Spot the Ball

http://projects.nytimes.com/interactive/sports/worldcup/spot-the-ball/2015/06/19

by mrbird • 11 years ago

282 82 11 years ago

projects.nytimes.com

_alastair13• 11 years ago

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.

corbinpage• 11 years ago

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

rwc• 11 years ago

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

_alastair13• 11 years ago

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.

BillTheCat• 11 years ago

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.

edgall• 11 years ago

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

ehmorris• 11 years ago

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?

_alastair13• 11 years ago

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.

dangrossman• 11 years ago

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

_alastair13• 11 years ago

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...

basseq• 11 years ago

Same problem: worked in IE (!) though.

h1fra• 11 years ago

Very cool, even for a non sporty person.

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

ak39• 11 years ago

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

vlasev• 11 years ago

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.

Swizec• 11 years ago

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?

_alastair13• 11 years ago

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.

wooyi• 11 years ago

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.

_alastair13• 11 years ago

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.

danso2• 11 years ago

How did Maureen Dowd do on it?

_alastair13• 11 years ago

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

pervycreeper2• 11 years ago

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

_alastair13• 11 years ago

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.

rattray2• 11 years ago

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?

_alastair13• 11 years ago

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.

dzhiurgis• 11 years ago

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?

Omnipresent• 11 years ago

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!

jmhobbs• 11 years ago

This is infuriating. I tried using where players were looking and it got me squat. Very cool though.

jpreiland• 11 years ago

Yeah I found that the players' eyes were mostly useless but often times the composition of the photo is helpful. For example, if players are gathered on the right edge of the frame and there's a bunch of blank space to the left, the ball is likely over in the open area

ars• 11 years ago

I used where the fans were looking instead. It worked sometimes.

I assume the players look where they are going, and at other people instead of the ball (since they can predict the trajectory they probably look where it will interact with something, not where it is.).

bherms2• 11 years ago

reminds me of bestofthebest's car raffle... since you can't really gamble in the UK, they do a skill based game that is exactly this... you buy guesses and then click where you think the ball is... the person each week with the closest guess to where a group of judges says the ball is wins the car of their choice

knodi123• 11 years ago

a group of judges? but... isn't there one objective correct answer?

thaumasiotes• 11 years ago

What fun is that?

bherms2• 11 years ago

Yeah, I'm confused as to why they need judges as well, but it's how they do it...

phaemon• 11 years ago

since you can't really gamble in the UK

Of course you can really gamble in the UK - that's what the bookies are for.

However, the regulations are a lot tighter for gambling, so it's easier to make it a "skill based game" which has much looser rules. Bingo and raffles have their own rules which, again, are a lot easier to comply with.

Zenst• 11 years ago

Indeed I recall Spot the Ball compititions in the national papers in my youth and nicely covered here https://en.wikipedia.org/wiki/Spot_the_ball in much better detail than I. So somewhat nostalgic seeing the new edition. Though recall it's demise in newspapers was down to the worry of technology enabling analysis of the picture reducing the human level of skill involved and with it being money related prizes then they moved on.

SmellyGeekBoy2• 11 years ago

Of course we have gambling in the UK. It's just that skill-based games are less tightly regulated than purely chance-based games. Hence those premium rate TV phone-in competitions where the player has a choice between one obvious answer and two stupid ones.

8_hours_ago• 11 years ago

John Graham-Cumming wrote an article about trying to hack a contest similar to this. I wonder if a similar technique could be used to find the ball on these images?

http://blog.jgc.org/2008/02/tonight-im-going-to-write-myself...

danso2• 11 years ago

I started to roll my eyes when I saw your comment because I thought, "the NYT made a fun web interactive with no-stakes and someone is pointlessly wondering if the answers are in the source code"...so now I realize what a reflection this is on how lame and limited my own thinking is :).

This was a great link...I've been casually brainstorming heuristics for detecting possible Photoshopping and John writes about exactly that. I'm going to walk through his algorithm using Python and PIL.

smackfu• 11 years ago

Neat idea, but is this something you can actually be skilled at? Even if you know the game, and can figure out what trajectory the ball was on, it seems very hard to figure out exactly where on that trajectory it was when the photo was taken, given the high speed of the ball.

emodendroket3• 11 years ago

Couldn't you? It seems like if you were skilled enough you could figure it out by the reactions of various players in the photos with more than one or two people in them.

jdrock• 11 years ago

Trying to triangulate based on where players' eyes are looking rarely works. In many cases, the goalie is looking in a completely different direction than where the ball is. I guess that's why they're getting scored on.

blazingfrog22• 11 years ago

After logging many thousand of hours of dutifully watching soccer (over 30 years of full-on love for the game), I was able to match the given situations to similar ones I'd seen before and get close enough most of the time. It's still all probabilities at the end of the day.

