Hacker News Reader

Comment by Ansil849 | Hacker News Reader

I was extremely confused about the distance values in relation to the percentages until I read the help:

> The distances displayed correspond to the distances between the selected and the target territory centers.

> For instance, the computed distance between United States and Canada is around 2260km even if they have a common border.

That's extremely deceptive and non-obvious unless you read through the documentation. One of the appeals of Wordle is that it is an extremely intuitive game which doesn't require rule reading.

View on HN

If you get a ‘country that’s 2000 km north of the USA’ and do not guess Canada (the ‘only’ country north of the US) then I don’t know what to say.

And you could presumably say the same thing about any other two countries, yes?

No?

Why not?

So what is the percentage indicate separately from the distance?

I think the "right" way to display the distance would be from the closest points in the guessed and target territories. So United States and Canada, sharing a border, would be "0 km", while Canada and Mexico would be whatever the shortest path from border to border would be (like 2,000km?)

I have a feeling that's a bit more difficult than centers, though. But it wouldn't be a problem as far as outlying territories go; the image is of a contiguous landmass, not the entire country.

But then you get Canada and you instantly know it's the USA next.

> But then you get Canada and you instantly know it's the USA next.

So in your hypothetical scenario....this hypothetical player would know that the only country Canada has a land border with is the USA, but not know what the USA looks like so as to be able to solve this without first guessing Canada?

This seems improbable to me. Chances are that if someone knows enough about a country that it's a 'single neighbor' country, or that it has a 'single neighbor' country, they will also know what the country looks like.

For example, if someone instantly knows that if they typed in Haiti it must then be the Dominican Republic, chances are very high that they wouldn't have typed Haiti because they also know what the Dominican Republic looks like. Much like chances are high that if someone knows that Canada only borders the USA, they also wouldn't have typed Canada in the first place, but would have guessed USA right away.

It's really a shame the developer didn't add this feature. One might be scared off by the large number of geometry files involved with this, but actually this is a great example of when hard coding can supersede dynamic/robust code. It would be better to simply store an object of all countries borders and the their relative direction to one another. I could see this being a slight issue with countries that share a border with another that is many directions of their border (Lesotho and South Africa, for example), but it would still be much less deceptive than saying "the USA is 2000km south of Canada."

France would be fascinating

(For context France has borders with Brazil for example, and is only a few miles from Canada and not far from Austrailia)

Yes, the fangs of colonialism bite deep.

Interesting because, between easy-and-wrong and hard-but-perfect, there is the solution of subtracting the size of each country from their distance.

Assuming by "size" you mean something like average/longest distance from the center this is the worst possible answer because it's super non-intuitive and can create crazy results like negative answers.

> I have a feeling that's a bit more difficult than centers, though.

This is quite an understatement.

Given it's a static dataset of countries, the distance of all country pairs could all be precomputed and loaded in memory, I guess ¯\_(ツ)_/¯

True. But to perform that precomputation, you have to find a database of country geometry outlines. And (manually?) filter out things like Hawaii and Guam from these. Then, assuming country geometry is given as lat lon polygons, it seems you then are computing either O(n^2 * m^2) distances where n is the number of countries and m is the average number of vertices per country. Or you might come up with some space partitioning algorithm on the vertices to reduce that to maybe O(n^2 * m log(m)). It's doable, but my comment was merely pointing out that it's much more complex than comparing country centers.

Oh, good point. Getting an SVG of the world and just brute forcing every country would be near instant.

Hmm. What issue might you encounter trying to solve this problem using an SVG?

What is the center of the US anyways? Does it include Alaska and Guam?

Great question! Is it the center of a convex hull containing the extent of territories? Or would you weight it by population? Is it in two dimensions on the surface of the Earth, or in three dimensions?

Consider the possible “center” of France. There’s probably some village in France proper with a monument, but unless you eject Guiana and Réunion I think it must be in the Atlantic somewhere.

Territories (eg French Guiana) are specifically separated in Worldle, even being possible answers themselves.

Also, this is a map game, so for me it's pretty intuitive that it is the center of the hull. Maybe with some error due to projection (not sure how they implemented the map), but should not be a problem for solving the puzzle.

> Also, this is a map game, so for me it's pretty intuitive that it is the center of the hull.

I don't follow. How does this being a map game make it intuitive that the distance outputs are based on hull centers? To me, closest border would seem be the most intuitive reading of the distance...this being a map game.

If you poll 10 random people, I guarantee 10/10 would say that they consider borders to be the intuitive means of measurement of distance between two countries, not their (hull) centers.

Please do

For the arrow hint to work at all, countries need to be a single point. I intuitively assumed the hull centre without really thinking about it.

> For the arrow hint to work at all, countries need to be a single point.

First of all; no, not at all. The arrow hint can work if countries are considered a mass, not a point, it simply means that there can be times when there are any number of correct arrow designations.

Second of all, even if countries are considered as single points for the arrow hint, that is not the issue being presently discussed: that of the distance hint, which is a separate hint from the arrow hint. Countries can be considered as single points for the arrow hint, and as masses for the distance hint.

I don’t know what you mean by “considered a mass” unless you mean “centre of mass”, which is a point. If there are times where there are a lot of correct arrow designations, how is that not more confusing?

In my opinion, using a different measure for the distance and direction would have been even more confusing. It’s also a lot harder to program for a silly web game.

