HNHacker News
TopNewBestAskShowJobs

tvladeck

1,670 karma · joined March 13, 2012

managing director of gradientmetrics.com

we do statistical market research. mostly survey-based experiments. for everyone from small startups to the worlds biggest brands to NGOs and political campaigns.

submissionscomments
tvladeck··on The Remarkable Brain Waves of High Level Meditators [video]
Not the OP, but have a similar story. According to my goal-tracking apps, I meditate about 65% of days, and I average about 10 minutes a day. (So, not a ton, but pretty consistent.)

The most common way that meditation changes me is to help me identify that I'm in a rumination cycle and to break out of it, and to be more present in certain moments. For me, it's not that I have noticed a baseline change, but rather that it's a skill that's super-helpful in key moments. Some examples:

- When my mind is racing while trying to sleep

- In moments of anger or frustration when I'm ruminating about some slight or when someone has reneged on a committment

- Noticing that I'm distracted or anxious when out with friends or at a concert

tvladeck··on Sex Differences in Visual Motion Processing
That quote doesn't seem to be particularly political. It's just saying including sex is important. Even the implication you read into it is not particularly political.
tvladeck··on Why is flaking so widespread in San Francisco?
Indeed — I chuckled when I read the headline. Here in NYC it's exactly the same.
tvladeck··on The scooter company that raised 400M in 4 months is what is wrong with SV
You’re not the one investing at that valuation, so why do you care?
tvladeck··on Bird is the fastest startup ever to reach a $1B valuation
Well, without knowing the liquidation preferences it’s impossible to tell what the common stock is actually valued at.

Better to think of it as the price of a call option on equity at an (unknown to us) strike price (the liquidation pref). And then when you pretend that’s the price of common stock and extrapolate out you get a $B, or two.

tvladeck··on Ask HN: Any recommended resources to learn market analysis?
Without knowing any more about what you are trying to do, I would just recommend a framework we learned at business school that I keep coming back to. It's for when you're looking at developing or altering a product. It's really just a checklist, and it's super simple, but it's effective. Can be tweaked to understand how different brands are attacking a market.

5Cs, STP, 4Ps

5Cs: Just list out a bunch of facts to get acquainted - Context: what are the wider trends (e.g. millennial habits, etc.). This is where you start wide

- Customer: what 'job' is your customer trying to do? Who could that be? What is your unit of analysis: a person, an occasion, a ___ ??

- Company: what are you good at?

- Competitors: Who else is trying to serve that job?

- Collaborators: Who could be a partner? (Vendor, complementary service, channel partner, etc.)

STP:

- Segmentation: What are the different customer segments? What are the dimensions that make two {people, occasions, etc.} different in a meaningful way

- Targeting: Which are the viable and nonviable segments? Who's your target?

- Positioning: What is your:

-- POP/POD: point of parity/point of difference

-- Frame of reference: who are you stealing share from?

-- Reasons to believe: why would someone believe that you can deliver?

4Ps/marketing mix:

- Product/brand: what is your product? what's it's functionality? what's your brand?

- Price: level and structure.

- Place: distribution channels. Pull vs. push.

- Promotion: where will you advertise? What's your message?

tvladeck··on Flow chart of cognitive biases
I have been trying to "export" but can't quite get the sizing/location right.
tvladeck··on The Fall of a Star Trader
My understanding is that the EURIBOR rate is an index that attempts to track the overnight interest rate that banks lend to each other at. These numbers are not public (the trades are made "over the counter"), so the body that sets the rate relies on information given to them by the banks. To calculate it, they do something like taking the median of all submissions.

At the same time, there is a lot of money riding on what the EURIBOR is at any time. The allegation is that this guy (and others) conspired to change their submissions to change the listed EURIBOR and make a lot of money.

I'm no expert though, and I could be wrong in my understanding.

tvladeck··on Why Is It Taking 20 Minutes to Mine This Bitcoin Block?
ahhh you're right
tvladeck··on Why Is It Taking 20 Minutes to Mine This Bitcoin Block?
> Correct, that is exactly what I am saying. If you pick a random point in time, you expect 20 minutes between the previous block and the next block on average.

