Loading scores...
Derek Stepan
Getty Images
Hockey Analytics

Arizona, Regression and NHL Tracking

by Gus Katsaros
Updated On: October 17, 2019, 2:00 am ET

I am a weekly guest on The Big Show on TSN1290 in Winnipeg, where I’ve been asked prior to the season kicking off about a team to be excited about, or to have heightened expectations.

Three years running, my response gravitated to the bubbling up in Arizona. They tweaked for the 2017-18 season, only to fizzle. Surely, the improvements in 2018-19 were enough to overcome obstacles, weren’t they? They weren’t. Arizona fizzled again – with injuries contributing to the overall poor showing last season.

But, in the summer of 2019, the ‘Yotes’ retooled, yet again. They added a big name in Phil Kessel, offering up Alex Galchenyuk as a trade piece, and my motor revved up once again, firing on all cylinders, full forward! Coincidentally, the Yotes broke through their shell against the Winnipeg Jets, of all teams, coming out with a 4-2 victory. They also scored their first two goals on the power play.

Editor's Note: Drafting is only half the battle. Dominate all season long with our Season Pass! Use our NEW Lineup Adviser, get our Weekly and Rest-of-Season rankings and projections, track all of your players and more on your way to a championship! Click here for more!

When asked about that team just prior to the 2019-20 season, by the Big Show, I went back to the well, for the third time in as many seasons, emphatically choosing the Arizona Coyotes.

Here’s why.


Entering Tuesday night’s game against the Winnipeg Jets, Arizona had played four games scoring a whopping total of seven (7) goals at 5v5, and sporting the third worst shooting percentage.  They scored twice at 5v5, and added their first two goals on the power play. A breakout game, indeed.

Early results are stingy defensively at evens, allowing only five goals against, with a respectable .947 save percentage. Goaltending – and to a lesser degree, defense – were points with some upgrade, but so far have held steady. News of an extended absence in the three months range to Niklas Hjalmarsson courtesy of a broken fibula from a blocked shot by Colorado’s Erik Johnson is distressing to the defensive game. If they falter, they’ll have to rely heavier on a rebuilt forward unit to score more goals than the Coyotes are bound to give up. This is the play.

Here was the shot block by Hjalmarsson pic.twitter.com/3hbcMwDnQR

— Here's Your Replay ⬇️ (@HeresYourReplay) October 14, 2019


If 5v5 play takes a hit, improvements to special teams – at least on paper – should make up any shortfall. Considering they ended up with the sixth worst power play efficiency last season, there’s room for improvement.

Prior to the Jets game in 2019-20, they’d been held off the scoreboard in 16 minutes of 5v4 power play time over four games. They’ve had the fifth lowest time with the man-advantage, yet the teams below them have scored two or more goals. They’ve fired 10 shots on goal – tied for the fewest shots in the NHL. Zone entry data would be beneficial here to further the conversation, but with only four games, there’s too small a sample to form a cogent argument.

Arizona Coyotes GF


The top unit consists Phil Kessel, Derek Stepan, Nick Schmaltz and Clayton Keller up front, with the lone defenseman Oliver Ekman-Larsson. Derek Stepan is an excellent example here because as ineffective as the Yotes power play struggles have been so far this season, they were downright dreadful in 2018-19. Stepan, had averaged a full season points total of 13 points – and a point on 65.7% of all on-ice goals scored – during a non-lockout season over his career, yet last season put up three points, one goal and two measly assists with a point on only 27% of on-ice goals. Dreadful results, and an example of just how bad special teams were last season in the desert.

Clayton Keller failed to score a single 5v4 goal in 265 minutes. Aside from Oliver Ekman-Larsson and the now departed Alex Galchenyuk, no other player topped 10 5v4 points.

The return of a healthy Nick Schmaltz (17-4-2-6) and the addition of Kessel should be making the power play more robust with added playmaking and firepower than … well, for a long, long time. This should be Arizona’s time to shine – and they’re floundering out of the gate.

This example brings us to a very important aspect of ‘regression’ – a term used often when discussing data, but with the context of a value dropping to an equilibrium point. As an example, a player that shoots the lights out carrying an above average shooting percentage is said to likely regress to the mean – or get back closer to career averages – over time. This has the same effect for underachieving players.

Stepan’s three points at 5v4 are a clear outlier in comparison to his historical work, so it’s likely that he, too, regresses closer to career averages. In this case, the ‘regression’ takes a positive swing, lifting the player’s totals back to a respectable level.

I’ll often use the term regression, but take issue with the subjective nature of the term’s use to mean only a downward movement and dismissing the opposite effect of underperformance and regressing back to career averages.


Expected Goals, Low, Medium and High Danger Shooting

You were probably expecting me to use some combination of expected goals, and/or some distance based shooting metrics. I, too, was expecting the same thing, however, the twins of Evolving-Hockey website pointed out a troubling trend in the 2019-20 NHL tracking data. The anomaly was discovered to change the ultimate location of shots by a wide enough margin to bring into question the data available publicly. This means, expected goals, shot and heat maps, and distance based scoring chances could be somewhat misrepresented in current data.

For a detailed account of the underlying issue, follow this twitter thread.

Yeah, so it looks like there is something very different about how the NHL is recording event location coordinates this season... and, umm, it's not great. Thread incoming

— EvolvingWild (@EvolvingWild) October 15, 2019


Some adjustments are required and other heavy hitting stats sites and personalities have weighed in. Micah Blake McCurdy has chimed in:

Just to corroborate what @EvolvingWild found in that thread, here's the change from 2018-2019 to 2019-2020, as a rate. Taken as a whole, the league is recording shot locations differently this season. pic.twitter.com/NkDMq3iJQc

— Micah Blake McCurdy (@IneffectiveMath) October 15, 2019

As well, Brad from Natural Stat Trick.

I've been comparing to goals from 17-18. This:

A) Appears like it could be the case
B) Doesn't seem to explain the overall difference. Most of the ones I checked had a smaller difference of ~2 feet, and in a direction that was more tangential than towards the net. https://t.co/q5Dj7jvdZl

— Natural Stat Trick (@NatStatTrick) October 15, 2019


Pointing out the tracking issues, it even looks like the NHL is investigating the possible cause.

This could be problematic and while these statistical heavy hitters have identified the problem, a solution has yet to be determined to fix the issue. For the time being, I’m skeptical about utilizing metrics that could be either outright incorrect, incomplete, or misleading and will refrain for the time being, but will be monitoring to ensure we can pick up and resume the discussion with proper metrics.


Oh wow https://t.co/ZJGnF9lB0j

— EvolvingWild (@EvolvingWild) October 16, 2019


Stay tuned.

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