HN's "flamewar detector" (I prefer "spiciness indicator") is one of those tools I'd been highly skeptical of, and
still have
significant issues with, but ... it
does in fact work much of the time. It's where it
doesn't work that the problems really manifest.
I'd come to that conclusion after scraping all "past" front pages from 2006 through May-ish 2023 and doing a number of analyses of that corpus. (I've commented on that a few times here on HN and on the Fediverse.)
One of the absolute spiciest discussions ever was a pg post about HN itself. Which suggests that when a topic of of direct interest and familiarity, people will tend to hop on it. There are also a great many flame-y threads, though note that by virtue of making the front page, my sample probably skews to less disastrous threads with the absolute sh*tshows being well below the top-30 fold.
Among other weaknesses, spiciness doesn't distinguish between pure troll/clickbait threads, and those on which there's a significant and justified spread of opinion. As such, the metric makes hard discussions even harder to have, though mods can and do turn off the penalty on request (often many hours after the discussion's started, for obvious reasons, which is its own penalty). I do wish that HN could have those discussions, and I've thought and written (both on HN and in emails to mods) about what that might entail. I'm coming to the long-delayed and somewhat regrettable conclusion that it's not the right tool for that particular job.
