. And some would have required regulatory approval.

As described in these musings, health functions are measured *under* the wrist where all the plumbing is close to the surface. You do recall that the nurse takes your pulse there, not on top? And that, for most people, the watch is worn on top? It was silly from the get go. Well, unless Apple could harangue God into moving sense points around to suit The Watch.

18 February 2015

Higher, Or Lower, Learning

I shan't name the title, since ... well, you know. Looking at the table of contents on Amazon for one of the machine learning texts reveals that linear regression is now a machine learning technique. One might expect this crowd to shortly claim air and water are theirs too.

16 February 2015

My Wedding Cake

One tries to limit the occasions where one says simply, "go read this". But today is one of those days. Read this.

Why? Read this, too.

Taken together, we, yet again, have more ammunition in the battle twixt client hegemonists and the RDBMS crowd. Oddly, the simple-talk piece (seemed) to elicit noise from client coders. Odd, because simple-talk is SQL Server centric, of course. Although it does produce more than a fair share of "coding" pieces.

So, the upshot is this: with 3D NAND, we will have terabyte SSD by the end of the year, my guess (well, we sort of do now, but they're kind of expensive). Mate such with the high-core count, high-memory count motherboards, and for that $10K (and an Organic Normal Form™ database), there's not the slightest need for client fiddling with data. None. As the Holub piece I referenced in one comment asserts:
The maintenance manager is happily sleeping in the back office instead of running around changing ROMs.

Or shipping client updates all over the world, hoping that none of them fails.

15 February 2015

The Final Nail in the Coffin [update]

Eventually, and today's news indicates that it's arrived, the eventual consistency meme promoted by that knucklehead at Amazon will be seen as that fine suit worn by the Emperor. The NYT reports on yet another bank heist, but the point is:
"We found that many banks only check the accounts every 10 hours or so," Mr. Golovanov of Kaspersky Lab said. "So in the interim, you could change the numbers and transfer the money."

I can't count the number of these knuckleheads who've claimed that bank systems need not be ACID, since banks only reconcile by the end of the day. How very 19th century. Green eyeshade and sleeve garters. Perhaps real-time transactions would help?

Just to be fair, so to speak, I did a quick search to find a knucklehead. First in line was this piece.
If you can fudge the consistency and stay up and compensate for other mistakes (which are rare), you'll make more money. That's the space most enterprises find themselves so BASE is more popular than it used to be.

Don't you just love the admission of "fudge"?? I wonder how much fudge they'd abide if they knew that they'd go to jail if they've left the bank vault unlocked?

Time Loves a Hero, Part the Second

Nick Carr's (got it right today) latest missive deals with machines, trains at the outset, and fits in with a nag I've had for years. Which nag got prodded recently over the Greece thing, the Japan thing, and QE discussions generally.

The nag: macros, especially, have a penchant for equating time saved by some device or method with some monetary gain by practioners. This has often been the justification for multi-lane highways and public rail transport. "Workers lose two hours a day in commuting, which means $X billions in lost productivity!!" BULLSHIT. Always has been.

In an economy where everyone, and I mean Banksters too, earns only on the basis of piece work with hours spent on the rack of their own choice, would such a leap of faith even begin to make sense. For the commuter, those two hours would have been devoted to Angry Birds or snorting coke or drilling some significant other. Or, in the worst case, cleaning up after mewling juveniles. How many of you, gentle readers, have used the excuse "gotta get to work!", in order to avoid futher "interfacing" with your nasty spawn? Hmmm? Of course you have. How many of you, gentle readers, always manage to work late, or at least arrive home late, in order to avoid having to prepare sup for the mewling brood? Hmmm?

Ford's assembly line was a productivity raiser for labor (and if Smith's marginal product dream were true, would have made them very rich; which it didn't), and its value was easy to calculate. The value of a GUI wordprocessor over a character version? Having been through it, the stunning panoply of styling options is what users cleave to. Before there was Angry Birds to distract the office worker, there was Word's sizzle of bling obliterating the steak of content. Pretty fonts never made an author more cogent.

So, in these days of fancy financial manipulations everywhere to be seen, and silliness of WhatsApp getting $19 billion (sort of, may be) for its freedom, demand some solid numbers when a tout is telling you that the next "innovation" in compute will save the world billions or trillions of dollars and make us all so much better off. Back in the 1950s, "Popular Science" and "Popular Mechanics" droned on about how automation would give us all teeny, tiny work schedules and leisure time as far as the eye could see. The "Jetsons" came from somewhere, you know: "George Jetson's workweek is typical of his era: an hour a day, two days a week.". Said leisure time used to make the world a still better place to live in. Hasn't turned out that way? Of course not. Capitalists are Rands, and won't be satisified until the 1% kill off the mooching 99%, who never seem to work. Where did all those middle class paying jobs go anyway?

The Wall Street Journal (huh??) collected a bunch, and it's pretty close.

This one, in particular:
1982: "By 2000, most Americans will be experiencing a new prosperity. The problems of shrinking energy supplies and spiraling energy costs will be offset by developments in computers, genetic engineering, and service industries that will bring about lifestyle changes that will in turn boost the economy."

