crdb

Chris Date's Introduction to SQL [1], or Edgar Codd's last book [2].

Actually both are worth reading to discuss differences in philosophy (e.g. "no NULLs" vs 4-valued logic including "unknown" and "not applicable" NULLs).

[1] https://www.amazon.com/Introduction-SQL-Mastering-Relational...

[2] https://www.amazon.com/Relational-Model-Database-Management-...

It gets more interesting with branded goods which are completely fungible. In theory they have a global MSRP or at least a globally optimised one, to avoid arbitrage, and brands get pretty advanced in fighting the grey market (for example, I read Chanel is now using custom molecules to identify fragrances destined for different export markets - the invisible ink serial numbers were too rapidly found and copied).

But the UK MSRP just dropped 10%. I'm looking forward to see how fast the MSRPs will adjust.

EntrePass has been catch 22'ed a few years ago. In order to even apply, you need to have investment from a limited list of approved investors. On top of that, you need to hire a certain number of people within a certain amount of time, and there are a bunch of other rules to be able to keep it.

The "standard" way would be to incorporate, capitalise the entity with around 40,000 USD, prove that the business will be a business (and not you buying a visa) with letters of intent from prospective customers and a detailed business plan, try and set things up so that you're not the shareholder AND the director AND the visa holder, and then apply for an Employment Pass (EP) under your own business.

That EP will be yours if you have received a high enough salary for the past X months (they ask for your salary history for years), if you have set yourself a high enough salary that you'll need to pay out every month, and - in many cases these days - the Ministry of Manpower (MOM) will probably still ask you for "manpower projections" which I think is a euphemism for "prove to us you'll hire a few locals". Oh yeah, and budget another 3,000 USD or so for the advice from Corporate Service Providers (CSPs) after your initial round of DIY applications get rejected for no reason given. The whole process should take you 6-12 weeks. Whilst a Singaporean applies for a company online for a few dollars and gets it done in a week.

Once you've jumped through those hoops, yes, things are pretty good. Low taxes, extremely efficient government that treats you like a customer, rule of law, low cost of living (compared to Western large cities, not Bangkok or Beijing), great weather, international community, easy travel to dream locations, and so on. I wouldn't want to be anywhere else. But it's not a pain-free process the way say, Hong Kong was before 2014.

As an example of a brilliant book with awful grammar, there's Heinlein's The Moon is a Harsh Mistress.

This definition of selfishness (which I agree is the most consistent) was strongly associated with Ayn Rand's [1]. Having grown up in Europe, reading her works was quite a shock.

The only thing I'd add is that in my experience "short-sighted self-interested decisions" are made by people who do not understand externalities, and/or who favour a locally optimal but globally sub-optimal scenario (defectors in the Prisonner's Dilemma). "Socialist" policies like a congestion charge or pollution regulation are thus consistent with individual rights and an individualist philosophy.

[1] see: http://aynrandlexicon.com/lexicon/selfishness.html for quotes from her various books.

It's very simple in Singapore too, arguably simpler. You go to the Small Claims Tribunal (you can get a date within days, anecdotally), the Registrar decides the claim in a few minutes, and it costs you a total of 5 quid to file the claim. The whole thing takes a couple of weeks at most [edit: or a day - see 1].

I think something similar exists in Switzerland, but the (German) landlady correctly guessed that as a non-German-speaking foreigner, I would not take the time and expense to attempt it. Imagine attempting to contest the necessity of "dusting the sides of books on bookshelves" to an arbitrator in a foreign language...

[1] "For a tourist claim, the Tribunals may fix both the consultation/mediation and the hearing within 24 hours of filing of the claim." - https://www.statecourts.gov.sg/SmallClaims/Pages/GeneralInfo...

Are you sure about the rent? I rented a 2 bedroom, with windows on 3 sides, next to a park and public transport, and a large balcony with BBQ, for 2 months (so short term let, higher rates), 20 minutes by tram from my office near Seefeld, for 1,500 CHF/month.

