I’ve been tracking the moneypuck playoff odds model lately since I’ve noticed the odds seem way too noisy, and wanted to highlight an especially egregious change overnight.

The Canucks lost to the Sharks, and other western results were Kings Avs and Yotes wins and losses by Flames and Knights. No other major things to cause any noise. Yet somehow the Canucks odds went UP 8% to 89, and their cup odds went from 2.9 to 5.8%.

I am all for models and analytics but when you have this much movement off of a loss something is deeply flawed with the underlying calculations.

19 comments

  2. 70% for VGK to make it seems low. I can see them falling behind LA and Vancouver but even then, you’d need Seattle or one of EDM/CGY to finish higher, too. And that still doesn’t rule out a WC spot.

    I imagine that Canucks jump is from whatever the model didn’t like about VGK losing to ARI. Setting the Canucks up for home ice/division win better

  3. I’d have to know exactly what the model looks at, but the Canucks had a strong game, just lost. Outshot them 33-22, had an xGF% of 61% according to naturalstattrick, and generated significantly more high danger scoring attempts. I have to imagine moneypuck saw this as a positive game for them.

  6. I think the point of the post is how swingy it is day-to-day. The % of a single game being really good or really bad for a team should already be factored, in so if one of those things happen the very next day it shouldn’t change everything so much.

  9. Damn. The league must’ve removed Washington as a team all together. Wish they would’ve said something.

  10. Last year Moneypuck had a team listed with odds less than 100% to make the playoffs. They had already qualified at that point.

    Whatever model they use, it doesn’t work. They’ve made a homemade calculator, and are trying to convince us that 1+1=3.

  13. 538’s hockey model was always better than MoneyPuck’s, but ABC laid off half the employees and now they’re not doing it this year

  14. Listen I’m glad we’re finally getting close to 50% odds to make it

    But us going up by 10% based on the results of one day of hockey is nonsense lol

  17. Not defending Moneypuck’s usefulness, but this is probably just a result of sample size. Since teams have only played ~20 games, each game is worth an average of around 5%. Having games that are considered more meaningful to a team’s playoff chances being worth slightly more than that would make some sense.

  18. It’s not flawed it’s that it’s far too early in the season to be drawing conclusions from the model even though it can generate numbers because the sample size is not big enough yet, as the majority of the games haven’t been played. Not even half of them have been played. So small occurrences have bigger affects. Most people into analytics know this and don’t start looking at these percentages until the 3/4ths mark of the season.

    (I believe it even says this someone on their site about early season results in the sums as well.)

  19. You can’t just assume that losing one game, even against a bad team, will result in a lower ranking for that team. There are tons and tons of granular factors that come into play and I’d say the most immediately glaring factor in the Canucks game is that Desmith played instead of Demko. They’re also drawing penalties at a greater rate than committing. Just as an example of 2 things that will come into play.

Leave a Reply