Wild West: Predicting Playoff Winners with Jelly Beans

Chris Kane

2018-04-16

So what do the NHL playoffs have to do with jelly beans? In truth, not much, but more on that in a second. At this time of year we are bombarded with everyone’s playoff takes on who should win and why, and the “X Factor” that all but guarantees it. For the most part though, only a few of the theories are based on something more than feeling. This article is going to summarize a few more data driven or evidence-based theories and kick off a mini-series where we compare their predictions in the Western Conference this year. The theories are as follows:

 

Expert Panel

Back to jelly beans. Imagine you and your friends are at a county fair and in front of you is one of those big bowls of jelly beans. There is a prize for the guess that is closest, but your group can only guess once. How do you do it? Talk it out? Everyone guess by themselves? It turns out there is some research on this phenomenon in behavioral economics (if someone can provide a reference in the comments I would greatly appreciate it). The idea is that if the group were to talk, the guess would be closer to what the person who speaks first and loudest suggests rather than what the actual group thinks. The better plan would be to have everyone guess individually and then average the guesses.

 

Great. What does that have to do with the playoffs? Well we recently released predictions from the Dobber Expert Panel. Take a look here. A great thing about the article is that it listed the “Consensus Picks” which essentially acts like averaging the jelly beans in the story above. It was done blindly, with everyone creating their picks independently of each other, so theoretically should be a better predictor than our individual opinions.

 

Goal Differential

Back on a March 27 Rambling, Steve Laidlaw shared some work he had done on playoff success and provided some more rationale as to why goal differential is the most important detail to look at when trying to predict playoff success. The general gist being that there are lots of strategies or skills a team can use to outscore their opponent, but the best predictor of playoff success is that they are consistently able to do so. In the case of Nashville v. Colorado, Nashville's regular season goal differential of 57 suggests they would prevail over Colorado’s 19.

 

Matchup Specific Chance and Goal Differential

My apologies to Travis Yost, but I had to call it something and the article didn’t really give it a pithy name. Over on TSN, Yost wrote a great piece outlining how he uses head to head matchups during the regular season to find out which teams out-chanced and outscored their opponents. The general theory being that analyzing chances gives you a larger data sample than just goals. If your team has been consistently out-chanced by their playoff opponent during the regular season, it's likely the opponent will in the playoffs too.

 

‘Expert’ Analysis

Just for comparison, a control, if you will, I have included the predictions