Whose Results Changed the Most in the Second Half of the Season?

Using half-season RAPM to check out the NHL's six biggest risers and fallers.

Something that hockey fans intuitively recognize is that players don’t always play up to their true talent level. They go on hot and cold streaks, struggle to adapt to a new team, trail off as the season goes on, etc. This is reflected by the eye test, point production, and underlying numbers as well. However, when we talk about how a player “started hot” or “really picked it up as the season went along,” we can only back it up with raw stats and not isolated models like RAPM, because while they are regularly updated throughout the season, they are only available cumulatively. This is for good reason, as they are strengthened by the inclusion of as much data as possible, but it does create a disconnect between the ebbs and flows of player performance and the broad assessment of a season-length stat. A player who started off terribly and ended the season with an elite stretch of play would have his performance graded as average, but that doesn’t necessarily tell the story of how his season actually went. With that in mind, I thought it would be interesting to look at a few players whose isolated on-ice results took a major turn (for better or for worse) from the first half of the season to the second.

These stats come courtesy of friend of the newsletter Patrick Bacon (a.k.a. TopDownHockey), a San Jose-based data analyst. He has developed his own take on EvolvingWild’s popular RAPM stat with a few adjustments:

A player’s Regularized Adjusted Plus-Minus (RAPM) is a point estimate of their isolated offensive and defensive impact on expected goals, calculated through a weighted ridge regression which accounts for:

  • Teammates

  • Opponents

  • Score

  • Zone starts

  • Venue

  • Period

  • Schedule

My RAPM only uses 5v5 inputs, as opposed to all even strength. In addition, the expected goal inputs themselves have been flurry-adjusted and stadium-adjusted to counteract scorekeeper bias.

Patrick ran his regression three times: once using the entire 2019-20 regular season, once using only games up to December 22nd (1st half) and once using only games after that point (2nd half). That means that we can compare first and second half results for the first time.

I’ve turned those results into a visualization, which can be seen below. The graph on the left side shows how a player’s impact on offence (xGF), defence (xGA), and overall (xG+/-) differed in the first and second halves of the season; the black line in the middle is league average. The table on the right expresses the same numbers in terms of percentile rank among players at their position who played 200+ 5v5 minutes in that half of the season.

In Stone’s case, you can see that his estimated defensive impact dipped slightly, but his offence made up for it - and overall he was still one of the league’s top playdrivers. (Visualizations for all players who meet the TOI requirements are available on my Patreon, along with lots of other stuff)

In this piece, I’m going to use these stats and visualizations to highlight six skaters’ whose performance as measured by these stats changed in a significant way. For each (offence, defence, and overall), I’m going to pick one player who improved and one who dropped off.

(Note: full season RAPM is a more reliable measurement of season-length performance than half-season RAPM is of half-season performance. That’s one reason I focused on extreme cases.)

Changes In Offence

Riser: Johnny Gaudreau, CGY

xGF RAPM Change: +0.205

Recently I wrote a deep dive on Johnny Gaudreau’s disappointing 2019-20 campaign. A number of readers asked if I thought that head coach Geoff Ward (hired November 29th) had put in place a new system that had somehow reduced his effectiveness. It can be easy to to connect the dots on things like that, but based on Gaudreau’s raw numbers it seemed unlikely - in fact, they were much poorer in the early going with Peters than they were afterwards. The RAPM backs that up, as Gaudreau’s scoring chance driving sprung back to normal in the second half.

Faller: Jonathan Huberdeau, FLA

xGF/60 RAPM Change: -0.251

Jonathan Huberdeau roared out of the gate in the first half; by December 22nd he was 7th in 5v5 scoring and had sparkling underlying numbers as well. Everything was going right for the league’s most unheralded 90 point scorer. Then something clearly went horribly, horribly wrong, because his scoring chance driving fell off a cliff. Not only did his raw xGF% fall from 55% to a putrid 43%, his production plummeted too. According to the RAPM isolates, he played no small part in this.

