[text_output]Welcome to the first edition in what will be a series of articles discussing some of the more complex advanced statistics and analytical concepts housed within the Blueshirts Breakaway Hockey Lexicon. For those who have yet to check out my Lexicon, it is a comprehensive resource that defines and contextualizes all major advanced statistics and concepts, and it also provides usage examples for some more important metrics, and provides recommendations for resources to use and people to follow if you wish to learn more.

In this inaugural post to the article series, I want to introduce a relatively new concept that even many analytically-inclined hockey fans may be unaware of: teammate relative statistics. Teammate relative statistics (abbreviated Rel TM) go a step beyond team relative metrics and attempts to further isolate a player’s performance by benchmarking his numbers against all of his individual teammates, instead of against his entire team in aggregate. Teammate relative statistics accomplish this by combining principles used in calculating team relative statistics and WOWY analysis (discussed in the next section). Teammate relative statistics were initially made publicly available by David Johnson, who ran the popular hockeyanalysis.com site, before he was hired by the Calgary Flames in the summer of 2017.

The key in calculating relative teammate statistics is by including the player’s on-ice performance as well as the average of all of his individual teammates’ on-ice performance when the player we are analyzing is not on the ice. So, for example, if we are discussing Corsi for per-60 (CF/60), the calculation would take the total on-ice CF/60 of the player we are analyzing, but also subtract the average of all is individual teammates’ on-ice CF/60s without the player on the ice. Another key point to the calculation is that the teammates’ portion of the calculation is weighted by individual teammate time on ice percentage (TOI%) with the player being analyzed. Each individual teammate is assigned a weight relative to their TOI% with the player being analyzed in order to properly account for how much of a potential impact that teammate may have on the player.[/text_output][image type=”thumbnail” float=”none” src=”2320″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””][text_output]Noted hockey statistician and Hockey-Graphs contributor Luke Solberg (better known as EvolvingWild on Twitter) recently wrote a fantastic two-part article series for Hockey-Graphs that discusses the various relative shot metrics (team, teammate and WOWY) and highlights the pros and cons of each of the metrics. The piece also includes a highly detailed explanation of his relative teammate calculation, and it breaks down all of the initial components to highlight exactly how the stats are formulated.

In the piece, Luke notes that the biggest issue with both team relative stats and standard teammate relative stats (such as the “RelT” data available on Corsica), simply put, is that it is difficult to isolate the performance of a player who is often deployed with the same teammate(s).

For example, for much of Ryan McDonagh’s tenure with the Rangers, he was paired with Dan Girardi; thus their team relative metrics are both greatly impacted by the fact that they are so frequently on the ice together. In very large sample sizes, this issue isn’t as problematic, but in small sample sizes (which includes a full season), this can be a huge issue when evaluating a player’s performance.

An additional issue pointed out by Luke is that team strength has an impact on both forms of relative statistics. He notes that, “players on the worst teams appear better and players on the best teams appear worse relative to the league.”[/text_output][image type=”circle” float=”none” src=”2322″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””][text_output]In order to account for the impact that both of these issues have on relative teammate statistics, Luke created adjustments that he discusses at length in his article. This isn’t the place to get into the nitty gritty calculations of how specifically he made the adjustment; if you wish to know the mathematical specifics, please check out Luke’s Hockey-Graphs article. The important thing to note however, is that Luke explicitly states that he believes his adjustments “do a very good job dealing with the innate problems the Rel TM method poses,” and he provides ample evidence to back up this claim.

The point in me explaining all of Luke’s calculations, weightings and adjustments that are involved in his relative teammate statistical model, is to lay the foundation for this claim, which Luke made in the conclusion of his piece: “for every type of long-term player evaluation, I feel the adjusted Rel TM method is vastly superior to the Rel Team method.” He also states that, “In general, I feel the Rel TM method – when adjusted for its inherent issues – is one of the best single-number “pen and paper” methods we have at our disposal for player evaluation.”

In other words, relative teammate statistics serve as a more reliable way to isolate and analyze single player performance than WOWY analysis or relative team analysis. Now, relative team and WOWY analysis certainly still are valuable evaluation techniques in their own right, but the new adjusted relative teammate versions serve as better long-term analysis techniques for individual players.

