To be clear, I'm arguing that the shot data you posted from Deathstroke does not prove what it claims to prove and that everyone should reject the idea that there is a universal rule in hockey that fewer shots against necessarily means a lower save percentage and higher shots against means a higher save percentage. At various times in league history there has been some sort of observed relationship between shots against and save percentage, but it has mostly been negative (i.e. fewer shots against means a higher save percentage, more shots against means a lower save percentage). And even in periods where we see some sort of correlation, there is still plenty of individual variation between teams.
Just for completeness' sake, in case anybody cares (and also since I haven't seen a full rebuttal on this subforum yet to the incorrect idea that save percentage is significantly positively impacted by shots faced), there are two other factors that also help explain the high-shot/low-shot games observance: Scorer bias and score effects.
Shot counting is subjective, so for every NHL game there is an officially recorded shot total and a "true" shot total, i.e. the shot total that would be recorded if some superhumanly consistent shot counter rated every game by the exact same standard. Assuming a basic normal distribution, simple Bayesian reasoning would conclude that the farther we get away from average, the more likely it is that the "true" shot count diverges from the recorded one.
For example, everyone knows about the six shot game Toronto had in the playoffs against New Jersey in 2000. Toronto players at the time argued at the time that they had a lot more shots than that (Garry Valk was quoted as saying they had
"13 or 14"). It is entirely reasonable to expect that there was something weird going on with that official scorer since six is such an extreme outlier given what we know about single game shot distributions in the NHL. The same logic applies to a 55 shot game, except in the other direction. Maybe it was only a "true" 52 shot game, but the goalie got credit for three extra stops during the many flurries around his crease that would have happened in such a game. In most games scorer bias is not a big deal, but for outliers it can absolutely have an impact, especially if you are looking at the ends of the distribution, particularly the very low shot games where the effect works to reduce recorded save percentages. The best way to deal with scorer bias though is not to focus on shots against bins, but to screen it out in other ways by analyzing all shots and saves for each rink separately and making appropriate adjustments to the numbers from each location, which will deal with the problem automatically.
Teams also play to the score, which means that teams in the lead tend to take fewer shots and teams that are behind tend to take more. Over the past three seasons, according to Natural Stat Trick, teams averaged 33.3 SF/60 when trailing and 27.2 SF/60 when leading in all situations. For the sake of comparison, the top team in the league (Pittsburgh) was at 33.3 SF/60 over that period and the bottom team (New Jersey) was at 27.6, showing that the score-based spread was actually greater than the observed spread based on team talent. What this means is that when a goalie is hot and his team is winning, he will tend to face even more shots than normal, while when a goalie starts poorly and his team falls behind, he might not get to face as many shots over the rest of the game to even out his performance. In other words, sometimes a goalie faces 30+ shots
because he has a high save percentage, rather than the other way around.
You might expect that goalies who faced 20 shots or fewer over a full game must have had a much better win/loss record than goalies who faced 40+, but that's not actually the case. Over the same past three seasons, if you look at goalies who played at least 55 minutes but faced
20 shots or fewer, their collective win/loss record is just 204-152-21 (.569). For the
40+ shots against group, the collective record is 340-153-130 (.650). That's score effects in action, and therefore it shouldn't be surprising to see that one bin has better results than another save percentage-wise given the difference in the overall win percentage.
The last point of evidence in my argument that this is merely a case of cherry-picking a biased sample is that the high-shot/low-shot relationship appears to be true for every goalie in history. There are few things that are true for every goalie in history, because there has been a huge variance in terms of team systems and defensive abilities over the years.
There have been many great defensive teams that allowed relatively few shots against (many different editions of the Montreal Canadiens come to mind), but there have also been great defensive teams that allowed a surprisingly high number of shots against (the 1960s Toronto Maple Leafs, for example, or the 1980s New York Islanders). There have been low-shot teams with great save percentages and low-shot teams with mediocre save percentages, basically every combination possible. This is a big part of the reason why we don't see obvious patterns in the data, and why each goalie/team case needs to be assessed on an individual basis when trying to determine how much of an impact the team had defensively. If something is true for every goalie in history, then it's pretty clearly something that falls in the same category as the example of goalies always playing better in wins than they do in losses, i.e. merely the case of a biased subsample that doesn't give any particularly useful information when you look at aggregate results.
I'm just pullin' your chain of course, k dog...
Chapeaux!
