Of course it doesn't really make sense for him to compare percentages. The total has grown by 150 million, so for me it makes more sense to say it like this:
Hololive - 61.8 mil hours to 90.7 mil
Niji - 59.8 mil hours to 76.6 mil
Indies - 54.5 mil hours to 145.2 mil
That said you also have to dig into the source data. The first chart says it only counts the 3000 most popular vtubers. The 2024 chart says it counts the 4000+ most popular. Of course adding ranks 3000 to 4000 would increase the watch hours total and also increase the indie watch hours share, as it's more likely ranks 3000 to 4000 are indie. In fact, even smaller corpo channels will rank in the top 1000 - the top 3000-4000 is virtually
all indie.
Streamcharts also had a categorization problem which has had constant improvement over the course of the last two years, so now more and more channels are correctly categorized as vtubers or correctly listed in the correct agency.
We can prove how many channels were missed in the 2023 data. Note how the Q1 2023 article listed 251.38 vtuber watch hours. In Q1 2024 Streamcharts published the following:
With 321.23 mil watch hours for vtubers in Q1 2023,
69.85 mil watch hours were not included in the Q1 2023 breakdown either due to not being in the top 3000 channels or due to miscategorization.
Platform distribution is also using the top 3000 vs top 4000+ which is sure to be loaded with Twitch 2-views and non-JP channels. It's a more accurate snapshot in 2024, but shouldn't be compared backwards.
In short, comparing the two reports is bunk. They aren't using the same dataset.
EDIT: I should
also add, while the Twitch API can pick up vtuber channels by looking at who is using the tag (and then manually removing those who don't count eg. xQc), the YouTube API has no such feature. You have to pick and choose YT channels manually too, you can't request a dump of the full YT stats. So a Twitch indie can be picked up automatically, but a YT indie has to be added
completely manually before you can pull any of their data. This naturally means YT streamers in the lower viewercounts won't be shown on any stat sites until someone submits them manually.
Streamcharts is the one charts website that has actually put effort into sorting through the vtuber data with a semblance of awareness of the issues, and the partnership with vstats (a well known JP vtuber stat site) is also reassuring for picking up the JP side of the market, but it's still an ongoing issue users of viewership stats should be aware of for anyone doing YT stats.