in 2019 & 21 I chaired election campaigns for a candidate. I opened 338 daily. In both cases they predicted the ultimate winner, but the margins were way off. The number polled even in national polls is so small I find the results somewhat dubious.
When you then try to extrapolate national or provincial polls to a local level with limited local knowledge it really is a "best guess". I like the site but treat the data with justified skepticism.