Something called the Hokkaido Bank Classic took place in Japan last weekend, and that means a new season of curling has begun. I’m not one of those people that wants a long curling offseason. Bring it on, I say. Having a regular summer schedule is probably a bit much, but a couple events in August is a fine way to ease us back into the season before we start getting multiple events every weekend.
In two weeks, the Baden Masters will begin, with the first cross-ocean field for a European event since before the pandemic. And I will be looking forward to some early morning (local time) draws involving Korey Dropkin, Colton Flasch, and the best teams that Europe has to offer.
Anyway, the ratings have flipped over and all of the fun of trying to handle new teams has begun. While teams have been training for a new season, I’ve also been doing some things to get ready for the upcoming season.
Most of my efforts in recent months have been spent liberating shot data from the depths of the dark web. We now have the largest shot database in the galaxy. Not that there was any competition for that honor.
I have over one million shots logged across 7500 games, which includes memorable games like USA’s gold medal win in 2018 and Carey’s double-steal to win the 2019 Scotties, and absurd games like Dunstone’s 33-1 win in the 2016 Canadian Juniors.
I’ve also updated the algorithm that estimates the shot difficulty displayed on the play-by-play. It no longer overfits as badly as it did before, and there’s internal consistency between the chance of a 4 and the chance of a zero. It does about as much as you can do with the information available.
I’ll have more to say about the quality of shot data in the future, but it’s definitely a problem that so many shots are erroneously recorded and that Curling Canada and the WCF have different methods for recording shots. I’m thankful the data exists but also hopeful that Curling Canada will take the data quality more seriously in the future.
Most of my time has been spent on developing a shot-by-shot win probability model. I’m so close to releasing something, yet so far away. There’s a lot of insight to get out of such a project and I’m looking forward to sharing something soon, even if it’s got a few warts.
One thing I haven’t done is touch the team ratings. If the ratings get screwy early in the season with all of the new teams, I will look at some tweaks. My favorite wrinkle in the ratings is that it will eliminate last season’s teams once it sees proof on the ice that they have broken up. For instance, once Marc Kennedy and Ryan Harnden take the ice with their new teams, Brad Jacobs will be removed from the ratings. So that issue should take care of itself pretty quick.
In fact, this has already happened to Japan’s Team Matsumura, who finished last season at 23rd, as Yuta Matsumura joined forces with Yusuke Morozumi at the Hokkaido Bank Classic, thus removing his previous team from the ratings.
That’s probably the extent of my Japanese curling coverage for the season. Hopefully, I’ll soon have more news for the site.
