That's the best kind of surprise! Congratulations.

The first program I ever wrote, back in primary school, was a game. The gameplay was rubbish, but hey. I had a few games under my belt by the time I went to university.

In my last year at uni I received an invitation to apply to a local games company - they went down to the Senate house, made a list of everyone who got a first or a 2:1, and sent them a letter. That wound up being my first real job.

It doesn't hurt to try to get into a university in a city which has a few games companies. Write games in your spare time. Go to industry conferences with your portfolio on your laptop and show it around.

Also play games, but do it with a critical eye: what does this game do well? What would I change?

It doesn't take that long, and it's the way that the formula you memorise was derived.

If you're needing to solve quadratic equations for a game, be aware that there are numerical issues with one half of -b +/- sqrt(b^2 - 4ac) when 4ac is small. The way to handle this is to have a case split on the sign of b and then to use the fact that the product of the roots is c/a.

There is only one way to get a useful answer to this: profile. I would start out using the standard library and only look at replacements if you see Math.sin or Math.cos actually showing up in your profiling reports.

Shift just about qualifies (it's about 5.3 years old), and Shift 3 and 4 definitely qualify (as does my Java4k demake).

On the exploration platformer front, I loved William and Sly (2009).

A pedant might say that you're not avoiding trig, just concealing it via de Moivre's theorem.

I would add feedback/criticism. It's very hard to look objectively at your own game, and even harder to see it from the perspective of someone who's playing it for the first time. If you can get some friends to try it and watch them play, that helps a lot; but it's also good to get comments (hopefully constructive) from people who understand game development.

So does debugging code you don't understand because you just copy-pasted it. And that's without getting into the legal issues.

Yes, the JIT will do that.

If you *really* want to over-optimise the fully generic case, you should look at a custom class loader which generates a balanced decision tree...

This is why I shouldn't write code after midnight. I've edited to correct, because the number of versions is getting confusing. Thanks.

Actually I stupidly lost the quantisation. The corrected version is

And of course there's always the option of offsetting things so that you have a half interval at each end and both true black and true white are possible.

Yes, because there's a bias up but no bias down. Actually it seems to be preferable to counter the bias up from the threshold with an intentional bias down so that on average they cancel:

(Corrected below)

That's not finding the nearest colour: it truncates rather than rounding.

As to the normalisation, the Wikipedia article says
So in your case with an 8x8 dither matrix you want to add bd * d / (8*8+1).

Putting that together,


and

(corrected below)

Or require MP3 and then you get data in the frequency domain already. Requires a bit more effort, perhaps, because decoders tend to be all-or-nothing.

In the simplest model, air resistance directly opposes direction of motion, so it affects the vertical component too. I once did some work on finding suitable air resistance parameters to fit known shell ranges and predict ranges for other shells for a historical warfare game, but I don't remember the values which I came up with.

For arrows the gyroscopic effect is likely to be quite significant in preventing the arrow from curving away from the straight line it would naturally follow. (Of course, it's never that simple: there's also precession to take into account...)

"The Mathematics of Projectiles in Sport" by Neville de Mestre might be worth getting hold of for people who are interested in realistic simulations.

PS HeroesGraveDev, stop digging.

In a vacuum. Arrows follow flatter paths.

Depends. In principle a file can be both a valid image and a valid jar. I don't know what their T&C say, though, or what processing they perform on uploaded images.

The obvious OpenGL binding for you to use with Scala would be... LWJGL. It's a JVM-based language which supports Java libraries natively.

If you want something faster than a full scan of the bounding box, it seems to me that the best way to do it is likely to be runlength-encoding the alpha channel as a pre-processing step. If your shapes are convex then you only need to store 3 numbers per row, and the overlap check would be two comparisons.

There is an art to constructive procrastination. If the washing up is accumulating in the sink, it's not so hard to tell yourself "I'll just fix this bug first..."

It's actually worse than Matheus says, because an array is an object and has an object header overhead. See http://www.javaspecialists.eu/archive/Issue078.html for an old approach; or experiment with java.lang.System's information on currently used memory.

Profile, profile, profile.

Basically your scores would have had less influence. The other two judges had almost identical means and standard deviations, but your scores were more tightly clustered towards the top end.

Edit: but most games wouldn't move in the positions at all; 7x7 would be the most drastically affected, moving down 5.5 places; no other game would move more than 2.5 places, but there would be one tie broken and two (one of them three-way) created in the top 11.

Not everyone is a Nintendo fanboy. There are enough of them that Nintendo remakes get a nostalgia bonus, but it's a good thing that the judges have different backgrounds: if they were all clones then there wouldn't be any point having more than one. And it's a good thing that the competitors have different backgrounds, because there were some great games on different platforms that get a second breath of life from someone who used to play them.

Appel commented that when he was judging Apo's games he made a conscious effort to judge each game on its own merits rather than against Apo's other games. Similarly, judges should make a conscious effort to judge games on their own merits rather than on the merits of the games which influenced them.

If you're interested in a real chess-Sudoku, check out http://www.toothycat.net/wiki/wiki.pl?PeterTaylor/ChessSudoku

On a separate note, for the past few years I've calculated the correlations between the judges and the community; interestingly, although this year the correlation between the judges is the lowest since I started doing this, the (Spearman's) correlation between the overall judges score and the community score is almost the same as last year, at 0.725.

1. I've had freezes on levels 1 and 2.
2. That's how it works already, isn't it? I'm sure I was using both mouse buttons to play it.

You forgot to take into account the Nintendo [dr]emake bonus.

My top 3 were Dord, Parasite Escape, Flywrench4k. I'm surprised to see Parasite Escape so far down. Half of my top 12 are in the community top 12.

Question needs more context. I assumed that the only person who would see the feedback was the game author, but it appears that teletubo and moogle assume that everyone will see it. Which is the case?

(Personally I don't mind either way, although I don't think that the individual point allocations should be public or you lose an important aspect of the distinction between community voting and judges).

That's good, because the logic's back to front as well.

Do you really need to find all entities within range, or would you be better using something like k-nearest, which is a well-discussed computational geometry problem?

I'm pretty sure there are better ways to compile without debug information in Eclipse.