My reading of it is that it's easiest to program if they just ask google maps the latlon of the country and use that. It's a lot more work to base it off the closest border.

Here’s my n=1, I guessed it was “center” as opposed to borders, but I feel like this might be influenced by what country you play the game with first.

Interesting, since my understanding of the matter is that the status of Guiana within France is as integral with that nation as Alaska is with the U.S.

Yes it is indeed an integral part of France — a full department as mentioned in a sibling comment.

(This is all news to me. I somehow assumed that French Guiana and British Virgin Islands were independent countries.)

I think French Guiana is not a territory, it's a full department in France. Equal to Hawaii in the US as far as I know, but I'm not 100% sure.

That’s my understanding too. Something about Napoleon wanting to be egalitarian. Has the interesting side-effect that France is both the western-most and eastern-most country in the EU.

About egalitarian Napoleon:

> In 1801, Louverture issued a constitution for Saint-Domingue [now Haiti] that decreed he would be governor-for-life and called for black autonomy and a sovereign black state. In response, Napoleon Bonaparte dispatched a large expeditionary force of French soldiers and warships to the island, led by Bonaparte's brother-in-law Charles Leclerc, to restore French rule. They were under secret instructions to restore slavery, at least in the formerly Spanish-held part of the island.

https://en.wikipedia.org/wiki/Haitian_Revolution#Napoleon_in...

As for French Guiana, it was actually an extreme penal colony at that time (also, it was briefly held by Portugal). It became a department in 1946.

The centroid, surely.

Safe to assume middle of the lower 48

That’s not generalizable to every country though.

Rule could be it’s the center of the geography they’re showing you. If they actually put Guam on the USA map then it’s included. If not, it’s not.

What about France? Is it half way into Spain because its weighted towards French Guyana?

What about Reunion? Then it should be roughly in the middle of Africe.

If you weight it by the actual size of territories rather than just the extreme boundaries then the center is indeed in Spain, because the overseas regions are still rather small in comparison to the European part of France.

If you use only the extreme boundaries then it gets a lot more interesting and I would say that the center is closer to the center of the Earth than to its surface.

Things probably change if you weigh by surface including the ocean EEZ, with French Polynesia being the most significant area of France at this point.

Fun fact, Google Earth had its coordinates defaulted to an apartment complex in Lawrence, Kansas, apparently the childhood home of one of the engineers, but also roughly the center of the United States.

https://www2.ljworld.com/news/2005/dec/21/lawrence_pinpointe...

I thought it was a farm in Kansas. They frequently get visitors looking for people who's geo ip data said they lived there.

A good measure would be to take the median distance between any two points in the two countries.

Median is great for removing outliers.

To be fair, Wordle also doesn't require much external knowledge beyond a basic vocabulary. This is aimed at a more narrow crowd and is just a fun experiment.

I think Wordle requires a pretty extensive vocabulary to be played successfully. Though you could say the same thing about Worldle: it doesn't require much external knowledge beyond basic geography.

For Wordle, a decent vocabulary and also some intuition about letter frequency, the likelihood of letter combinations in various positions, etc. For this "basic geography" I guess but that would include knowing the names and approximate locations of some fairly obscure and not distinctive looking countries. (That said, if I cared enough to do so, I could probably bone up on my geography for a couple hours and I could probably do better.

Worldle also doesn't require rule reading. I've been playing it for a week and never read any documentation. It's quite intuitive if you're not interested in nitpicking exactly what the distance measurement means.

It sort of does to the degree you need to know (for some words) how things work if a letter appears more than once.

Yeah, I agree that’s pretty confusing. That being said I had to read the Wordle rules the first time I played. Had no idea what the differences were between colors and that a letter could be used multiple times in the same word.

Perhaps they should have picked:

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

Wouldn't that make countries like the USA and Canada extremely far apart because they are both large countries? The distance you would get is the distance from Hawaii to somewhere on the northeastern edge of Nunavut/Newfoundland/Labrador

Not quite, you'd get the maximum distance from the extreme end of one territory to the nearest point of the other.

> Not quite, you'd get the maximum distance from the extreme end of one territory to the nearest point of the other.

But which country do you pick the extreme edge of, and which country do you pick the nearest point of? For example, if measuring US<->Canada, do you pick the extreme end of Canada or the extreme end of the US?

You pick whatever gives you the largest distance.

For every point in both territories, find the closest point in the other territory. From these pairs of points, find the pair with the largest distance. That distance is your Hausdorff distance.

Oh I see, so still pretty large since both countries are large, but if one of the bordering countries were small the distance would be small.

No, you would get the distance from the border to the other end of the large country, since that is longer.

I think the percentage could be relative size of countries instead. Still a hint and kinda intuitive.

How is it deceptive? To me it's an extra tidbit of information that doesn't seem important. It's clearly linked with distance, so I know it's something to do with distance.

> How is it deceptive?

When people talk about how far one country is from another, chances are extremely unlikely they are referring to how far their central points are from one another.

That's true, but it takes a moment to realize the intuitive first thought couldn't work in practice. So it must be something like the centers. Plus the exact distance doesn't matter in this game.

they could have simply used distance between geometries like st_distance in postgres. curious about edge cases here where for example one coordinate is 178 longitude and another -178 longitude. also does this game show whole countries which have territories over great distances like france denmark or norway

Reply on news.ycombinator.com