News

Advanced stats vs. the eye test: models split on who will win Stanley Cup

Dom Luszczyszyn
By:

(Photo by Justin K. Aller/Getty Images) Author: The Hockey News

News

Advanced stats vs. the eye test: models split on who will win Stanley Cup

Dom Luszczyszyn
By:

This Stanley Cup final will be hotly contested and tightly played. Here’s who both schools of thought think will win it all.

The long road to the Stanley Cup final is finally over and depending on your outlook, the two teams playing are either very surprising or exactly what you expected.

From the East we’ve got the Pittsburgh Penguins, a team with a sketchy D-corps and – at the start of the playoffs – an injured starting goalie. A team that has disappointed a few times in past playoff seasons despite their talent.

From the West we’ve got the San Jose Sharks, a team not many would believe could actually go far in the playoffs without actually seeing it for themselves first. A team that has choked year after year.

I can see why some people would be surprised.

Before the playoffs began, we here at THN previewed three different sets of predictions: one based on stats, one based on the eye test, and a combination of the two.

The eye test was based on a survey of 15 people inside hockey. Together, they ranked Pittsburgh and San Jose as average playoff teams; eighth and ninth best respectively. Instead, they looked unstoppable as they toppled the best (Los Angeles) and second best (Washington) teams on their way to the final.

The issue was how the panel weighted each aspect of the survey, with too much emphasis placed on defense and goaltending over offense, a strength of both clubs. With new weights based on the forecasting accuracy throughout the playoffs that put more emphasis on offense, San Jose ended up as the sixth best team, with Pittsburgh as the seventh. A little better, but not really legit contenders.

The ‘eye test’ gave this specific Stanley Cup matchup a one percent chance of happening. The most likely was Los Angeles vs. Washington at 7.2 percent.

If you paid attention to the underlying numbers for both teams, though, you’d be much less surprised. By the statistical model we’ve been using for these stats vs. eye test posts, Pittsburgh was actually considered the best team before the playoffs started while San Jose was third.

The chances of this exact playoff matchup was 4.3 percent according to this stats model. The only reason it was that low too was because the Sharks first round opponent happened to be one of the only two statistically superior teams: Los Angeles. A Kings-Penguins final had a 6.2 percent likelihood.

This post-season the stats have been much more prescient. Here’s a recap for each round and how many series each model ‘called’ correctly. The numbers were better by one series in each round.

Round 1

Stats: 5-3

Eyes: 4-4

Both Stats and Eyes picked Los Angeles, Anaheim, and Florida

Eyes picked Chicago

Round 2

Stats: 4-0

Eyes: 3-1

Eyes picked Washington

Round 3

Stats: 2-0

Eyes: 1-1

Eyes picked St. Louis

Total

Stats: 11-3

Eyes: 8-6

It’s only a three series difference, but it’s also 79 percent accuracy vs. 57 percent. There’s other stat models around that have been even more accurate, calling a mind-boggling 13 of 14 series correctly (the honours there go to Shane O’Donnell of Hockey Prospectus and DTMAboutHeart of Hockey Graphs).

As a whole, stat models have been more accurate when using log-loss – a measure of prediction accuracy – this post-season compared to pundits with an average log-loss of 0.583 compared to 0.634 (lower is better).

Of course, there’s a sample size issue at hand here. It’s only been one year and it’s easy to get lucky. Still, this has been a good playoff year for the numbers crowd and that shouldn’t be ignored. This isn’t meant to be a referendum for the entire stats vs. eye test debate, but merely a revelatory experiment for anyone who still doubts the powers of hockey’s new numbers movement.

The biggest thing that analytics types have preached in the past is the predictive power of the new stats they’re peddling. This experiment and this post-season should be an eye-opening experience in just how true that can be.

With that being said, here’s who both schools of thought think will win the Stanley Cup.

pit-sj

What The Stats Say

It’ll be close, but this series is Pittsburgh’s to lose. Since Mike Sullivan arrived as coach, the Penguins have been an entirely different team. They’re the second best shot rate team on the strength of their elite shot generation. What’s interesting though is that their shot suppression was nearly just as good since Sullivan took over. Only Los Angeles and Anaheim allowed fewer attempts. On paper the 'D' may look sketchy, but they get the job done thanks to Sullivan’s system.

In that same time frame, the Sharks have really come around. Their shot suppression has been just as good and their shot generation is decent too, just not at the same level as Pittsburgh’s. Together, they’re the third best possession team since Sullivan arrived and pose a very real threat to the Penguins. In terms of goals, they were fourth best while the Penguins were second. They’re an elite team that deserves to be here, but aside from their powerplay, Pittsburgh has them beat in most statistical categories which is why they’re considered the favorites here.

What The Eyes See

For the fourth time this playoffs, the eye test has a differing opinion to what the stats say. Each other time the eyes were proven wrong, but perhaps this time will be different. Instead of seeing the Penguins as favorites, the eyes see this as closer to a coin flip, with a lean toward the Sharks.

There’s a lot to like about this Sharks team, but their biggest advantage is their strength on the backend. It’s been mentioned a few times that Pittsburgh is weak there, but the difference looks even more lopsided when you place them side-by-side with San Jose’s group. The Sharks third pair isn’t strong, but there’s very few teams in the league with a better top-four than Brent Burns, Paul Martin, Marc-Edouard Vlasic, and Justin Braun.

Will that advantage be enough to win it for San Jose? I’m not sure. But in terms of player skill, it’s the biggest mismatch here. Both teams have an abundance of high-level forwards complemented by competent depth that could make offence a wash. Ditto for goaltending where both Martin Jones and Matt Murray are inexperienced, but serviceable.

In both cases, you’d probably lean Pittsburgh, but I don’t think it’s enough to forget about the discrepancy on defense, which is why San Jose is the eye test favorite.

Why Not Both?

Once again, the main theme is combining both ideologies. There’s a lot to like about both teams, but the mismatch on defense means San Jose probably looks better and deeper on paper. But it’s just impossible to ignore that Pittsburgh has played better ever since Sullivan took over the bench.

As is always the case, the truth lies somewhere in between. This Cup final will be hotly contested and tightly played. It has the makings of being one of the best finals in recent history because both teams are so good, so fast, and score a lot of goals. It’ll be fun, it’ll be great and it’ll definitely be close.

But if you had to pick one team to win it, it’s hard not to have a slight lean toward Pittsburgh. And that’s exactly what combining numbers with the eye test suggests.

Comments
Share X
News

Advanced stats vs. the eye test: models split on who will win Stanley Cup