I’m currently computing it for every single day since tracking started so I’ll let you know in a few hours

Higher values means it’s more likely it’s used by people high in the T500 compared to low. Blaster at +2 means it’s ~7.4x less likely than pure chance to be used by high skill players. So it’s used more successfully by the top 100 than ranks 400-500

Of the Top 500 in Tako TC, 60 of the spots are taken by a single weapon. No other weapon has such high representation

It’s funny how one of the major criticisms of AI use is outsourcing thinking when most of the applications of said criticism are outsourcing thinking to random Bluesky posts

With today’s update you’ll find loading splat.top profiles is a lot faster

Few more updates on splat.top
- Past Seasons now have a tab for their leaderboards, under Archive
- Skill Offset can now be split by mode, region, or mode+region
- Some minor bugfixes

pyproject.toml

Very cool little UX touch added by @hfcred.dev

You can now type w./u./t./o./to. to quickly change what you are searching for on the global search eliminating those unnecessarily clicks when you just have to go...

Register for In The Zone 51! Happening Saturday 25th April w/ $240 prize

#sendouink

It’d be really funny if in Harry Potter’s world, Isaac Newton was renowned for his alchemy and made fun of for his physics as a complement to the real world Newton

Ok Glocket is an insane stomping build at my ELO it’s basically freelo. Chunking late game carries for over a thousand per shot, facetanking and killing ultbot Haze at the 40 minute mark, dodging doorman ult entirely… why didn’t I pick this up earlier oh my god

It doesn’t actually look very large, it looks like noise!

LUTI Divs vs. sendou ink seeding power

How much difference there really is between Div 7 and Div 10? Otherwise I guess you see the hierarchy of it quite well.

Has there been a single pokemon game in the last few years that didn’t need a six page thesis explaining “why it’s good, actually”? It feels like they’re batting zero across the whole franchise

I’ll give it a look, I’m fixing a few bugs I didn’t catch before deploying that’s probably one.

Results update once a week because it takes five hours to complete the calculations :)

I’m not gonna do the top 500 across all past seasons mostly because each season is independent and cannot be compared with each other, but the previous seasons should be easy I already have and use the data

Like the final results? Yeah I could probably wire that up as well

Additionally, now there's player pages for comp.splat.top! What value does this add? Well, if you log in with discord, it'll link to your sendou.ink account and you'll be able to see your top ten most impactful wins and losses! Automatic results!

splat.top got a makeover! Profile pages are slicker and align with comp.splat.top a lot better. Additionally, a new page for Race to 5000 XP that tracks any player over 4000 XP and compares with successful climbs to 5000 XP!

TLDR high skill offset means high XP players use it disproportionately, negative skill offset means low XP players. Zero means expected

It’s essentially a skill-oriented usage pattern on how it appears on the Top 500. Given k appearances across all leaderboards, create a prob. distribution of the median XP for 4000 choose k players. Then compare against median XP for that weapon. Skill offset is the log prob difference.

I was surprised by RPD being so high myself, it’s just… six orders of magnitude less likely than baseline?? Really?

This brings the row count up to 87 million rows. Getting closer to 100 million 🎉

splat.top is currently scraping again, and I'm slowly backfilling all the data that's missing (albeit at a slower 1-snapshot-per-hour cadence rather than the scraped 6-snapshots-per-hour). Here's the Skill Offset of the current meta

Just replays. In theory could be used live if it’s ever figured out how to pull the data in real time, framework should be the same. But I’m just assuming replays for now

It’ll have to wait for Easter, the church has more important matters to tend to

This is a replay yeah!

And dark magic

WIP