Ice Hockey’s knowledge war hits the playoffs

Friday, April 18th, 2014

With the Stanley Cup Playoffs under way, Zoe Coleman McNair looks at the struggle for acceptance of advanced stats in the NHL and breaks down the match-ups.

In 2012, the LA Kings barely scraped into the 8th seed in the Western Conference, and proceeded to practically cruise through the playoffs, only dropping four games before winning their first Stanley Cup. Everyone loves a Cinderella run, and the Kings’ story seemed to tick all the boxes, except one: had they really been an average team that got good when it counted, or a good team whose bad luck disappeared when it mattered?

The answer isn’t as clear cut as that, but it definitely leaned towards the latter, putting the Kings squarely into the middle of the advanced statistics debate – or as they’re known in hockey, ‘fancystats’. Unlike in baseball, where analytics are common, hockey media (and many team organisations) have been reluctant to embrace anything more advanced than the most basic statistics. Part of the problem is that unlike in baseball, which is more easily distilled down to a series of single events, hockey is a messy sprawl where multiple things are happening at once, making it that much harder to track.

The movement towards embracing fancystats is growing, even though the stats themselves are imperfect. The problem is that the stats used in hockey don’t really correlate exactly with a team’s record. Instead, they can be used to look at the underlying strengths and weaknesses of a team, and if a team’s record is outperforming their underlying numbers, the stats will show us where the eventual crash back to earth will come from. Unfortunately they can’t tell us when that crash will happen, so many have dismissed them as useless.

The most commonly referenced stats are Corsi and Fenwick, which (in slightly different ways) puck possession, counting how many shots are aimed towards/around the opposition’s net during a game. They tends to correlate with success, because the more times you aim the puck at the net the more chances it has to go in, and if you have the puck it means the other team can’t aim it at your net. PDO roughly measures how lucky a team is being, by taking two numbers that over enough time should add up to 100, and measuring how far above or below that a team is.

This year’s playoffs feature multiple teams whose positioning doesn’t correlate to their underlying numbers, which makes the early rounds ripe with potential upsets. Colorado have one of the best records in the league, finishing third in the regular season, but their possession stats are so bad that the only teams worse than them are outright terrible, and the only thing keeping them as alive as they are is both their goaltenders performing above and beyond. Montreal are ninth in the league though they’re barely ahead of Colorado in possession, but have the reigning Olympic gold-medal winner in their net, which has kept them alive enough to make the playoffs. Anaheim rode the second highest PDO to the top of the Western conference, but their possession stats don’t even break the top ten, which means they really don’t have any leeway for their shooting percentage (the amount of shots on goal that become actual goals) to regress very far before they could be in serious trouble, especially with a new question mark over their goaltending after their regular season starter put together a string of underwhelming performances.

The other side of the coin is New Jersey and Vancouver, both top 10 possession teams with low shooting percentages and uneven goaltending dragging them out of the playoffs. In comparison, Florida are solidly middle of the league in possession and the second worst team in the league by their actual results.

Toronto are perhaps the ultimate example used in the fancystats versus results debate. They managed to take advantage of exceptional goaltending in the early season to keep themselves in the playoff hunt, while the stats fans kept pointing to their terrible possession numbers as an indication that it wouldn’t last. It didn’t, and Toronto finished the year in 23rd place, but even that was better than their dead-last possession numbers.Their seeming refusal to accept that their success was unsustainable, even as their season collapsed around them, has lead to even those who are less dedicated to fancystats wondering how Toronto could have ignored their underlying problems for so long.

This despite the fact that other teams – like Colorado and Montreal, had far greater discrepancies between their possession numbers and their eventual positions. If there’s nothing else you can rely on in hockey, you can at least rely on the rest of the league delighting in the failures of the Maple Leafs.

But while the Leafs eventual downfall provided ammunition for the pro-stats side of the debate, the other teams’ continuing success in the face of subpar possession shores up the other side – if the stats can’t be relied upon to be predictive across the whole league rather then just individual teams, then how much use can they really be as an analytics tool? The fact that regression is mostly inevitable doesn’t help us if we can’t predict when it’ll begin, and if a team can string together sixteen wins with very few losses during the regular season even with terrible possession, there’s theoretically nothing to stop them doing the same in the playoffs.

But the fact of the matter is, relying on best-in-the-league goaltending to provide your wins isn’t something that’s sustainable in the long term. Even the teams that do have generally outstanding goaltending over a long period of time – New York, LA, Boston, are able to back that up with above-average to strong possession numbers, meaning that they have more leeway to still succeed even if their goaltenders hit a string of below-average performances.

At the end of the day, teams like Toronto ignore the statistics and put together chaotic and eventually doomed seasons, while teams like Chicago scrupulously track analytics and enjoy repeated seasons of success and two cups in four years. Even if hockey’s fancystats are imperfect predictors and still young, with the increasing visible correlation between good statistical performance and team success, the debate isn’t going away anytime soon.

