clock menu more-arrow no yes

Filed under:

Turning the eye test into a number

New, comments

A look at where the stats and the eyes agree or disagree.

Philadelphia Flyers v New Jersey Devils Photo by Andy Marlin/NHLI via Getty Images

Whether it’s wide-scale team evaluation or at the player level, hockey analytics have really begun to claim their stake in the game over the last decade. Teams are continuously growing more accustomed to the use of data, and are willing to use any and all information available to them. The hockey analytics wars are long gone, even if the egg in your Twitter mentions suggests otherwise; they’re here to stay.

Something that is also here to stay in hockey evaluation is the eye test. Analytics aren’t here to replace old-school analysis, rather inform and enhance it. When used incorrectly, the numbers can be misleading. They can be cherry-picked, and void of context. Further, in-zone defense remains tough to capture, as well as the true impact of intangibles.

On the flip side, the thing with the eye test is that it’s extremely subjective. Everyone values things differently, and so differently at times that an action held in high regard by one, can be seen as a complete non-factor by another. Luck also plays a big role, with a bad play that leads to a goal against being focused on, and remembered much more than an even worse play that doesn’t lead to a goal against.

The reverse is true as well, with Ivan Provorov’s perceived impact on the Flyers’ power play last season (and through the early stage of this season) serving as a recent example. Those seven power-play goals in 2019-20, the most among NHL defensemen, look even more misleading now than they did back then. Overall, the eyes and the numbers complement each other well, supporting the other’s weaknesses, and more often than not coming to similar conclusions.

With that in mind, I wanted to see how much my own eye test agreed or disagreed with the stats. The first step was setting a few ground rules to keep me in check.

The rules were as follows:

  1. After each game, rate each player with a score of 1-100 based on how they impacted the game. A one being the worst imaginable performance possible, a 100 being the best imaginable performance possible, and a rating of 50 meaning that the player had little to no impact on the game positively or negatively.
  2. Rate the players before going out of your way to look at any numbers from the game, to keep the influence of the numbers at a minimum. (Ideally, there would be a way to avoid any influence entirely, but it is close to impossible to remain oblivious to things like point totals while watching a game.)
  3. No grading on a curve. Expectations don’t matter, only how their actions impacted the game.

Simple enough. Onto the numbers.

Average Ratings

Player Games Played Rating
Player Games Played Rating
Claude Giroux 54 54.02
Sean Couturier 45 52.72
Wade Allison 14 51.43
James van Riemsdyk 56 50.09
Shayne Gostisbehere 41 49.17
Joel Farabee 55 48.39
Ivan Provorov 56 47.82
Travis Sanheim 55 47.61
Travis Konecny 50 47.57
Oskar Lindblom 50 46.9
Michael Raffl 34 46.76
Jakub Voracek 53 46.75
Justin Braun 53 46.1
Scott Laughton 53 45.87
Nic Aube-Kubel 50 45.16
Kevin Hayes 55 45.06
Phillippe Myers 44 44.51
Robert Hagg 34 44.18
Connor Bunnaman 18 43.06
Nolan Patrick 52 42.21
Samuel Morin 16 41.05
Erik Gustafsson 24 40.65

*Only skaters that appeared in at least 10 games included; Samuel Morin’s games at forward have been removed from the dataset.

A couple of things stand out. First, there’s not a big spread in the ratings. The vast majority fall between 45 and 50. Second, by the ratings, only four Flyers on average positively impacted the team’s performance this season. Harsh, but given how the year went, not too surprising.

At the top are who you would expect to be there in Claude Giroux and Sean Couturier. Then there’s Wade Allison, who at this point might just be the fanbase’s most beloved player. Rounding out the top four is James van Riemsdyk, whose two-way game really seemed to turn a corner last season and who wound up in a three-way tie in the Flyers’ points race by season’s end. There’s clear agreeance between the eyes and numbers here, with my top four among the top five Flyers in Expected Goals-For percentage (xGF%), as well as the top eight in Corsi-For percentage (CF%) at 5-on-5. However, the final player needed to make up the top five won’t be found at the top of the ratings.