The landlady did keep my deposit (sorry, "had a professional cleaning done for the exact amount of the deposit", and sent me the detailed list as proof in German which I do not speak), bringing that to 2,250 CHF/month. A quick back of the envelope on real estate inflation since then brings me to around 2,500 CHF/month (1,700 CHF/month if you ignore the deposit trick).

That's not as cheap as a small town in Europe but it's quite cheap, I think, for a major international city and definitely when you take into account the starting salaries. Vaguely remembering, a graduate from a good university starting in strategy consulting could expect 90-110,000 CHF/year in Zurich vs 35,000 GBP (49,000 CHF) in London despite higher COL and higher taxes (same companies, which may or may not start with M or B). I have no idea about the numbers in San Francisco but I bet they are not so good either.

There's other issues with working in Switzerland if you come from Anglo-Saxon big cities, but COL isn't it if you are paid locally, unless you are dealing with an employer exploiting the unemployment rate in your home country to underpay you substantially (which is definitely the case with some French speaking companies in Geneva).

This is a common dilemma in finance, the key is the possibility of outsized returns made possible with leverage. Let's say you have done well at your bank for the last X years and you have an audited track record somehow and a decent amount of saved bonuses. Time to go work for yourself.

Option 1: hedge fund: take a large amount of outside capital, but also sink enough skin into the fund that your interests appear aligned. If things go well, you can become a billionaire within a decade. If things go badly, you're burnt out of the industry (with famous exceptions) and you've lost a significant part of your net worth.

Option 2: start a prop shop or just trade your own assets, take no outside capital, go more slowly and take less risk (as less hunger for returns) and you might have a lower probability of burning it all, but you can only increase your net worth so much within the same decade. At most a few multiples of what you started with.

If there's a bull market, the first strategy is especially attractive because a. you can level up like crazy because debt is cheap (on top of the investments in your fund which are also coming in cheap) b. everything goes up, so just find the more volatile stocks (e.g. tech) as they go up faster. This strategy has made many people rich but you need to find a way out before the crash, because you'll find the perfect storm of your financing terms tightening suddenly, your high volatility stocks cratering faster than the rest of the market (that's why you selected them!) and your LPs (investors) all wanting out at the same time. There'll be white knights but they'll take their pound of flesh.

The prop shop partner on the other hand can wait out the crisis and buy the bottom or safely trade the volatility. He has comparatively less pressure on all of the above. These are the moments of greatest opportunity.

I always thought the same applied to startups. Eschew massive financing if you can, aim for lower volatility markets, and start the thing in a downturn when you have less competition and people are more sensitive to quality, is a lower upside but higher probability of upside strategy.

If they're talking about traders "in Geneva", they're talking about physical traders. Can't speak for oil which we didn't trade but for ags and metals, you would take delivery since it's your business (getting stuff from A to B).

Speculation on price is not done by taking paper positions in one direction, but by limiting or increasing your paper hedge, aka trading the "basis" or margin between your physical and your paper.

The physical side requires a fair amount of "RTFC" ("caring what the harbormaster is saying") and being aware of operations.

As an example we were given in training, one of our grains trader went to inspect a ship before loading it with grain, and the local guy thought something was dodgy so he closed the doors, and shone his torch at the walls. Thousands of "stars" blinked back at them. Turns out the ship had been used recently to carry broken glass, which obviously made it highly unsuitable for grains.

But the more common stuff is e.g. overloading the ship on one side so that it leans on the side you know the controller will measure the load, or playing with the percentage admixture (basically dead beasts and dirt) to net yourself a bit more margin.

I had a really shameless colleague who flew to London twice a week. He never checked in luggage, stayed in his suit, and arrived at GVA about 20 minutes before flight departure. If there was no queue, he'd make it easily; if there was a queue, he'd use the final call as an excuse to skip to the front and said people always let him (hence my calling him shameless).

Without getting that extreme, there's definitely a case to be made for knowing expected processing time from arriving to the airport to being in the plane (and generally, the distribution of probable door to door time when you buy your flight). I've done LCY in under 10 minutes once, but was turned away at LHR because I arrived there 44 minutes before the flight and BA cutoff at 45. The biggest queues in Europe are usually going out, rather than into the flight.

