Hacker News Reader

Comment by _alastair | Hacker News Reader

I'm one of the people that made this interactive - funny to see it on Hacker News! If anyone has any questions about it, let me know.

View on HN

Very interesting to see the density of others' guesses. It doesn't seem to show every other guess though — how do you choose which ones to show?

Correct, it doesn't render all clicks (I dread to think how browsers would react if I tried) - all the results are put through a k-means cluster analysis at regular intervals to produce approximately 30 visible results. The percentages are calculated relative to all the clicks, though.

Nifty. So do each of the guesses shown account for a roughly equal number of clustered clicks, or do some of them aggregate more clicks than others?

I hate to throw to Wikipedia but I won't be able to do a better description than they do: https://en.wikipedia.org/wiki/K-means_clustering

No worries thanks, it's been a while since I've read the details about it.

Do you know if these kinds of interactive content perform better (ie clicks/shares/views) vs other forms, ie longform or just plain data visualization pieces? I personally enjoyed it a lot.

This isn't really my area of expertise - we have people at the company dedicated to studying this kind of thing in far more detail than I know how, who may end up reading this - hi! - so I don't want to speak out of turn.

But in my experience there's a lot of variation - things like this interactive aren't as tied to the news cycle as many articles are, so it won't peak as high but may make up for it in longer term traffic.

How long does it take to "hide" the ball per pic?

One possible way of generating such images is to use more images since a lot of sport photos seem to be taken as part of a burst sequence. All you need is two such photos where the ball has moved over enough and the rest shouldn't be difficult.

But won't the players have moved roughly as much as the ball from photo to photo (since they're moving at about the same speed)? I suppose as long as the players haven't moved into the space where the ball used to be you could still trivially use that space to replace the ball in an earlier image.

This is a guess but maybe if they have multiple moving objects they just figure out which one is shaped like a ball?

First, let's assume the ball is shaped like a sphere...

What is its radius in meters? Can I assume it is a unit sphere?

What am I doing wrong? Clicking and dragging the "drag this" has no effect. Nothing moves or changes in Chrome 43.

Oh, damn. I think you've found a bug, sorry about that. I think it's expecting a touch event when it shouldn't be. Let me take a look at that.

EDIT: should be fixed now. Apologies again, embarrassing bug caused by a last minute bug fix. The way these things always go...

Yep. Chrome on a Surface Pro 3. Doesn't react to click events, only worked if I actually touched my screen.

Same here.

Try with the stylus, I could be wrong but last time i played with a Surface Pro, I noticed that using your fingers on the touch screen game touchEvents and using the stylus gave mouseEvents.

Re-opened to test in Chrome on Ubuntu (XPS13dev) and to my astonishment, touch events worked!

Same problem: worked in IE (!) though.

How do you construct the background image that replaces the ball?

Suspect they used Photoshop Content Aware Fill. http://www.photoshopessentials.com/photo-editing/content-awa...

That's the work of Sam Manchester, deputy editor on the Sports desk and chief Photoshop wizard. I believe most of it is just cloning different parts of the photo to cover up the ball, though it can get more complex. For example, on the 4th photo of this previous round:

http://projects.nytimes.com/interactive/sports/worldcup/spot...

he actually cloned a players face from a different photo and pasted it in to cover up the ball. I have no idea how he does it so well.

I was wondering how he did that one!!! Part of my thinking was that you'd choose pictures that had the ball in an easy to photoshop location, and that threw me off

I think he enjoys a challenge. Case in point, the last photo in this set:

http://projects.nytimes.com/interactive/sports/worldcup/spot...

Maybe there's a way I just can't find, but it would be handy to be able to toggle everything 'off' again to get back to the original ball-less image.

Sports photographers also take many shots in rapid succession. I bet you could clone the background from a shot a half second before the one in the article.

If you look at the third picture, the guy in green is cloned and blended in where the ball is meant to be.

Are you going to share analytical results?

Like, does countries who adore football, score higher.

Do IPs who regularly read NY Times sports section score better?

Or people who actually seen the game on TV, do they score better?

Very cool, even for a non sporty person.

Interactive, fun and not too complicated. Really liked this idea

Would love to find out more about how it was made. Can you point to a small example of how to accomplish this? seems so cool!

How do you calculate accuracy? Sometimes I clicked in almost the opposite corner of the ball and it said I did better than 80% of readers.

Are you just calculating how close my guess is to the most densely clicked spot even if the ball is not actually in that spot?

The accuracy is calculated compared to other readers, so if you all clicked on the wrong corner you can still be better than 80% of them!

I mostly did it this way because there's no hard number that makes sense here - we don't know ft/metres, and pixels aren't a unit everyone is used to thinking about.

How did Maureen Dowd do on it?

I'm trying to come up with a good joke for this, Dan, and I'm failing. So I will tell you that last year Nicholas Kristof scored a respectable 54%:

https://www.facebook.com/nytimes/posts/10150426093994999

I wanted to ask how long it took to make this, but that's an impossible thing to answer - so how much lead time did you (and the team) have before the first version went live?

I just checked my e-mail - it looks like we decided that we were definitely going to do it approximately two weeks before the first round went live. That's not typical but not necessarily unusual, if that's a sentence that even makes sense.

Atypical in which direction? Is that more or less time than you'd usually have?

Makes sense to me! (I would interpret it to mean 20-30% of projects have <3wk lead time, fwiw)

What was the concept behind this, and what is interesting about it?

Thanks for the question, pervycreeper!

The concept is not new at all - Spot the Ball is a competition that ran in UK (and possible other) newspapers going back at least as far as the 70s. It was a cash prize competition and was pretty popular, though it's died out in recent years.

I wanted to bring it back to get people to interact a little more with a highlights photo gallery - it's a lot more fun that way. IMO, it's interesting because it's just the right level of infuriating.

Interesting, thanks. Definitely a tradition which is very well suited to revamping for the digital age.

Here's a mention of it in the 40s in VS Naipaul's autobiographical fiction.

https://books.google.com/books?id=A2bTQbKZRf0C&lpg=PA110&ots...

Early 1960s: Grandad tried the intersecting eyelines approach, Dad did the grid of guesses approach (you got a number of guesses). Neither won much!

One thing that people don't realise about newspaper spot the ball competitions is that the winning position was not where the ball originally was in the photo, but where the competition organisers thought it should be.

That removed any actual element of skill ("where are the players looking?") and turned it into pure guesswork.

People could buy rubber stamps of a grid of crosses so they could make very many simultaneous guesses.

MLSSoccer.com did this last year during the playoffs.

Yep - after we did it for the World Cup last year! Great to see more people taking it up.

Reply on news.ycombinator.com