It just doesn’t add up

Out of 22 skaters, Kevin Hayes comes in at no. 16, ahead of just two forwards; Connor Bunnaman and Nolan Patrick. Yet Hayes was one of the team’s better play drivers this past season, produced more even-strength offense than he did a year ago on a per-minute basis, and was on pace for a 46-point season under a regular 82-game schedule. The discrepancy between the eye test and the numbers on Hayes’ season is the biggest in the entire experiment.

There are a couple of things at play here. First, we can’t ignore the new, shiny toy effect wearing off as Hayes’ second season in Philadelphia began. Hayes was always going to be more scrutinized in year two than he had been in year one, but this was exasperated by his big-play tendencies disappearing. His impact on the penalty kill was a big reason behind its’ resurgence last season, maybe the biggest, and this past season it felt like the Flyers no longer had that player. At even strength and on the power play, far too often it felt like Hayes wasn’t as engaged as he had been a year ago.

Part of this can be attributed to playing through injury, as Hayes will reportedly undergo core muscle surgery over the off-season, but it was a legitimately poor season from Hayes. One that the Flyers cannot afford for him to repeat.

After Hayes, the next biggest disagreement comes from Robert Hagg’s numbers. Yes, the defenseman largely viewed as an advanced stats black hole actually graded out positively in 2021. The on-ice metrics essentially view Hagg as the team’s third-best defenseman, at worst, this past season. The eye test ratings have him sixth, behind every other regular defender. Perhaps Hagg really was OK in limited minutes this season, and a lack of standout plays combined with implicit bias sunk his rating. Or these 34 games were a one-off driven by excellent numbers when paired with Travis Sanheim, his most common partner, and surprisingly good results alongside Shayne Gostisbehere — a pairing that hadn’t fared well by the numbers in the past.

Either way, Hagg managed to come out of the 2021 season with the best underlying numbers of his career.

Finally, there’s Phil Myers. After an inconsistent start to the season, the scratchings began and his play only worsened from there. There were real breakdowns in defensive zone coverage that led to his poor rating via the eye test, but it can’t be ignored that Myers’ on-ice save percentage at 5-on-5 was one of the worst in the league at 87.27 this year. His defensive shortcomings this year contributed to that figure, but we also have to keep in mind that, among skaters that played at least 250 minutes, Flyers players made up half of the bottom ten in on-ice save percentage. At least some portion of any issues you might have had with Myers’ season goes back to goaltending.

The shot metrics lend credence to that notion, with passing grades in both CF% (52.23%) and xGF% (50.74%). Still, that doesn’t excuse his defensive miscues or the number of times that he was burned off the rush this season. He did show signs of improvement later on, and because of this, his eye test rating was trending up once more by the end of the season. From an optimistic angle, these can be seen as reasons to believe that the Myers we saw for a large chunk of the 2021 season won’t be the Myers that we see next year.

Sanheim’s rollercoaster season

One of the more interesting ratings to take a deeper look at is Sanheim’s, especially compared to the rest of the defense. Through the first 20 or so games, Sanheim had been the Flyers’ best defenseman by the numbers and had been passing the eye test with flying colors. But as March rolled around, Sanheim’s play fell off dramatically right alongside the Flyers’ playoff chances.

Like Myers, Sanheim’s mistakes were loud and costly. Both of their ratings had a similar mid-season plummet, but Myers’ drop-off, as well as his recovery, happened a bit more gradually than Sanheim’s. Despite a disastrous stretch, it’s not hard to see why Sanheim joined Provorov and Justin Braun as the only Philadelphia defenders to not be healthy scratched this season. His lows were quite low, but the full picture points towards Sanheim being the Flyers’ second-best defenseman this year; maybe even their best if you’re not as high on Gostisbehere’s 2021 campaign.

After a tough start, Braun had a solid enough season the rest of the way and should be in the “best defenseman” conversation as well. Provorov was maybe the most consistent of the bunch, without many peaks or valleys.

Closing thoughts

Overall, the results were about what I expected to find. A small handful of players were seen in a better light by either the eye test or the stats, but largely the two came to the same conclusion. Even when it came to one of the Flyers' leading points-scorers, Jakub Voracek, both the eye test and underlying numbers agree that his raw point total isn’t a totally accurate representation of his season on the whole.

This is how it should be; the two working together to come to similar conclusions. It won’t be that way all the time, or with every player, but the disagreements should be few and far between.

Statistics referenced courtesy of Evolving-Hockey