After tons of optimizations I was able to get it up to 10% of the data I have. I’ve intentionally introduced a bias towards winning builds. Very slight for now, but it’s slightly improved this. Once complete I’ll offer an API for inference on both weapon and builds embeddings.

holy shit this worked way better than it had any reason to. This is only on 4% of the data I have and it's already surprisingly good. What the fuck lmao. This is weapon similarity via kit choice as per @/stat.ink data

X Battle Top 500 player activity. Note that the increase after 6.0.0 is due to the changes in the way XP is calculated after series, which shows the Top 500 is significantly more volatile ever since they implemented those changes. This is a bad thing for a ranking system.

Ever wonder who the strongest players using a weapon are and how often they used it to get their xp? Want to compare peaks? Wonder no more! I've released a leaderboard for this. It's late so I can't fix a bug that says "Service Unavailable" but it works! splat.top/top_weapons

More yap coming soon, courtesy of applying economics to a video game about squid kids

Here's a preview of the upcoming rework of splat.top! Ontop of everything you can see in SplatNet, I've added which region the player belonged to last season!

I made a video of the evolution of the Top 500 distribution vs X Power, each frame represents a different day. You can see how it evolves, it's kinda neat actually