Train is great if it is well planned. I did MRS-GVA by train and it took 5 hours as it was going 30kph between Lyon and Geneva. SNCF oversold the TGV and there were 10-20 people standing per carriage. The queue at Geneva Station customs added a further 10 minutes. Next time: Blablacar.

Having done ZRH-GVA many, many more times than I care to count in both directions, may I recommend the plane.

The flight itself is REALLY quick and very scenic, even (especially?) at night with the moonlit snowy mountains [1], which are to your side, not under you (try to look south), and the stars. You're barely getting used to being in flight and they are already handing over the little chocolate and asking you to fasten your seat belt. Both ZRH and GVA (especially GVA) are ultra fast airports where check-in and security are a breeze (I recommend the mobile boarding pass) and there is fairly little walking. You can definitely door-to-door it in much less on the plane even with checked-in luggage (unless it's substantially changed in the last 3-4 years). Price wise I used to fly that route for 30-40 CHF, very competitive. It was sometimes cheaper to fly to London via ZRH than direct from GVA.

Meanwhile, whilst there are a few direct trains, most have a stop in the middle to change trains (from memory: Bern? maybe Biel or something that starts with L) and you have to carry your bags back out and stand in the cold for a few minutes and stress about whether you're on the right platform. Also, the bag space in second class is quite limited. On the plus side, you avoid taking trains to/from the airports, or getting frustrated with Switzerland's low speed limits on the highway.

I'm surprised the OP uses North American cities as examples of airport connectors. Hong Kong, Tokyo (NRT) and Shanghai to cite three big international destinations have great airport connectors which make sense considering the distances at play. I've always wondered why planners don't send their team to visit the top 10-20 airports in the world to check what the competition is doing.

Singapore is an example of the airport being close enough that you might as well take a taxi and a rapid transit to the "centre" (what is the centre, anyway - CBD, Orchard? there are a lot of hubs) makes less sense. Montreal's situation looks more like Singapore's than Tokyo's.

London is an example of where the extremely expensive airport connector (Heathrow Express) ends up somewhere with nothing except cheap hotels and slow connections, which is why they're finally getting round to building Crossrail, which goes from the Heathrow terminals all the way to all the major hubs (including the City and Canary Wharf, great news for those who have to fly transatlantic).

[1] https://www.youtube.com/watch?v=_hvWN3jOvvM&t=1m20s

I should be more specific: this applied mostly to Grab, an Uber competitor (the South East Asian Lyft if you will), in its early days before it implemented its own surge pricing (which is invisible - the price changes upwards, but is quoted upfront after you enter your destination).

Today, both Uber and Grab have relatively short dispatch times (usually < 1-2 minutes where I live) at all times of the day and night.

The interesting phenomenon is that both are running incentive campaigns and depending on which one has the better incentives, the other one will end up with dispatch times over 10 minutes. Like many customers, I run both apps and pick whichever is the flavour of the week based on dispatch time. I have a very slight preference for Grab since they quote the price upfront, but generally, taking UberX after a Grab quote, it comes to the same number or within 5% either way.

Talking to drivers (I take Grab/Uber over 10x a week), I've noticed:

- virtually all (one exception in the last 4 months) drive for both, and have a phone for each (Grab does not have an iOS app, apparently);

- virtually all drivers have done it for less than 6 months, lending credence to the "fast food joint" theory of Uber driving (that it provides temp unskilled jobs with no long term prospects like fast food joints in the US);

- maybe 80% of drivers are part timers, often unskilled workers (e.g. shop staff, construction) who are moonlighting, often because they have a lot of kids (highest was 8) and their wage is not amazing;

- taxi drivers tell me virtually all bookings are now from either of the apps; one said he did only one physical street flag in the last 6 months! I anecdotally knew this to be the case for me, as even when the taxi queue is free I still book my car via the apps, so I don't have to pay in cash and because dispatch times are so quick;

- some of the incentives seemed pretty good: if you made $150 in one day (the average trip being around $6-8) you got a whopping $23 extra per trip!