First Round Matchups:

Anaheim (1) vs Dallas (8)

For my money, this has the potential to be one of the best matchups of the round. Anaheim have the best record in the West (only one point behind league-leaders Boston), but their underlying numbers imply they’re just waiting to regress. The only problem with regression is we never know when it’s going to arrive, and while Dallas’ underlying numbers are good, they still have a bunch of question marks over their defense. If both goalies stay on form and the top line pairings of Getzlaf/Perry and Seguin/Benn contribute as we know they can, this could easily go the distance

Colorado (2) vs Minnesota (7)

Expectations for Colorado started out so low it would have been hard for them not to eclipse them, but it’s safe to say nobody predicted them winning their division (probably team-for-team the toughest in the league). However, the majority of their success has been down to stellar goaltending masking some atrocious possession stats (none of the teams below them made the playoffs) and most people are expecting a collapse to come soon. Unfortunately for Minnesota, their possession numbers aren’t much better, and a string of injuries and illnesses mean they probably don’t have stellar goaltending to rely on. With any other central opponent, Colorado would probably be done, but this could be their ticket to the second round.

St. Louis (3) vs Chicago (5)

On paper, this matchup is a bloodbath, with many fans disappointed it’s not going to be in a later round. Chicago are last year’s champions and have seemed to avoid the worst of a cup hangover, and St. Louis spent most of the season rolling over everyone in their path. But injuries to key players and late-season skids left both teams out of the running for top of the central. Chicago finished the year barely able to beat anyone in their division, and St. Louis went out on a 6-game losing streak. This at least means the series should still be evenly-matched, though perhaps not quite as good as fans had hoped. It’ll likely come down to which team rebounds from injuries faster (and has fewer players playing through them), and whether or not Ryan Miller can get back to his early season form.

San Jose (4) vs LA (5)

Similarly to St. Louis vs Chicago, this is another series fans were hoping to wait a little longer for. Both teams have put together strong regular seasons despite bouts of scoring difficulty, and when they met in the second round last year it went to seven games, with five of those being decided by a single goal. San Jose are hoping Tomáš Hertl’s return from injury will add to their scoring and that other key players don’t falter if they’re going to keep on track for their first cup, and the Kings need goaltender Jonathan Quick to turn in another excellent post-season, because the Kings score fewer goals than the majority of the league. However, if he’s back on top form then the Sharks getting anything past him is going to be a tough sell, and there’s every chance they take it to seven games again.

Boston (1) vs Detroit (8)

Boston have the best record, the best defenseman, the best defensive forward, and probably the best goalie in the league. They made it to the finals last year and are definitely planning on making it back. Detroit have the best coach in the league (arguably the sport), but are missing key players in Zetterberg and Datsyuk. It’s not impossible, but it’ll be incredibly hard for Detroit to pull an upset. If Detroit’s younger players can push to a next level AND Boston have an unexpected meltdown then maybe, but most watchers have already called this one.

Pittsburgh (2) vs Columbus (7)

Almost every possible prediction has been made for this series. On the one hand Pittsburgh have been in the playoffs every year since 2007, including one cup, one cup final and one conference final, and have the best player in the world in Crosby (on top of Malkin, who’s top 5), and the second best record in the east despite more injuries then any other team. On the other, their bottom six is a possession black hole and they have a goalie who many doubt is capable of winning playoff series’. Columbus have a lot of players who’re desperate for post-season success (they’ve only ever played four playoff games, all losses) and a goalie who won the Vezina (Best goaltender in the league) last year. Either team winning is a possibility, but there could be big shake-ups in store for Pittsburgh if they suffer another premature playoff exit.

Tampa Bay (3) vs Montreal (4)

Tampa Bay are another team few were expecting to make it this far, after ending the 2013 season 27th in the league, but they’ve outperformed expectations to a huge degree. Tampa have better possession and Montreal has better goaltending, but with Tampa having lost one of their top players at the trade deadline and their starting goalie for probably the first couple of games at least, they’re left starting the playoffs with their backup. Lindback has been good so far, but hasn’t had enough starts to really anticipate how he’ll fare over a best of seven series, and they need Stamkos and the rest of the team to be on top form, whilst also hoping Montreal’s scoring woes continue, and that Price’s goaltending doesn’t keep papering over the cracks in possession.

New York (5) vs Philadelphia (6)

Philadelphia were so bad at the start of the season that their coach got fired after three games. Their possession stats are middle of the road, their goaltending is bad, and they’ve barely managed to score more then they’ve allowed. They’ve lucked into their starting goalie going above and beyond a few times, and their forwards getting it together when it counted to claw their way to the post-season, but their defense core has some serious holes. New York also started the season off badly, but rebounded better, and their possession numbers are better and their goaltending is near the top of the league. The Rangers will be aiming to take an early lead and shut the door, and when Lundqvist is on top form, there aren’t many players that can get many past him.

One Response to “Ice Hockey’s knowledge war hits the playoffs”

  1. Can You please write more? It’s very useful. :)

Leave a Reply