photon137

In my view, as an investor in Japan, these firms are the best entry points to getting access to business relationships (and key projects) across Asia. One can imagine these to be huge aggregators of business relationships and as central points (besides the banks which are getting away from cross-shareholder relationships) into the major _keiretsus_. This is the entry fee for being really in the know in corporate Japan.

And, as another commenter mentioned, getting a sweet arbitrage on the low borrowing rate and earning a nice dividend yield doesn't hurt.

Even if you're looking for causation, detrending is usually necessary to obtain consistent estimators (in statistical terms). For example, Granger causality works with two stationary timeseries:

https://en.wikipedia.org/wiki/Granger_causality

Adding a common deterministic trend to two random time series does not make them correlated. Intuitively, you still cannot predict the deviation from the trend of one timeseries using the deviation from the trend of the other timeseries.

His statement (after adding the constant trend) is misleading:

"Now let’s repeat the same tests on these new series. We get surprising results: the correlation coefficient is 0.96 — a very strong unmistakable correlation"

What he's calculated is the correlation between a set of points from y_1 and y_2 - and that will be large, of course - their (deterministically increasing) means have correlation 1.

The quantity that qualifies as correlation for predictability purposes is actually the correlation between the deviations from mean.

This is all fairly clear if you use the actual formula (Pearson correlation):

E[(y_1 - mu * t)(y_2 - mu * t)] / (sigma_y1 * sigma_y2)

What the author is trying to explain are the concepts of cointegration and stationarity. A useful introduction here: http://www.uta.edu/faculty/crowder/papers/drunk%20and%20dog....

Peak-power output would be less for an inclined rail (vs the vertical shaft solution).

Using a cluster of Owin (self-hosted) servers with NancyFx on Mono to host REST APIs behind nginx for a distributed-computing solution on AWS. Also use EntityFramework on Npgsql+postgres - experience has been positive.

Only downside is that for something like a customer-facing portal, the identity-management is tied to Windows-specific features, for now - so have had to build a custom solution for this that runs on Linux. Waiting for ASP .NET vNext and its Identity solution when it becomes slightly more mature.

EDIT: Also evaluating a move to Azure. Primary driver (for choosing the .net stack) has been prior experience with C#, .NET and Visual Studio

Great universities.

English as the primary language (great learning resources available for it all across the globe).

Real economic mobility - you can really make it big if you're smart, persevering and slightly lucky. Who you know matters to a certain extent in all countries, but the significance of that factor is less in the US as compared to the other countries, in my opinion (others may disagree, of course).

Tom Sargent, the econ Nobel Laureate, has a great website with examples in numpy/scipy solving some well-known micro/macro problems.

http://quant-econ.net/

With Chrome rolling out AES GCM (and other browsers likely to follow suit) - this seems to be a really plausible future state of the world.

No you can just download the free isolated/integrated Visual Studio shells - PTVS works just fine. And, imho, it's the best Python IDE I've worked with (I mostly use pandas/numpy/scipy)

Yes, but what is a non-differentiable curve if you and I don't even have the same definition of a curve? What about a curve which is differentiable in one coordinate system but not so in another?

EDIT: My point is - mathematics is axiomatic in its very basis - the axioms have to be agreed upon by people who agree upon a conclusion derived from those axioms.

Education here is not an overriding factor. The way a congenitally blind person "imagines" abstract representations of shapes is substantially different, primarily because the physical wiring of the visual cortex and indeed a lot of the brain would be very different when compared to a person who could see. The dimensionality of their world, despite being the same as that of the sighted ones, may not involve the same descriptive system - ie maybe "polar" (just an example) coordinates would be more "natural" to them than a Cartesian system.

It's simply a problem of what gets mapped to what in the absence of mapping channels we take for granted.

Education is probably not a significant variable here.

But the principal eigenvectors can change if scales are changed - and the directions in which they may then point can be completely different.

That's also true. But having the ability to hire people who can hit the ground running with production code is an advantage.

I've often wondered about using Go as an architectural choice - it's near-native fast, the syntax and semantics are good, it's GC'ed and runs on Linux (edit: my language of choice is C# - and Go is the only one that comes close to it _and_ runs on Linux, IMO - any others out there?- performance is critical).

However, if I decide to go with Go as a framework choice, is there an active market of Go developers out there from which one can hire?

And he worked on Tintin (which they label as Tin Tin - which makes me cringe like anything!)

If anything needs to grab my attention, 6 seconds is probably overkill. Twitter did it with 140 chars - they're doing it with 3/6 secs whatever.

Hell, C pointers have been doing it with 4 and 8 bytes for ages! :)

It's not an issue of installer checkboxes - it's an issue of abuse of trust. Would you feel the same way about the no-doc home-loans that were handed out just so that some brokers could make some money? They were simply some checkboxes too that nobody cared to read through - and look where it led us.

Extremely disappointed. If integrity in all aspects of a business is a lower priority than growth, then I don't suppose there is any difference between Wall Street and Silicon Valley investors anymore.

Questionable practices should be just that - questionable - and remain that way. This "ironing" over by stalwarts like pg poses the danger of this stuff becoming the norm over time.

Natural convection in a system heated from below - try and detect when the heat transfer switches from conduction to convection - maybe compare Nusselt numbers computed versus what's predicted by Navier-Stokes (no inter molecular interactions) ?

Edit: Another one - model the human circulatory system - trying to find optimal postures for maximum heart efficiency. Also simulate blood-flow keeping track of anti-coagulant factor concentration to detect high-risk areas for blood-clot formation.

This is just plain rhetoric and has nothing to do with real world trading. Anybody trading options who takes an outright view is doing it wrong (unless they have extremely good reason to do so).

