Loading scores...
Matt Duchene
Getty Images
Hockey Analytics

Blue Jackets Bounce the Bolts

by Gus Katsaros

Two teams knocked out in four games. The playoffs are certainly one unpredictable animal.

We’re going to look at some of the underlying numbers for the remarkable sweep of the Tampa Bay Lightning by the Columbus Blue Jackets, but before we get into some data, we should address sample size, luck and the playoffs.

 

Don’t forget, for everything NHL, check out Rotoworld's Player News, and follow @Rotoworld_ HK and @KatsHockey on Twitter.

 

One of the caveats of using shot metrics in small samples is ‘noise’. Data that doesn’t reach a specific threshold for analysis can be really noisy and portray a different iota of ideas than what may be expected.

Take for example Game 1 of the Leafs/Bruins series, where a single game summary where Toronto’s Auston Matthews was deemed to only have contributed 0.07 expected goals. Only a handful of players fared worse, including defensemen Travis Dermott, Jake Muzzin and Nikita Zaitsev, and forwards Connor Brown and Frederik Gauthier.

 

Auston Matthews xGF example

 

I’m citing expected goals (xGF) here to indicate single game sample since xGF has an element of context. While Matthews may not have personally contributed an acceptable level of xGF generation, he countered that by having one of the best Corsi For percentages in Game 1. In fact, the line of Matthews, Kapanen and Johnsson tilted the ice in the Leafs' favor dramatically at 5v5.

 

Auston Matthews Corsi For Game 1 Versus Boston

 

So, while not really generating a lot of scoring chances via an expected goals model, the shot differential shows that while on the ice, Matthews and his line were dominant in creating more shooting chances than the Bruins.

This is an inherent danger of scrutiny to decipher effect using a small sample size. The context will require multiple inputs to support the outcome – and even then, the noise can be overwhelming. You’ll hear about ‘luck’ in the playoffs. Luck in statistical analysis refers to the lack of repeatable events that have been determined to influence winning. Sometimes these repeatable items can be parsed through data, with indicators such as inflated goal scoring, but without the requisite shot attempts, concluding with an inflated shooting percentage, individually and on-ice.

Going beyond the ‘eye test’ – a phrase I am loathe to utter, using a single game summary is full of noise, but you too, dear reader, can cut through the ruckus.

During the regular season, it usually takes about 20 games for Corsi to contain predictive capacity. A bigger sample size, the better. With the playoffs really being about each individual game, a team that is out in four games in the opening round won’t have much, statistically, to offer as part of predictive analysis. But analysts can decipher the components to provide contextual analysis. Even teams that ‘trust’ the process will have to make adjustments along the way and that can also skew some results.

This is where video and data meld to really produce a well-rounded picture, as we shall see with the Blue Jackets shortly.

Similar to the ideology that spawned expected goals – trying to use all available shot metrics data with a numerical value attributed to each shot location, rebound angle and situation, provides a greater sample size, and some qualitative context in which to decipher effectiveness. If you’re going to be using more encompassing metrics, stick to expected goals and measure they’re effectiveness to real goals scored. Even then, after only three or four games, the numbers are still forming and not finalized enough for full context. Tread carefully.

 

Columbus Blue Jackets Bounce the Bolts

First, congrats Columbus. An absolutely remarkable sweep of a record setting team in Tampa Bay.

Before getting too heavy into the Blue Jackets series performance, make sure to check out the primer and tactical outlook by Alison Lukan of The Athletic. Alison uses simple concepts and language to integrate numbers with video and commentary, breaking down the series and offering an in-depth examination of the smothering Columbus 1-2-2 forecheck. And boy, did it every play out in Game 4.

So is a four game sample enough to remove some of the noise?

A practical example of using expected goals here are the Blue Jackets heading into Game 4 of Round 1. The Columbus Blue Jackets have the reigning Presidents Trophy winners against the ropes, leading 3-0 in the series.

One would have to figure that to have accomplished that, they would have had to outplay the Lightning by a significant margin. But sometimes a playoff series emulates death by a thousand paper cuts, small mistakes, sloppy play and tactical execution taking on more significance as the series gets away from the Lightning, more and more out of reach.

Have the Blue Jackets really outplayed the Lightning? Let’s check the data.

 

 

Round 1 Situational Expected Goals

 

There’s a lot going on in this table, and before we hit Columbus, check out the xGF differential for the Carolina Hurricanes at 5v5! The Canes have scored seven times with a 6.86 xGF and allowed five with a 3.88 xGA. The 63.8% xGF% differential is astounding. Yet, the Capitals hold a 2-1 series lead. Seems like a lot of defending when taking into consideration the 8.64% shooting percentage, in comparison to the Caps 11.11%.

Focusing once again on the Columbus xGF and (xGA), in three games leading up to the pivotal elimination game, Columbus had generated 3.69 xGF and scored six at 5v5 (all stats presented here are at even strength 5v5, unless otherwise noted). They’ve racked up 4.74 xGA and allowed four goals.

Taking it to the power play – which is a diminished data set to keep in mind – they’ve generated about 1.11 xGF, but scored four goals, and a league wide high of 30.77% shooting percentage. They’ve outperformed their expected goals in both situations and overall it’s even more skewed, with 5.41 xGF and 12 goals scored. Aside from Vegas who topped them by small decimal percentage points, the Blue Jackets firing at 14.5% overall is remarkable. There’s a lot of defending in that regard too, but the Jackets did an excellent job smothering the Presidents Trophy winners.

After the Game 4 sweep, Columbus scored eight goals at 5v5, and allowed six with an xGF of 5.32 and xGA of 6.46, while only controlling 45% of the xGF.

In Game 4, all situations they looked like this, courtesy of Natural Stat Trick. Now, there are a few empty net goals as good measure to pump up the stats, but even then, they scored four times before the empty netters. Game 4 was the first in the series where they scored less goals (in all situations) than their xGF.

 

Columbus Blue Jackets Game 4

 

With the man advantage, the expected goals measures are even more baffling from a Tampa Bay perspective. Tampa Bay, who finished with the season’s best power play only achieved an xGF of 0.81 in 10.03 minutes at 5v4. They scored once, in Game 4. It was simply put, a horrible performance.

 

*************************

Be mindful of the number in small samples.

Congrats to the Blue Jackets and New York Islanders fans. They’ve taken a beating over the years. But this is sweet moment in which to rejoice. On to Round 2.

Gus Katsaros
Gus Katsaros is the Pro Scouting Coordinator with McKeen’s Hockey, publishers of industry leading scouting and fantasy guide, the McKeen’s Annual Hockey Pool Yearbook. He also contributes to popular blog MapleLeafsHotStove.com ... he can be followed on Twitter @KatsHockey