Not all Primary Assists are Created Equal

We often use assists as an indicator of playmaking ability and there is often a preference for primary assists since these are the ones that directly lead to the shot that led to the goal. However, not all primary assists are created equal if we want to use them to measure playmaking ability.

Over the previous 3 seasons Brady Tkachuk has 69 primary assists, or 0.328 primary assists per game played. Drake Batherson has 56 primary assists or 0.304 primary assists per game played. These per game numbers are relatively close so one would assume that they are similarly good passers/playmakers.

However, not all primary assists come from passes and thus not all primary assists are an indicator of passing or playmaking ability. For the purpose of this article I will state that there are two types of primary assists, those from a pass and those from a shot attempt. This is where the Tkachuk and Batherson worlds diverge.

For Brady Tkachuk, 49.3% of his primary assists came on plays where he took a shot in the 3 seconds prior to the goal being scored indicating the sequence was Tkachuk shooting the puck, a teammate collecting the rebound and scoring a goal or alternatively Tkachuk shooting the puck a teammate collecting a loose puck after a blocked or missed shot and then scoring a goal. In these scenarios there is no passing or playmaking on the part of Tkachuk but he would get credited with a primary assist because he was the last teammate to touch the puck.

For Drake Batherson these primary assists from shots accounted for just 16.1% of his primary assists. If we pull out these shot based primary assists and just look at the pass based primary assists Batherson would have 47 such assists or 0.255 per game while Tkachuk would have 35 such assists or 0.161 per game. Tkachuk and Batherson are not looking like equally talented playmakers now.

Some of the top players for percentage of primary assists that came after their own shot are Zach Hyman (67.3%), Logan Couture (54.0%), David Pastrnak (53.2%), Auston Matthews (52.9%) and Timo Meier (51.0%). At the other end of the spectrum are players such as Robert Thomas (9.7%), Alex Wennberg (12.5%), Pavel Zacha (13.2%), Ryan Strome (16.67%) and Nicklas Backstrom (19.4%).

Over the previous 3 seasons the top players in Primary Assists via the pass (excluding likely rebound assists) are McDavid (98), Draisaitl (77), Marner (75), Gaudreau (70), Panarin (68) and Huberdeau (67). These all pass the smell test when we think of the top playmakers in the league.

(Note that all data in this article is all situation data, not just 5v5)

So, why does this matter? Well, several weeks ago on Twitter I commented that a line of Tkachuk, Norris and Tarasenko that the Senators were trying lacked a true playmaker. I consider Tkachuk, Norris and Tarasenko all more shooters than passers. I used a fairly rudimentary statistic of the ratio of goals to assists as an indicator of whether the player is more of a shooter or more of a passer and it supported my belief. I quickly got a response from Micah Blake McCurdy who took issue with my rudimentary statistic and suggested I use his Setting metric instead and suggested that Tarasenko and Tkachuk are both good ‘Setters’ or playmakers. McCurdy’s model had Tkachuk and Batherson as equally good Setters.

You can read about McCurdy’s model including a description of his inputs into his model for calculating Setting ability. Essentially it is looking at primary shot assists. Where shots that are goals he uses the player the NHL recorded as the primary assister and for shots that didn’t result in goals he attributed percentages of likelihood the player had the primary shot assist to players on the ice at the time the shot was taken with adjustments based on their position (forwards were assigned a greater likelihood of being the primary assister).

However, all shots are included in his model including shots after rebounds where there was no pass immediately prior to the shot being taken and thus no ‘setting’ taking place. McCurdy’s Setting metric is likely a good predictor of primary assists, however because some players get a significant portion of their primary assists from shots not passes made I don’t believe his Setting metric is a good indicator of passing/playmaking ability. An easy fix would be to remove rebound from his Setting model.

A second (probably lesser) concern I have about McCurdy’s ‘Setter’ model is he assigns equal probability that the forwards on the ice got the primary assist and an equal probability to the defensemen on the ice that they got the primary assist. If Ovechkin took the shot it is far more likely that the passer was Nicklas Backstrom or maybe John Carlson than it was Tom Wilson or Martin Fehervary. I am not sure how much this will impact results for the majority of players but I have to think for some players at the extremes it will have an impact.

Long story short though, Brady Tkachuk is a great player, but I stand by my position that he is not a passer/playmaker.