The market is not dumb and prices volatility accordingly (hence a skew).

What you're talking about is a tail event - most sane option strategies protect against tail events (unless you're writing options and not covering your behind).

Most views via options are on spreads or on volatility where the maximum downside is known (in fact, betting in favor of a tail event is an extremely cheap strategy - via a strangle centered around the current spot - the deeply-out-of money options trade near zero while if a tail event does happen, you're entitled to a handsome payout - the max you lose is the amount you paid upfront - which might be as little as a few hundred dollars - essentially it's a lottery).

So even though your underlying stock may fall off a cliff, you wouldn't care if you had a sane option portfolio because what you get paid for is the distance covered by the drunkard - whether he covers it by falling off a cliff or going zig-zag ad infinitum doesn't matter.

Suggestion: In practice, for stock-options, volatility is strike-dependent ie implied volatility has a skew (lower strikes have higher implied volatility - the leverage effect). Also, the volatility is dependent on time to expiry of the option.

Also, as I've mentioned in the HN thread for your previous article, single-name stock options which are exchange-traded have an American exercise type. The valuation of these cannot be done using Black-Scholes if they are (a) long-dated (ie time to expiry is quite long) or (b) they are deeply in the money/out of money or (c) have an underlying which has a significant dividend yield (in which case you'd want to own the stock rather than the option).

Puts behave different than calls if the exercise-style is American (puts have a limited upside - so you wouldn't wait too long if the underlying stock has fallen far enough - your payoff is not likely to be larger).

You may want to discuss this and option greeks the next time.

Whenever you're talking about vanilla options, _always_ mention the exercise type - European or American (or Bermudan/whatever). Single-name stocks like Facebook, Google have exchange-traded options which are American. Index-options such as those on the S&P500 are usually European.

Also, it's a good intro for buy-and-sit strategies but rarely does anyone sit on options till expiry. They try to make some gamma-based profits (delta-hedged option) or some vol-based profits (strangle/straddle etc).

They are not Gaussian random vars - a very dominant growth factor ("mu" as you would call it in a stochastic process) applies to both - which is higher in Google's case - without a doubt.

A domain in which GAs are slowly finding resurgence is algorithmic portfolio management - I work in this area a bit (ie portfolio management - but not directly involved with the GA guys - plus their research is proprietary) - but their premise is that evolution has done pretty well surviving against nature and products of evolution itself.

The financial markets are a bit like that - stochastic exogenous factors (headline risk like the Euro crisis, the FOMC meeting decisions etc.) and evolved responses based on the principle of no-arbitrage (ie other traders) determine portfolio performance and trading decisions along with transaction costs.

But GAs as they exist are unsuitable for this, from what I've gathered from conversations with those quants. The modifications required appear to be in the direction of structured intra-generation genome modification (or adaptation) rather than pure randomness and crossover to produce agents with a specific trading behaviour (much in the same way as evolution allows a sensible change in the DNA to produce a particular protein suited for a task).

There is growth indeed via Apple, Samsung and other licensees. But that's mostly within the purview of Apple, Samsung, Qualcomm etc. ARM doesn't have to scale for that growth - Apple and Samsung do - yet ARM's designs reach millions of us.

As I understand it, the whole reason Valve built Steam was to cut the publisher out of the value chain (who, to be honest, had too much power and added little value for the customer).

Incremental patching and DLCs are (extremely beneficial) side-effects.

> "Businesses that are good for society need to grow quickly to actually reach a large fraction of society"

That is not true (and is more relevant to manufacturing than to tech companies). Look at ARM, for example, it never needed to grow rapidly and it doesn't actually produce stuff - it just licenses intellectual property - but the existence of ARM benefits millions. Would this fledgling firm in the early 90s Cambridge not be called a startup (was Apple dumb to fund it as a joint venture?) - would VCs not have liked to have a piece of it in the early 90s?

EDIT: Same goes for id, Valve and other game studios - you see sometimes it's better to have the best engineers and researchers take their time, let them develop and mature the technology they're working on and build something long-lasting that affects millions of people.

I am not terming that as BS because it's impossible - it's not. I'm saying it's not suitable for all businesses (it might be downright hara-kiri for some of them) - that alone should not disqualify them from being known or viewed as startups.

Couldn't agree more - let the entrepreneurs define and live the term "startup" - not the VCs.

(PS: After sleeping over pg's essay, I found it much less convincing the next day - implicitly, he's mixing up price with value, ignoring the temporal nature of markets and of opportunities - Google couldn't have been a startup in 1989 and Ford was a startup in the 1910s - and disregarding multi-product and B2B startups (premium Tibetan to Hungarian services) altogether. The clincher from yesterday - weekly growth of 7% - sounds like utter BS today.)

EDIT: Not sure why I'm being downvoted. I'm just pointing out that this article only and only defines a startup in terms of its worthiness to investors. It's as misleading as any other single-formula characterization of a complex multi-agent system. Startups != Growth, period.

"For a company to grow really big, it must (a) make something lots of people want, and (b) reach and serve all those people"

Very valuable insight.

However, (b) in its own right, can serve as a fast growth business model - where the delivery or "clearing" of value between those who demand and those who supply is the value proposition itself - because everyone wants delivery (making it, by default, a big market).

Most banks work on this principle, in an abstract sense. A business like FedEx or UPS is a more physical example of this.

Online takeout-ordering services are examples of this - the customer wants the food and the "online ordering website" startup does not produce food - but what it produces is "clearing" ie matching demand to supply.

This - as an idea, in my experience, always scales and grows fast as well while falling into the category (b) that pg mentions.