I thought this sounded funny, and I did a little simulation to see if it was correct. Given his assumptions (poisson with lambda 10), you do not get that answer. I got right around 10, which is what I would expect.

https://gist.github.com/tvladeck/e7a164dfe70fa765b10c1af64b3...

tvladeck··on Go 2017 Survey Results
Update: I asked the author!

The reply was discouraging in one sense and encouraging in another:

https://twitter.com/spf13/status/968259340619698176

"Custom Go application that parses the raw XML data from Qualtrics and the generates the SVGs"

tvladeck··on Go 2017 Survey Results
Tangential question: what was used to produce those charts? I love the simple formatting and how the data is laid out.

My company* uses ggplot exclusively for reports that don't look nearly this well done. I love the information density, love the use of colors, love everything about them!

Based on my inspection of the source, these are "native html" charts in the sense that it's a mix of text and svg rect elements — not an "image" that is included.

Does anyone know how these are produced? I would love to shamelessly imitate if possible.

*tiny market research firm

tvladeck··on Ask HN: How do you manage payroll and compliance for foreign remote employees?
> What are the problems in employing people as contractors rather than employees?

No problems for me, but there would be huge problems for the new employee. In the Netherlands, it's illegal, as the contractor has to have >= 3 clients. It also causes problems with access to social services and healthcare.

It looks like, for the Netherlands at least, that there are some local providers that handle this. One I found (and will probably go with) is Orange Tax — www.orangetax.com

tvladeck··on Ask HN: How do you manage payroll and compliance for foreign remote employees?
Unfortunately while they do international _payments_ they don't seem to do international _payroll_ and _compliance_.

For other reasons, looks interesting. Thanks for the recommenation!

tvladeck··on NYC Subway travel time map
Umm... wow! I did not know this, which shows my level of knowledge in this field (0). But this is amazing! Thank you!
tvladeck··on NYC Subway travel time map
This is awesome. Sort of similar idea, different approach, definitely doesn't look as cool, but my team and I published this today:

http://blog.gradientmetrics.com/2017/12/07/new-york-city-in-...

tvladeck··on New York City in Timespace
Hello! I am the author. This was a super fun project and something I've been randomly obsessed with for a while. Any Qs and I'd be happy to answer!
tvladeck··on The often-overlooked random forest kernel
> 1) Random forests are by nature supervised learners

You can run unsupervised random forests. The way this procedure works is to first make a "copy" of the original dataset (the reason for the quotes will be evident in a second), and then train the forest to distinguish the original data from the copy with a binary target vector.

The "copy" is made by sampling column-wise from the original dataset, so that, in the copy, each variable has the same univariate distribution as the original, but any interdependencies are destroyed.

The forest then learns to distinguish the original from the copy, and the only information that it has at its disposal are interdependencies in the original dataset.

tvladeck··on Tesla Passes Ford by Market Value
Not a very meaningful metric. Ford is still 3x Tesla in terms of Enterprise Value, which accounts for how much of the cap table is debt.
tvladeck··on Why your startup should be a Delaware C-Corp, not an LLC
yes this is commonly done in PE but for whatever reason it's not done in VC. usually they say it's b/c of compensation via options but i don't actually believe there is a principled reason behind it.
tvladeck··on Why your startup should be a Delaware C-Corp, not an LLC
The reason is that many of their LPs (e.g. pension funds) are non profits, and they can't have taxable income flow up to them or their Unrelated Business Taxable Income will threaten their nonprofit status. VCs are flow-thru entities so any income hitting them from _their_ investments would hit their LPs. Therefore they can only invest in blocking entities.
tvladeck··on Why your startup should be a Delaware C-Corp, not an LLC
This is not true. It's very easy to convert from LLC to C corp but not vice versa. As a general rule it's easy to go from flow-thru entities to tax paying entities but not vice versa. This asymmetry of irreversibility, combined with the fact that most startup exits are asset, not stock, sales, makes the thesis of your post incorrect.
tvladeck··on Starting a real business
Both of the points you brought up are not specific to c corps. Any company will list salaries as an expense on the P&L, and an LLC also provides a layer of protection between "business and business owner(s)". The first two Ls stand for Limited Liability...
tvladeck··on Starting a real business
If you're just starting out, the worst decision you can make is to incorporate as a C-corp. An LLC makes vastly more sense, even for the tiny proportion of companies that want VC funding (and specifically institutional funding - funding from other sources would not matter as much - I'll explain why below). Stripe Atlas really ought to make the default an LLC.