Taxis in Singapore get substantial discounts both on car rental and gas ($0.70/l diesel for taxis vs $2/l petrol for everybody else) so the economics are funny; Singapore is one of the few countries where taxis are competitive with UberX. During surge hours, I take taxis on Grab since the rate does not seem to be surged unlike Uber, and there's always plenty around.

This is not true of all other countries I've used Uber in. For example, in Sydney, in the last 2 years UberX has been about half taxi equivalent fares; in Paris, Uber Black (UberX is now famously illegal) is slightly cheaper than taxis but you get a really nice car and a much friendlier driver.

Emotional anecdote: I had an older Malay driver who struggled a little bit with English and upon driving past Little India, pointed at his old school (he now lives a good half hour away as the area has massively gentrified). He said he got in with a bad crowd at school and spent many years in jail due to gang activities, but he climbed back out of it, got a job, married and had two kids. He drove Uber/Grab in part because the money was exceptionally good compared to his other options as an ex-con, in part because it was a way to have a car in Singapore where the permit for 10 years of ownership costs upwards of $100,000, the apps income offsetting the rental cost for the day. He wanted a car because his kids were proud and happy when he drove them to school instead of having to take public transport. He said this practically with tears in his eyes as if it was the most important thing in the world to him.

In Singapore, drivers told me to put a guaranteed tip in the "notes" section when it's difficult to find a driver, as many people add 5-10 dollars that way and get picked up first. Personal surge is also what came to mind reading the headline.

> Two popular shopping companies in India exactly did this, they completely abandoned their websites and went native app only. This combined with large set of permission requested by apps lead to worse experience in terms of privacy for consumers.

I'm surprised nobody has commented on your comment yet. I was in a meeting just this morning where my interlocutor assured me that over 70% of advertising in 10 years will be native apps since everything else is getting blocked or abandoned (and presenting it as an opportunity to do all the stuff you "can't do anymore" on browser).

That's not quite my argument. It's subtler than that: 1. the world is not univariate and there is rarely one variable that explains 90% of the variation; 2. each user has a different utility function and it is rare to find one that applies to all aka "objective quality".

What I am arguing is that one variable that DOES often matter is signalling - including the subtle signalling for taste that is linked to high social status, and that can be found in e.g. spotting the 1-inch long straight boutonniere and matt buttons that signal a Savile Row tailor. And sometimes the pure signalling in itself, the completely intangible value created by a marketing department, becomes as "real" as any of the intrinsic qualities of the product.

Objectively, you do not need these things. A Ferrari is ultimately just a car. To some people the utility is getting from A to B, and they have a Bentley cos that's what people in their social circles do. Others truly love a well crafted product and value the Ferrari for the driving experience. Others yet have a different utility function, like a race driver who told me he did not enjoy driving the "borrowed" Lamborghini his friend lent him for a few track days, because his Evo had better visibility and turned much better, even if the Lambo was better in straight lines. And then you get the signalling where the Lambo is bought for its price tag in a flashy colour a la Entourage agent whatever his name was. I even remember a hedge fund manager who whilst he was interviewing me proudly showed me his F430 racing photos and cups.

(was based on original comment that GL > BL) Blue Label is an excellent choice when another company is footing the bill at an open bar, and I know a lot of people who would be genuinely pleased if you served them a shot of Blue Label after hosting them for dinner. They are rarely whisky enthusiasts, but there is genuine utility in the happiness that you give them in the perceived value that has been created by Diageo's marketing department around the brand. These intangible things become, in my experience, increasingly important the closer you get to sales.

> I'm pretty sure people still buy those things because they see the value in investing money in quality products, not to signal something. Signaling is a consequence here, not a purpose.

