This season I decided that I’m going to start to try get into predictive modelling for NHL games. It’s something that tried out in baseball before, with a small amount of potentially unsustainable success, and although hockey is nowhere near where baseball is stats-wise, following Corsica’s prediction contest from last season inspired me to do a bit of it myself.
Plus, I’m supposed to write about stuff on this website semi-frequently and I can only do so many deep-dives into a seventh defenceman’s case for more playing time so often until it gets boring for me to write, and for you to read.
My hope is that my model will produce a modicum of gambling success that I will gladly pass along, in the form of profitable projections. But, be forewarned, this model is heretofore unproven, so tail me at your own risk. I promise that I’ll track its progress diligently, and if I’m being taken to the cleaners, I’ll shut it down and stop passing along bogus projections.
Before following any model’s projections, it’s a good idea to have a cursory understanding of how it works. The model I designed is based on a simplistic expected goals model that I designed, which is based on both shot volume and quality, with the quality being based on shot location groupings (high, medium, low danger) that can be easily found at https://www.naturalstattrick.com/. These stats are also broken down by strength state, so that I can determine how many goals each team would score and allow given its actual shot rates and general shot locations, but a league-wide average shooting and save percentage from each shot area at each strength state (meaning pp, pk, even strength, etc).
I then blend my expected goal model’s outputs with a team’s actual on-ice goal results, thus accounting for mostly goaltending defensively, and skill-based factors which aren’t easily captures in xG models offensively. Think shooting talent and pre-shot puck movement. I then adjust for home ice advantage, which is based on league-wide home/road splits from 2015 to today.
The Oilers may actually be a solid bet right now. Their xG numbers have been surging lately, and while they’ve been doing pretty well, even against formidable opponents, they’re still getting priced like the likely playoff outsider most books had them as prior to the season (not that I disagreed with that assessment at the time).
Tonight I have the Oilers as 52% favourites on the road in Washington, while they currently have implied odds around 47-48% at most books. And, tomorrow, I have them as 42% underdogs in Tampa, but many books have them in the high 30s.
This is probably worth keeping an eye on if the Oilers keep playing well. It’s obviously early on, but the Oilers’ early-season underlying stats suggest that they may be an above-average team for the time being. Which would mean that they’re likely a smart bet as long as they’re still getting priced like a team with a season-long point projection in the 80s.