Changes In Defence

Riser: Leon Draisaitl, EDM

xGA RAPM Change: -0.271

Two things cannot be overemphasized: how bad Leon Draisaitl was defensively in the first half of the season, and how much he improved in the second half. No forward in the league was on the ice for more scoring chances and goals against than Drai up until December 22nd, but the Hart Trophy winner turned a corner in a huge way in 2020. While he dynamic duo of #97 and #29 put up serious points, its shameless run-and-gun style constantly left the Oilers’ defence and goalies out to dry. But once Draisaitl was moved back to centre with Ryan Nugent-Hopkins and Kailer Yamamoto, his play completely transformed, and he was actually above-average defensively as the anchor of a more possession-oriented line that was solid at both ends of the ice. This doesn’t erase what happened in the first three months of the season, but it is an encouraging sign moving forward and an indication that he belongs permanently at centre (and away from McDavid).

Faller: Quinn Hughes, VAN

xGA RAPM Change: +0.200

Quinn Hughes’ rookie season was beyond impressive: he emerged as one of the league’s premier transition players, led all rookies in points, and was pretty easily the best blueliner on a team that made it to the second round of the playoffs. He entered the league with a rare confidence with the puck that led to great outcomes at both ends of the ice. Considering all of this, many Canucks fans were bewildered when some analytics writers (like me) didn’t give him as much Calder love as they expected. The reason is what happened in the second half, when the wheels fell off defensively for him and his partner Chris Tanev. RAPM spreads the blame pretty evenly between the two of them, and it’s not fully clear what the cause was: did Tanev’s game collapse and Hughes didn’t adapt? Did Hughes start taking more risks? Did the rookie just run out of gas, as many have before him? Either way, I think there’s little to worry about here, and the Calder runner-up is poised to be one of the league’s best for a long time.

Changes Overall

Riser: Rasmus Dahlin, BUF

xG +/- RAPM Change: +0.299

To say that Rasmus Dahlin didn’t initially take to coach Ralph Krueger’s system would be an understatement. After putting up arguably the best D+1 season by a defenceman in the cap era, Dahlin performed hideously in the first half, ranking near the bottom of the league at both ends of the ice. What happened? Chad DeDominicis of Expecting Buffalo explains:

Krueger tried to get him to focus on his defensive side of the game, admitting it was going to negatively impact his offense, and it did. He got in his own head and struggled with turnovers by overthinking.

Krueger’s plan backfired, as deliberately suffocating Dahlin’s offence made his defensive play even worse. After recovering from a concussion, Dahlin returned to his rookie season style and form for the most part, carrying the puck more frequently and participating more offensively. This seemed to have improved his actual defence a lot more than forcing him to sacrifice the strengths of his game did. Who would have thought?

Faller: Ben Chiarot, MTL

xG +/- RAPM Change: -0.341

To put it gently, Ben Chiarot was not the analytics community’s favourite player when he was a member of the Jets. So when he started his tenure with the Canadiens playing the best hockey of his career, Habs fans justifiably took a victory lap around the nerds who had told them they were getting Alzner 2.0. Then the second half of the season happened. Chiarot kept blocking a lot of shots (1st on the team) and hitting a lot of guys (2nd), but the bottom completely fell out of his on-ice results. He was pretty evidently a hindrance on his partner Shea Weber, who had much better results without him whether you look at shots, scoring chances, or that super fancy stat “goals against“. In other words, he turned from a real contributor back into a guy who “brings physicality” and “plays tough minutes” This extended into the playoffs, when he did not look good on the ice or the stat sheet; of the Habs’ defenceman, he had by far the worst rate of on-ice goals against, shots against, and scoring chances against. It seems likely that Chiarot made an excellent first impression and then reverted back to his usual self in the second half. It will be interesting to see which player the Habs get next year - I’ll admit I have a hunch.