Rather than write up my position on this, I will Defer To Authority and post this excellent Quora answer on signalling and its relationship with quality: https://www.quora.com/Why-would-anyone-buy-a-Rolex-watch/ans... (although I disagree on her premise that there is no objective difference in quality, especially when it comes to wine, but then I am a card carrying Frenchman so hardly unbiased). I particularly liked the ending:

> Of course, none of this applies to me. I honestly love Van Gogh paintings, not because it signals good taste within my social circle, but because they're intrinsically beautiful.

Well, I love my Port Ellen, Karuisawa and Mortlach and I would look forward to adding Urushi Seikos or Chanel's Cuir de Russie (vintage, of course) to my collections. For their intrinsic qualities, naturally. Unfortunately, my wife does not see those intrinsic qualities as clearly as I do, so my plans have been put on hold for the moment (and, probably, all future moments).

My impression has always been that drinking aged whisky was more about signalling (both financial status and knowledge and taste), including in the largest markets (France, etc.). Same reason people still wear suits and Goodyear-welted leather shoes for work. So there'll always be a large demand for aged stuff. If you just want to get drunk, vodka is better for hangovers and a lot cheaper. For example, a liquor trader here told me Indonesia is really into Chivas 18 year old, which is by far his largest export there.

That being said, economics have made it happen already: there have been no age statements on the "basic" offering from all mass market Japanese single malts (Yamazaki, Yoichi, Taketsuru...) for a few years now because the vastly increased demand effectively emptied the stock of the 10-12 year old stuff that was previously the default. I also get the feeling that a lot of Scottish houses are cutting their product with stuff that meets the age requirement for labelling but is not as good as what their products were made of 10-15 years ago.

In terms of taste, I think the "fast aged" stuff is a different product. Younger whiskies have a very different flavour profile (not necessarily inferior - try a Port Charlotte PC5-7). Can it become a new market segment, certainly - in fact it already has (e.g. Laphroaig Quarter Cask mentioned elsewhere in this thread). Replace, I'm less sure.

I had to do it with one client. My philosophy is one of automation and never doing something twice - the less hours the client needs per unit work done, the better - and the philosophy of some of their managers was the opposite, that you didn't get what you paid for unless people sweated for it. In other words, they incentivised their own workforce and consultants not to automate anything.

So I made a read-only log, shared it with them, and started writing down everything I was doing to make their products. It ended up being really interesting as I never really analysed my development process, all the false tracks you go down for a while before reverting, all the data format changes, type errors you meet, avoiding repetition, documentation, flashes of inspiration, or how waiting for a query to complete breaks your flow... and I think they got a better idea of what is involved in producing stuff (no, it's not "just updating a few graphs" if you went and changed all the definitions and they come from 7 distinct data sources).

A side effect is that they got precisely the hours they paid for.

With other clients, I routinely used to put in a weekend or late nights if things didn't quite work, or if I had an idea for an interesting product that could be helpful to them. With these guys, it was "this is probably going to take me 10 hours. Can you approve these hours?" and 2 weeks later they'd say "OK" and I'd get started only then. They were so hours-driven that when I passed them a new product I did do in my spare time, they refused to even have a meeting to look at it in fear of having to pay for more hours. Quite amazing.

I guess so. Retaliation is the best strategy in iterative PD, no? (I'm rusty on this stuff).

Nash equilibrium in a regular PD would be DD, but it assumes that the other side will defect. I think from experience that the payoff grid is different in business in that instead of CC resulting in a lower payoff for C than for D in CD, it's a much higher payoff; and the D in CD is not getting much more than in DD (maybe 20% more).

I wish some of my clients understood that.

One just ruined all the goodwill they built up with me over the past year (which got them countless free hours, crunch time at short notice, etc.) by stiffing my company of about a grand - literally a few hours' work - on some badly structured justification, the real reason being he felt his colleague was not negotiating hard enough with me and he "showed him".

Now I'm going to stick to the contract hours and probably prioritise other work over theirs; I also changed the payment terms to be upfront to avoid stiffing, and if they want extra hours, guess what, their sticker price went up from their discounted price for longstanding cooperative clients to "market rate". I guess the thousand bucks is visible and the value they lost isn't, so he probably thinks he was very successful.

