Good question from @LLysowski on the 26/23 pair how Don balances eyes vs the numbers.

Don’s response on the analytics:

“Analytics are tough b/c there are so many analytics. There are so many variables that you can question some of the integrity in some of the analytics” (1/?) https://t.co/JGXZS5ktvl

— Chad DeDominicis (@CMDeDominicis) December 28, 2023

…I mean what else do you want us to write about?

Do I think that quote is deserving of a negative reaction? No. Do I think that quote encompasses “how analytics works”? Also no.

Starting With Ogres

Analytics are layered, just like how in calculus the derivative is the rate of change of the primary function. To win games, you need to score more goals than the other team. To score goals, you need to maximize the quality of shots your team generates and minimize the quality of shots your team allows. Part of that optimization of quality is an optimization of quality. You need to possess the puck in your own zone to generate shots. To get the puck to your own zone you need to enter the zone successfully from the neutral zone. You also need to get the puck out of your own zone.

Have I covered everything? This almost feels like a broken record but it’s worth re-iterating because it helps get a major point of data-driven decision making across. Each layer’s “why” can drill down to a lower level. At least now the next time we have to have these conversations, there’s a sign to tap.

So in this context, the base numbers for the Rasmus Dahlin and Mattias Samuelsson pairing are that in the 351:26 of 5v5 play that they have been on the ice, they have been outscored 17-10 for a 37.04% GF%. First question: Is this indicative of the shot quantity and quality they have been giving up? Not necessarily. A 49.63% share of the shot attempts and 48.98% of the expected goals hovers closer to 50/50.

So we ask why again. The sample size is small, but three months of hockey can’t be “just” luck-related. Plus when you need to win hockey games, there’s really no throwing your hands up in the air and saying “Oh Well it’ll turn”.

The next level would be the micro stats, which Chad looked at earlier this month, particularly in the context of zone exits, so I won’t be repetitive, but fortunately, the private data would be available for that in our hypotheticals.

The context of the pair is interesting as well, as the other pairs appear to be better for both Dahlin and Samuelsson compared to playing together. Naturally, our next question is “Then why are they still being paired together?” The only thing left to do is watch the film, but without the true context of their systems, the analysis drops off compared to what the staff looks at. Film is an important part of analysis, especially at a coaching level — the issue is when film watching is not assisted with the records of what is being shown (which at the most literal level is the data).

Which brings us to the classic statistical answer: I don’t know. My opinion is that I think the organization expects their players to turn it around and have something click at some point and they still believe in 26-23 as a top pairing with this current roster. And until we get the Sabres playbook, we don’t know for sure what those film observations are.