One good example of how we can use Luke’s adjusted Rel TM model is by comparing J.T. Miller to Vladislav Namestnikov, both of whom were involved in a blockbuster deal between the Rangers and the Lightning at the 2018 trade deadline. A point that many fans brought up with regards to comparing the two players, was the fact that Namestnikov may have been a benefactor of playing on a better team (Tampa Bay) and on a far superior line to any that J.T. Miller ever played on. During the 2017-2018 season, Namestnikov’s most common linemates were MVP candidates Steven Stamkos and Nikita Kucherov, which obviously had a positive impact on Namestnikov’s point production. Conversely, J.T. Miller had a rotating cast of linemates throughout the season, with his most common partners being Mats Zuccarello and Michael Grabner; not bad, but a far cry from Kucherov and Stamkos.[/text_output][image type=”circle” float=”none” link=”true” target=”blank” info=”tooltip” info_place=”bottom” info_trigger=”hover” src=”2323″ alt=”Photo Credit: The Canadian Press via AP / Darryl Dyck” href=”photo credit” title=”Photo Credit: The Canadian Press via AP / Darryl Dyck” info_content=”Photo Credit: The Canadian Press via AP / Darryl Dyck” lightbox_caption=”” id=”” class=”aligncenter” style=””][text_output]We can use Luke’s data, which he made publicly available here, to gain a better understanding of the impact that J.T. Miller and Vladislav Namestnikov had relative to their respective teammates. Of all the data shared in the Google Drive doc linked above, what stands out the most to me is what Luke refers to as “relative teammate total impact” statistics. I urge you to read Luke’s Hockey-Graphs piece for a full breakdown, but in layman’s terms, the statistics attempt to measure a player’s total impact on his team, relative to his teammates, for a respective metric (such as Corsi or expected goals).

The impact statistics take into account all 5v5 data and include the adjustments we discussed above. They work similarly to any differential number (plus-minus style), which is important to note because players with more ice time can have a greater variance in their numbers. As much as we all like to isolate per-60-minute production, total impact statistics are also vitally important, because a player who can produce at a high level while receiving high usage is obviously more valuable and has a larger team impact than a player who produces at a high level but receives minimal ice time. After reaching out to Luke on Twitter, and confirming I was interpreting these statistics correctly, he added that, ““I would probably describe that [an impact stat] as an *estimated* net total contribution.” So, for example, if a player has an adjusted relative teammate expected goal total impact of 15, that means that his estimated net total contribution to the team’s expected goals total is +15 over the course of the season.

Now, let’s finally look at the data for Miller and Namestnikov, shall we? As of the most recent update Luke made to his data (March 2, 2018), Vladislav Namestnikov has an adjusted expected goal total impact of 0.9, while J.T. Miller sits at 0.4. As a reminder, these total impact statistics are differentials (plus/minus, with 0 equating to no impact positive or negative), so these numbers by Namestnikov and Miller indicate that both have had a slightly positive impact on their teams’ expected goal differential while they are on the ice, relative to their teammates. In terms of the NHL ranks in the statistic, Namestnikov is 275th while Miller is 299th out of 600. For context, Dougie Hamilton leads the NHL at 16.1, and Brooks Oprik is the worst in the NHL at -13.4. It is worth noting that this data uses Luke’s proprietary expected goal model, and not the Corsica version (which is Manny Perry’s model), which is likely the most common expected goal model you see referenced.

The Corsi impact numbers more strongly favor Namestnikov to Miller, as Namestnikov has an adjusted Corsi total impact of 43.8, while J.T. Miller sports a -33.9. Using this data, we can conclude that Namestnikov has had a positive impact on his team’s shot attempt share rate while he is on the ice, relative to his teammates, while J.T. Miller has had a negative impact relative to his teammates. Namestnikov ranks 204th out of 600 in terms of Corsi total impact, while Miller is 425th, so roughly one-third of the qualified players fall between the two. For context, Dougie Hamilton also leads the NHL in this statistic with a 337.2, while Justin Braun brings up the rear at -348.4.[/text_output][image type=”circle” float=”none” link=”true” target=”blank” info=”tooltip” info_place=”bottom” info_trigger=”hover” src=”2326″ alt=”Photo Credit: Anne-Marie Sorvin-USA TODAY Sports” href=”photo credit” title=”Photo Credit: Anne-Marie Sorvin-USA TODAY Sports” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””][text_output]Finally, we get to the conclusion: what does all of this actually mean, and how does it help us with the J.T. Miller versus Vladislav Namestnikov debate that was prevalent on social media after the trade? Using this relative to teammate data, we can reasonably state that even when one accounts for the fact that Namestnikov was playing on a far superior team with far superior linemates than J.T. Miller, he still impacted the game in a positive way more than J.T. Miller did, at least in terms of expected goal and shot attempt differentials.

Now, before some of you get up in arms about this claim, I want to make it very clear that I am not sitting here and trying to tell you all that all you need is Luke’s new teammate relative statistics model to determine which players are better. What I am saying however, is that this adjusted relative to teammate model of Luke’s is a new valuable tool that we can use to help round out our analysis of the two players in this discussion, which alongside with traditional scouting/the eye test and all of the other data we have available to us, can help us better analyze the performance of each and the impact they have on their teams.[/text_output]

Author: Drew Way

Diehard New York Rangers fan since 1988! Always has been fascinated by sports statistics, and is a big proponent of supplementing analytics with the eye test. Also a big Yankees, Giants and Knicks fan.