The main thing that models got wrong on the Blues last year was their implementation of expected goal models. These tend to overvalue a guy whacking a rebound in the crease into a goalie's pads repeatedly (cough cough Brady Tkachuk) and devalue effective snipers like Perron who routinely score goals from the circles that aren't high probability shots.
The models hurt the Blues last year because our offense was extraordinarily predicated on cross-ice passes. Get the goalie to commit, reverse grain, bang it in the open net. We both attempted the most and scored the most off of them than any other team by a wide margin, and our teamwide shooting percentages were ridiculous last year as a result.
This approach also means lots of intercepted and deflected passes, depressing overall shot totals and subsequently expected goal totals. Last year's Blues traded a lot of .3 to .5 xG shots for sometimes hopeful passes to guys with even better shots.
It was fun to watch and gave us the most effective offense we've had in 30 years. It also didn't model as well as if we'd just taken the first shot.
Edit: Look at Perron's EV offense. It's stupid. EV defense seems off too, but that finishing total is spot on and boosts the overall.
View attachment 575714