Ah, LeanTossup, the same tosspots
that projected Biden would get 383 electoral votes, winning Florida, Texas, and Ohio and had Biden as winning the popular vote by 10%, double the actual margin. Although to be fair, they were fairly accurate in Alberta in 2019. If they shared any methodology whatsoever, I'd have more confidence in them, but right now it's a black box, a pretty map and wishful thinking.
Certainly recent polling hasn't been great for the UCP; they are of course wildly unpopular in Edmonton, and
recent polling in Calgary has them at 29% versus 42% for the NDP, with them leading 33% to 27% in the Rest of the province. But there's a very large undecided share; 13% in Calgary and 19% in the Rest, plus substantial shares in the Wildrose Independent party (7% in Calgary and 13% in the Rest). The undecided voters are almost all conservative leaning and low-participation voters. The dynamics of a campaign can really change these voters; as a comparison, models in the 2015 Federal election tended to show a slim Liberal or Conservative minority, because there was a large share of undecided voters who weren't being included; in reality, most of them hadn't decided who they were voting for, but had sure as hell decided they were sick of Harper and jumped on Trudeau's bandwagon as it became clear he was the guy to beat Harper. I'm not sure I'd bet money on all of those undecided and third party voters remaining so if it becomes clear in a close election that voting third party will hand the reins over to Notley.