spiral-abyss VS genshinstats

Also if you're interested, here is a github that has a much larger sampling than what you've done here: https://github.com/appsample-com/spiral-abyss

Our team does think that our sample size is still small. Even though the sample size might be statistically enough, the problem lies in our method of data collection, which is via a form. This creates a bias towards players who are willing to participate, rather than doing a random sampling. I assume spiralabyss.org got their sample by random sampling, which is done by scanning random accounts. We've asked them about how they were able to collect such a large sample size , but they haven't replied yet.

The GitHub discussion: https://github.com/appsample-com/spiral-abyss/issues/34

In comparison with the previous Characters Usage, Diona's usage rate hasn't changed that much, there are other characters that have a more significant usage rate boost.

Sorry, I was referring to their Github. They only provide 1000 records.

That's understandable. If it can make you more comfortable to know what we can get with that token, it's all listed here

I'm pretty sure spiralabyss.org also collects their data from Hoyolab. We collect our data using this library, which only asks for the user's UID. spiralabyss.org also only collects the user's UID in their form.

Sorry, we couldn't find a way to collect the average clear times with the library that we use.

You can modify the source code of the hoyolab page to skip a check that only show that information to the mobile app. Here we had a discussion about it: https://github.com/thesadru/genshinstats/issues/36

The way we do it is using this library that plugs into Hoyolab's API. To use it, we need to (essentially) log in using a hoyolab account, but from there we can call for user's data using their UID (there's also a way to convert from CID to UID using the same API).