The two major reasons why:

1. When you're a C-corp, you pay taxes twice. Once at the corporate level, and once at the individual level. This matters both for ongoing income but also for any liquidation event - as most liquidation events are asset, not stock, sales, you're getting taxed twice here too.

2. You can go from LLC -> C-corp easily but not the reverse. Why would you make the decision before you have to? Start as an LLC. In the very unlikely event you are taking institutional funding you can convert; in most cases, you'll happily stay as an LLC and keep the extra tax dollars you'd be giving the government.

Finally - all this business that VCs prefer to invest in Delaware C corps because of the legal knowledge there is - sorry - bullshit. They do it because they have to invest in taxed entities, because they themselves are partnerships so their interest in any flow-through entity will flow up to their investors, some of which are non-profits. Non-profits, like pensions, risk losing their non-profit status if they have unrelated business taxable income. There is a solution here, which is to have a special purpose blocker corp that sits in between the VC partnership and the LLC. This is done all the time in private equity but not in VC and there is no principled reason why not.

So there you have it. Don't do Stripe Atlas because you're forcing yourself to make a decision you don't need to make right now, that's irreversible, and that may end up costing you a lot of money.

IANAL, but I am a Wharton MBA

tvladeck··on Data from over 100 failed ceasefires shows most are followed by offensives
This is false (or at least a poor interpretation of the data). Ceasefires make it more likely that future ceasefires will succeed. Steven Pinker and others have looked into this extensively.

Here is a write-up of the opposing take.

http://www.nytimes.com/2016/09/16/world/middleeast/another-c...

tvladeck··on Computers do better than pathologists in predicting lung cancer type, severity
The article mentioned they only had a dataset of 2,186 slides. Isn't this a _tiny_ dataset to do any learning on images? They also mentioned they blew up the predictors to over 10,000 features. This sounds like a recipe for overfitting to me. I can't access the actual paper (getting a 500 error), but does anyone know how they built these algorithms?
tvladeck··on Growing Unease as British Mutual Funds Block the Exit Doors
I was more speaking to the origin of the term (as in, a society of people pooling their money to build buildings); not how it's used today. You may very well be correct and that would make sense.
tvladeck··on Growing Unease as British Mutual Funds Block the Exit Doors
You're confusing different definitions of the "money".

In theory, they retain enough _assets_ to pay off their creditors, but not enough _cash_. In the event of a bank run they'd have to be handing over their stake in mortgages and other debt that hasn't yet been paid off. But of course these are illiquid, and the creditors wouldn't accept that in general.

In the vocabulary of money supply, you're confusing M0 and M1.

tvladeck··on Growing Unease as British Mutual Funds Block the Exit Doors
This seems to me very analogous to preventing a run on banks. Traditionally retail banks would invest their depositors' savings in real estate by financing mortgages. Indeed - in England many banks are still called "building societies".

Depositors are in principle able to get their money back at any time, but this is possible only because a small fraction will want to at any one time.

But when there's a "run on the banks" in a financial crisis, banks can't satisfy the withdrawal demands of every depositor simultaneously - except through distressed sales that only exacerbate the problem. Governments sometimes step in to call a "bank holiday" until the panic subsides.

The situation, down to the main class of assets held (real estate) and the liquidity promised to creditors, seems to be pretty much exactly the same as at the retail level. That being the case, this seems to be the prudent move.

tvladeck··on Jane, John, Leslie? A Historical Method for Algorithmic Gender Prediction
I'm very good friends with one of the authors (Cameron Blevins) and some context here is helpful: He is an expert at applying computational tools to historical records. For example, his thesis at Stanford is about using information about the expansion of the postal network to map the westward expansion in the US. For him, the interest here is mapping names to gender conditional on a specific era, not about doing so today when you have the option of asking the person.
← PreviousPage 2 of 8Next →