--Omni Future Almanac
Given that this was written within memory of the OPEC oil embargo, not so much.

I wonder, other than the communists and rabid socialists, who predicted that the 1% would hold so much more of the national wealth now than they did in 1950? Or, more importantly, how this wealth transfer has distorted all the data we use to make predictions? If all of business adopts the Apple mode (produce for the top X% and ignore the rest), how long can the global economy remain stable? See here, the first graph.

Source: http://www.infinitejeff.com/permalink/tag/income-distribution/

11 February 2015

As Pretty As A Picture

I won't name names since, in this instance it makes no difference, but there's been a swell of visualization and data exploration meme a growin' in the gloming. Part of this is driven by the R contributions of Hadley Wickham which have made pictures as easy to make as Excel is for spreadsheets. The problem with ease of use is that the ignorant can do great damage; I'm talkin' 'bout you London Whale. This is especially true of pictures. They're really easy to use for propaganda or obfuscatory purposes.

So, if you're serious about data exploration and visualization, not just looking for some cute way to support The Boss's Agenda, thou shouldst read up a bit.

First, Cleveland's "Visualizing Data". He spends a good deal of time pointing out the ways in which visualizations is nefarious.

Second, one of these on data exploration:
Tukey's "Exploratory Data Analysis" if you want the source of the Nile kind of thing.

Or, Pearson's "Exploring Data in Engineering, the Sciences, and Medicine" if you want something nearly hot off the presses. He's had a blog about the book, which has been linked through R-bloggers, but I've not seen a new posting in a while.

Picasso is quoted, "It took me four years to paint like Raphael, but a lifetime to paint like a child." With data pictures, go the way of Raphael.

I Sing The Body Electric

It was sometime in school, might have been junior high or as late as undergraduate, when the instructor in some class brought up the topic of electric cars. Now, this was decades before Tesla, so it was somewhat abstract. The question raised: would an electric car be less polluting than the current gasoline/diesel ones? The class agreed that it would. Well, except myself.

My point: one can only answer the question with some specific technical information. To wit: at the point of use, yes the electric car emits no exhaust, so is cleaner than the internal combustion alternatives. But that's only half the point. The juice to charge the car has to come from someplace, and at that time, the answer was, mostly, coal fired power plants. And, at that time, these were about the dirtiest creatures on the planet. So, my answer was: it depends. Much to my surprise, and gratitude, the instructor smiled and agreed. One small victory.

What that has to do with today, is that Joe Nocera, in discussing a book on battery tech by Steve LeVine, normally sane and intelligent, falls into the same intellectual trap as my classmates of decades ago. Now, of course, today's cars are cleaner and many (most?) coal fired generation have scrubbers and such, but the issue remains: only testing knows for sure. Joe, on the other hand, is sure:
And with so much less exhaust spewed into the air, the effect on climate change could be lowered.

The standard article of faith. It is just faith. But he does, in paraphrasing LeVine, make one of these endeavors' major points:
[LeVine] told me recently, "a loss of confidence in the U.S. in our ability to create a real economy" -- one based not on financial instruments or a real estate boom, but real products that would help create entire new industries.
In other words, what QE exercises were supposed to do, but ended up just funneling all that moolah to the banksters and others of the 1%.

It is a worthwhile read, nevertheless. If only to emphasize that humans don't make the laws of physics or chemistry, just find out what they are. Whether chemical batteries can ever have the energy density of gasoline, or even diesel? It is the portability of gasoline that makes all the rest of automotive society possible. I wouldn't bet on it. Better to structure a transport system of public and private conveyance that fits the limits of chemical batteries. But that's Socialism, so it won't happen here. We'll choke to death on exhaust and road rage.

05 February 2015

Faster Stud Muffin

Most guys have had a paramour who raised flattery to great heights. One aspect of this apple polishing is the sobriquet, Stud Muffin. Faster, stud muffin, faster!!

I've not spent time with Sql Server in a while, being on linux exclusively. On the other hand, I've kept an eye on Hekaton news as it passes by. simple-talk has a new discussion, with this delightful bon mot:
To date, Microsoft is stating that organizations testing In-Memory OLTP have seen transaction speeds improve by up to 30 times compared to past performance, with the best performance gains achieved when the business logic resides in the database and not in the applications.
[my emphasis]

Take that, javascripters!!! You're no stud muffins.

Note the first comment, of course, for a somewhat dissenting view. Note also that different engines use different caching mechanisms, so the simple notion of "more memory than disk footprint" will mean a sorta, kinda in-memory database won't necessarily work.

Johnny Knocks Ville

I'm no where near Knoxville, so I won't be able to attend (I'm not a fan of streamed/innterTube teaching, alas), but this announcement on R HPC contains this nugget:
But R has a reputation for being sluggish and inappropriate for large datasets. However, much of R's problems with performance and scalability are due to bad practices of individual programmers rather than being inherent limitations of R itself.


"The fault, dear Brutus, is not in our stars, but in ourselves, that we are underlings."