Recent battle check might be too strict?

[Frostbyte0x70]

Hello! I just uploaded my first batch of battle data to stat.ink, and I discovered that unfortunately all of them are marked as "Used in global stats: ×No". After looking around in the code, I believe it's because the battles took place a couple days ago, and the limit to count them for global stats is 24 hours.

stat.ink/models/api/v3/PostBattleForm.php

Line 937 in 2605458

$startAt < time() - 86400 ||

What's the reason for such a strict time limit? I'm guessing there will be many more people like me, who don't upload battles right away, and I'm worried that a lot of valid data might be getting discarded. Perhaps the limit could be increased or removed?

[fetus-hina]

Thank you for using stat.ink and for taking the time to look into the code in such detail.

stat.ink is designed under the assumption that users upload their battle data almost in real time. In the past, this was a reasonable assumption, and most users were uploading their data immediately after each battle.

Currently, since almost all use cases rely on transfers from SplatNet, you are right that a 24-hour limit is probably too short, especially for users who upload data in batches or with some delay.

However, this 24-hour restriction is also used as a practical measure to reduce the amount of data that needs to be processed when generating global statistics. Without some kind of time limit, the volume of data to aggregate would increase significantly and could have a negative impact on performance and resource usage.

So while the limit may not be ideal in all situations, it plays an important role in keeping the global statistics system manageable.

[Frostbyte0x70]

I see. Performance issues can be tricky to solve, yes.

Is there no other alternative though? The way it's implemented right now, a lot of users will have their perfectly valid data discarded without them even realizing it. For people like me, who want to upload their data to help make global statistics more reliable, that limit means we won't actually get to contribute anything most of the time.

I could understand if it's not possible to remove the limit though. But even if it isn't, I would consider other alternatives, like dropping data at random rather than based on upload date, to reduce bias. Or keeping data for the most important modes only. Or increasing the limit a bit and testing how much of a performance impact it actually has, maybe it's not that bad.