It bewilders me that someone can be in business for years and not understand that cooperation is a better strategy, but there are a lot out there. I've found the short term cost of cooperating with defectors worth it in the long run, though.

McKinsey flew me and a few other fresh grads to Dubai for a week because I was helping run the entrepreneurs society. 5* hotel, mock case study for 1/3 of the time and the rest tourism.

It was a holiday for the McKinsey guys running that thing as well, they appreciated being paid to go waterskiing and BBQing on the beach. For me it was luxury unlike anything I experienced until then - the rooms were so big we had to almost shout to be heard across them, and there was a sofa on the balcony! I took home loads of Hermes toiletries and cologne (I didn't have enough money on my bank account for the room deposit, and one of the consultants had to deposit with his card). In fact we even missed our Paris connection to Dubai, so McKinsey sent a consultant over to baby sit us, which meant taking a Mercedes to the Brasserie La Coupole for an epic feast followed by a night in a neighbouring 5* (I think in the Tour Montparnasse).

Unfortunately, I did not graduate with high enough grades (2.2 on the BA, cutoff was 2.1) for McKinsey, which is why they rejected me the first time. So by the third time they tried to headhunt me, I just told them upfront that their system/HR would reject me because of my grades and it was not worth our time continuing the discussion. They insisted, and it happened as predicted, making me wonder why they still care about that for more senior hires.

There was also a period of time in my life where I accepted conference speaking slots if the hotel was nice enough and close enough to my home. Can't resist a proper 5* hotel lunch buffet, especially if they have giant prawns and olive-roast beef and a decent beer on tap. I made my talks increasingly ridiculous, I think by the end I was explaining Starcraft cheese to middle-aged CTOs and quoting Delta Force founder Col. Charlie Beckwith before following up with droid-driven tractors (all relevant to ML in e-commerce, I promise). They loved it, got voted "best talk of the conference" a couple of times, not hard when 90% of the other talks were "so here's the 3Vs of Big Data" verbatim copied from Wikipedia.

> If I want to keep it without being chained to servers that will go away, I break the law.

I was thinking of that today as I had to download (for the 6th time since I started) a Kindle book I'm reading to my Android.

I'm another 1-2 downloads away from going on a torrent site and grabbing a DRM free copy that will actually stay on my local storage, be read on an app that does not take 30 seconds to boot, and not require large downloads with every open.

The natural evolution of that is to buy a copy on Amazon to support the author and industry, then download a free copy somewhere so you can actually read it in peace.

Brilliant marketing:

1. get people worried about something they previously didn't care about;

2. as a solution, add a new part to the car that will need frequent replacement and maintenance, thereby increasing the CLV.

The relational model, by being based directly on set theory and first order logic, is the most robust method for managing data today.

Yes, SQL is not the best implementation; theoretically we could try out Tutorial D/D4/whatever, but it's good enough especially in the Postgres flavour which is built on decades of academic and commercial research. PostgreSQL is almost entirely declarative (the RM is declarative) - how do you propose to improve on that?

As a quick reminder, these are the things you have to build manually into your application if you're not using a relational DBMS (or check are included in your new black box solution for managing data):

- data being shared (concurrent access, writes, etc.);

- avoiding redundancy, inconsistency (which you get for free by centralising the data into a single copy instead of having one copy per thread/program/user);

- transaction (= logical unit of work) atomicity (all or nothing) - say you want to transfer money from A to B, decrease A, then increase B; what if "increase B" fails? A relational database will ensure your transaction does not half go through;

- integrity - impossible things are avoided by constraints, anything from "an employee logging 400 hours of work this week" to parsing different date formats (because the date is stored as string because it's an Entity Attribute Value antipattern because "the schema has to be flexible");

- easily enforced security ("nobody but finance accesses payroll tables");

- and obviously data independence, such as freedom from having to specify the physical representation of data and access techniques in your application code. (OK, this is one place where SQL is not perfect; but it's pretty good)

(*thanks to C. J. Date for the list)

The sad thing is that this stuff does not seem to be taught anymore; I get a lot of business from the fact that most frameworks encourage antipatterns by design (Bill Karwin's book on the subject [1] is a great, easy read for those who can't stomach C. J. Date's 1000 page "Introduction" [2]).

