Silver Nuggets: Playoff Probabilities

Today's nuggets, as well as a look at how playoff odds are generated.

So how about last night's game, eh? The Senators kept taking the lead, and eventually hung onto it against the Bruins. It was very exciting hockey, and all worth it because the Sens won in the end. We all know how big of a deal the game was:

Sports Club Stats says that last night's results (including the Capitals' win) jumped the Sens up to 50.4% while the Bruins dropped to 54.7%. Those are amazing odds considering how bleak the Sens' season looked even a couple weeks ago.

All this got me thinking about how the odds are actually calculated. Full disclaimer: I love the process of stats. Not specific to hockey, just in general. Statistical analysis is fascinating to me, how people have developed systems to plow through large amounts of data. If you don't love stats, this next section might not be too interesting to you. But if you love clicking on buttons and watching numbers change, you might still find it interesting.

Sports Club Stats has this great section where they explain how they calculate their odds. You can read the whole explanation if you want, but in short, they use what are called Monte Carlo methods. These work exactly how you'd expect big simulations to work: an algorithm runs a whole bunch of randomized simulations following a certain set of rules, and then the outcomes are averaged. In the case of playoff predictions, they simulate the outcome of every game for the rest of the season, and see where each team ends up. There is an interactive section at the Sports Club Stats explanation link above where you can click to see the 2006-07 playoff race results after 1, 2, 3, 4, 5, and then powers of ten up to 10,000,000 simulations. It really illustrates how the technique works.

You may notice I said the simulations follow a certain set of rules. For Sports Club Stats, they just use a team's earlier home and road records, and apply this to the number of home and road games upcoming. Obviously, there are flaws in this. For one, some teams are better than their records (such as the Kings) and some are definitely worse (perhaps the Flames). As they point out themselves, it ignores factors such as mid-season serious injuries to players who have now returned to the lineup. It also assumes that home-road win-loss rates are repeatable, which is a contentious claim.

Micah Blake McCurdy (@IneffectiveMath) with the username tolstoi around these parts wrote an article a little while ago explaining his simulation technique, Pip. He also uses Monte Carlo methods, but instead of using a team's win-loss record, he uses Fenwick (USAT) as a predictor. Basically, he looks at each team in each game's typical Fenwick for and against. He's developed weights for the four inputs using years of past data, and combines them to predict the likely outcome of a game. The simulation takes these into account when randomizing each game outcome. This does away with some of the issues of the Sports Club Stats simulations, since Fenwick is a better predictor of future performance than straight wins and losses, but still doesn't account for things like injuries and recovery.

Thanks for indulging my little stats spiel. No model is perfect, but the models are getting better. And in my opinion, talking about the importance of games is more fun when you can quantify it. I'm just thankful there are people smart enough and patient enough to put together this information for the rest of us to use.

Tonight's games that matter:

Really, there aren't any. There are only three games, and only Detroit at Tampa Bay could have implications for the Sens. If you want Montreal to win the division, cheer for Detroit, who are less likely to catch the Habs. But if Detroit still leaves you with a sour taste in your mouth (like they do for me) then follow your heart. It probably won't matter much.

Sens Links:

Non-Sens Links:

  • Matthew Wuest, the founder of CapGeek, passed away yesterday morning from colon cancer. He put immeasurable time into CapGeek, made it available for free online, and never went looking for the attention he could've for having such a crucial site in the hockey universe. He will be truly missed. [Metro]
  • Last night Jaromir Jagr scored his 718th career NHL goal, passing Phil Esposito to take sole control of the fifth-highest total in NHL history. Just think how high he could be without those years in the KHL. [NHL]
  • Team Canada has made the 1-2 Page Playoff at the Women's World Curling Championships, on the back of a 9-2 record in the round robin. Jennifer Jones' group has looked shakier than usual this tournament. The Canadians take on the Swiss tonight and if it's anything like when the teams met earlier this week, it'll be a doozy of a game. [TSN]

Not everyone can afford to pay for sports coverage right now, and that is why we will keep as much of the site's content free for as long as we can.

But if you are able to, please consider subscribing to help keep our articles free (and get a few extra perks).

Erik Condra
  • Ability to comment and participate in our community
  • Twice monthly newsletter available only to subscribers
  • Ad-free reading
  • Our undying love and appreciation
Brady Tkachuk
  • Everything from the Erik Condra tier
  • 10% discount on all merch
  • Access to any future paywalled content
  • A personal thank-you from the Silver Seven staff
Daniel Alfredsson
  • Everything from the Brady Tkachuk tier
  • Inner peace knowing you are supporting quality, independent coverage of your favourite sports team