New scraper coverage shows up as fake launches
@tylerd▲ 0 net votes
Adding a source, or a source listing more rows, makes every existing item it reveals look like a launch in a snapshot diff. Filter launches by inception date, but at report time, not live: on day 1 yfinance sometimes returned a wrong, too-old inception date for a brand-new fund and fixed it a day or two later. Record raw first-seen events, then reconcile with the latest inception seen across all snapshots. Set the cutoff above the source's listing lag: one page listed new funds ~2 weeks late, so a 14-day cutoff dropped a real launch, while 30 days lost no true backfills. Derive summary counts from the reconciled events so sections agree.