[1] http://www.amazon.com/SQL-Antipatterns-Programming-Pragmatic...

[2] http://www.amazon.com/Introduction-Database-Systems-8th/dp/0...

However, if you ping me on my profile email after I put up a job on Who's Hiring, and I send you back an email saying "hey, can you answer these few filter questions, take your time" and one of them is "what is multicollinearity", chances are you'll google it and tell me what you typed above and it will be fine. It's not like you can't use Google or take your time.

What I'm dealing with is people who are dumping their CV and a standard cover letter on me at high frequency and I do not want to put an HR filter between them and me, because I want to spot the guy without a LinkedIn profile (HR red flag), who dropped out of college (so no brand - HR red flag) but has plenty of useful experience (so no buzzwords, because real experience looks unremarkable on a CV - HR red flag). I know no more efficient way. Would love to hear of any.

I did make the mistake many years ago of rejecting a PhD in stats because she admitted not knowing about neural nets ("but my lab buddy is doing some research in them"). I'm still kicking myself, but have learnt a bit about hiring since then.

I run my own data consulting firm based in Asia.

Originally, we were going to build high dimensional statistical learning systems crunching large datasets on GPUs in Haskell.

90% of clients cannot feed me the data, so I end up building them a data warehouse first; in some cases we even redo their data model. 100% of clients have datasets too small to justify getting out of R and single CPUs, so our Haskell ML libraries are, so far, on the wishlist only (I think Tweag has had more success there, but I still have feeble hope that one day...) Hence the RM focus as well as ML. The ML side is sort of picking up this month.

That being said, I'm not actually that well versed in ML. Just the guy running the company. I've built (and read) just enough to know how it works, how it applies to clients and who to hire to get the job done properly...

There's a difference. Imagine a simple one dimensional linear regression.

A value with x close to the mean, but extremely high y would barely move the line as a whole up or down as you vary y. The gradient would not be much affected.

A value with extremely high x does not need to move much to change the gradient of the line. You might even change the sign of the gradient.

Both are outliers but the second has higher leverage than the first.

Edit: illustration: http://imgur.com/vVKwjHh (right hand side are high leverage)

What do you call high leverage observations then?

I'm referring to leverage in statistics [1], and particularly how high leverage observations may affect your model and how people deal with them. Clearer?

For context: one of the candidates I interviewed told me that the best way to pick a model was to pick the model that would have the highest R squared when fitted to the whole dataset. I asked him about overfitting and he didn't know what I was talking about. Same guy whose CV showed 4 years in a research lab doing stats [2]. A lot of people are just going through the motions.

[1] https://en.wikipedia.org/wiki/Leverage_(statistics)

[2] reposting for fun: http://4.bp.blogspot.com/-IJo_Tkw95-o/VANnTdCFgHI/AAAAAAAADV...

"Trivia questions" are not useless. I've put out job ads only to receive hundreds of applications from all over the world, many not even in English or without even a word accompanying a CV. Of the "proper" applicants, I've then had to deal with dozens of candidates that claimed a competence they did not have. And guess what - that's a lot of effort and time which is taken away from the business. In my case, that's a big opportunity cost with salaries burning a deeper hole for every minute that they are not offset by client time. I suspect it's much worse for famous organisations.

So now I ask a dozen questions by email that should take 5 minutes to answer in order to filter out both these groups, and that cover many basic CS concepts. Works really well, even if it's a bit annoying for candidates and probably turns off many vs just uploading your CV for HR to auto-scan for keywords.

For example, if I'm hiring someone to do data modelling work and they tell me it's called the relational model because of the relations between tables, it's unlikely they'll know how to unfold EAV antipatterns. And if you have actual machine learning experience, it's unlikely you'll be stumped by my asking what high leverage means.