All this said, we’re working with small enough numbers here that just a handful of entries can shift the average pretty drastically, so grain of salt, etc.
2024’s median game is, hilariously, ROD MCSCHLONG GETS PUNCHED IN THE DONG, at just over 1MB.
The largest game is The Quest for the Teacup of Minor Sentimental Value, which single-handedly accounts for over a quarter of the total file size. (I chose 2024 as an example both because it has the highest average and because I knew QftToMSV would be an outlier; note that I enjoyed it and this isn’t a criticism!)
The smallest file is a 1KB redirect to where A Death in Hyperspace is hosted on the primary author’s website.
Only ten games are above the average size, and a significant majority are in line with pre-2019 sizes.
(My uncompressed download of the 2024 archive is actually slightly smaller than the dottore’s, 1.29GB with 1.28 of it being game files, so YMMV.)
2025 is more difficult because all games in the big ZIP file have their own zipped folder, even if they’re single files, and compression ratios aren’t necessarily consistent. Cart is the median when sorting these zipped games, and its full size is under 1MB.
Once again, a handful of 100MB+ games at the top with only 20% of games above the average, a handful of redirect files at the bottom, and a majority of games at pre-2019 levels, many of which wouldn’t even be out of place pre-2010.
It’s almost like from any given year, someone could, in bad faith, cherrypick examples that suggest files sizes aren’t growing, or are even shrinking, while a clear trend for growing file sizes across the board remains self-evident.
It’s worth noting that the zip file for QftToMSV actually contains three versions of the game: the HTML5 one to be played in-browser, plus Linux and Windows versions for those who download it. I wanted to include a Mac version as well, but that would have taken it above the maximum allowed for IFComp. It’s inevitably going to be an outlier regardless, but the size of the submission is vastly larger than what’s actually necessary for the game to run.
Might this be a case where geometric mean is more insightful than arithmetic mean? I’m no statistician, but I understand arithmetic means can become heavily skewed when the data spans several orders of magnitude while geometric means are lessprone to such.
Though, a table showing entries per year grouped by size magnitude(e.g. <10 kb, 10-100kb, 100kb-1mb, 1mb-10 mb, 10mb-100mb) might provide a summation appropriate to the data in question.
Of course, there’s also the question of how to properly compare games running under different paradigms. Naturally, a story file that relies on an external interpreter is probably going to be much smaller than a game that provides a native executable for a modern OS… and for games that ship multiple, platform-specific versions, which version do you take for its filesize?
I’m going to guess number and size of images is going to be driving some of the larger game sizes. The size of games I was putting online years ago were smaller than now because image size and quality have increased. Phone data allowance and internet download speeds used to seriously restrict what you’d want to add to a game a lot more than it does now. (It’s still a consideration, but you no longer necessarily have to use tiny jpgs to not get complaints.)
Probably not as large a consideration for text only games (including many parsers), but for games using systems like twine with images or music, it’d be a bigger one.