Following the direction players are looking is just one parameter and can be wildly inaccurate without any information on the angle and velocity of the ball at the time of the shot. One of the trickiest situations is when 2 players jump for the ball, usually coming from a goalkeeper's clearance, so high and fairly fast, and collide in the air (often closing their eyes): the ball could literally be anywhere.

_alastair13• 11 years ago

...sort of?

That's kind of the genius of puzzles like these (we can't take credit for inventing it - Spot the Ball has existed for a long time in the analog world) is that it's a mix of skill and luck, so you can fool yourself into thinking you're amazing at it, when in reality the next photo might trip you up completely.

biot• 11 years ago

I can relate. I mostly ended up around 55% overall per round (only slightly above average) but I did get a 100% score on one photo which made up for some spectacular fails.

anigbrowl• 11 years ago

As I have mostly scores in the vicinity of 95-99%, yes indeed you can. In the UK and I think many other countries where soccer is popular, 'spot the ball' games have been a popular newspaper and prize competition entertainment for many years, probably long than I've been alive. It also helps to have played a lot as a kid and have a basic intuition of how the game is played.

following the eyeline of players (and sometimes background observers) is the most basic clue, but it's also important to remember that in soccer you're not allowed to touch the ball with your arms or hands if you can possibly avoid it, so when you see a player's stiffened arm you can often guess that they're fighting to avoid an accidental contact that would result in play being interrupted by the referee.

It's not a very useful skill but I hope that soccer gets much more popular in the US as I might be able to win some easy prize money if spot-the-ball competitions become popular. Or for a small fee, I will be happy to provide individual tuition in this potentially lucrative derivative sport :)

hackuser• 11 years ago

It's odd to see a casual game, with no other significance or meaning, in the NY Times. It would be great to see them using interactive tools to tell news stories more often.

emodendroket3• 11 years ago

They do that too. Like the story they did about the Philippine Coast Guard in the South China Sea.

Klinky• 11 years ago

Next they'll be putting crossword puzzles in the New York Times! Of all places!

tabrischen• 11 years ago

This is very fun for me, even as someone who's not a sports fan. Great way to engage this other segment of readers.

mholt• 11 years ago

Did anyone actually read the article?

DLWormwood• 11 years ago

I remember seeing an arcade redemption machine that worked on the same principle. It had a database of hundreds of screenshots from soccer/footy games and the player had to guess where the ball should be. The more accurate the guess; the more tickets the machine dispensed.

usaphp• 11 years ago

If you notice the outline of the ball has different size for each photo, this allows to identify a relative distance where the ball is compared to height of players

_alastair13• 11 years ago

Glad you noticed that! It's proof of the difference a great designer (in this case, Rumsey Taylor of our News Design team) can make.

As a developer it can be quite humbling to realise how many seemingly tiny yet actually really important touches exist in a project you think you know inside out.

_puk• 11 years ago

Good fun!

My understanding of the UK version of this was that the ball was actually placed by a group of pundits[0], rather than being in the original location, so even if you found a freeze frame of the original match you'd still not be able to cheat the system.

Would be fun to crowd source a position taken from wrong guesses to provide some variance.

0: http://www.theguardian.com/football/shortcuts/2015/jan/14/ho...

SmellyGeekBoy2• 11 years ago

I like that - a "most popular guess" vs "actual position" would be really interesting. As we seem to have at least one of the developers in this thread, maybe we can make that happen (or at least a post-mortem blog post analysing the data?)

namuol• 11 years ago

For me this quickly became "spot the photoshop artifact". I only noticed it clearly for one photo and did marginally better than average (58% -- some guesses were way off) Still very fun.

Rexxar• 11 years ago

The second time, after removing the cookie, is much more easy (~46% to 99.67%). The conclusion is that there are, for the moment, few cheaters.

ChrisArchitect• 11 years ago

as a footy fan I enjoy this a lot - tho this time around I scored terribly compared to world cup a year ago

thanks _alastair for getting involved here - tons of great little insights

Spot the Ball

Related Stories