So, about those goalies: What would have mattered

A conversation with Tina Robinson the other day led me to do some thinking about the season and to what extent the goaltending may have been the weak link in the Lightning's game this year. It certainly is easy to assume that if the goaltending would have been better, the season wouldn't have gone the way it has.

The problem is that it's really, really hard to disentangle the performance of a goalie from the performance of his team. In fact, that's something folks argue about a lot. Sometimes, for every person who blames the goaltender for that "soft" goal, another person sees a defensive breakdown. The fact of the matter is that hockey is a complex system, and we don't really have good ways to talk about how goals get made, much less how winning gets made.*

So I wanted to know whether we could somehow control for goaltender performance--or conversely, control for shot performance--and determine what the effects of the goaltending were this season, and whether it really was as simple as "change the goalie, save the season."

It wasn't.

The process:

The Lightning have 75 points in the standings with 74 games played. Their current record is 34-33-7. Their season save percentage is .893 and their season shot ratio [SF/SF+SA] is .476.

To tease out the differences that goaltending and puck possession made, I first changed the save percentage while keeping everything else the same (shots and goals for) in each game played. This allowed me to see in what games an increased save percentage would have changed the outcome. The shot ratio experiment was slightly more complicated. I assumed the same number of total shots per game and the same save and shooting percentages per game. I changed only the proportion of shots for and against. [Total shots * new shot ratio = new shots for and new shots against.] That allowed me to calculate how goals for and goals against would have changed.

For every new game scenario, I then calculated the goal differential on a per-game basis. If the game's goal difference rounded to at least 1, it was counted as a win. If the difference rounded to -1 or below it was counted a loss. Anything in between was counted as "too close to call." Close games could go either way, but for the purposes of teasing out goalie performance levels, it doesn't matter.

In determining standings points, I counted 2 for a win, 0 for a loss, and 1 for a close, on the assumption that it would have gone to overtime because neither team was ahead by at least 1 goal. (In fact, since these games all include overtime without distinction, we can say that since there's no clear lead, they essentially go to a shootout.)

What if...

A. The goalies had performed at their career averages?

I calculated Dwayne Roloson & Mathieu Garon's career SV% excluding this year (to be fair). Garon's is .904 and Roli's is .910. Roli has played a lot more than Garon over the years, though. Their combined career SV% excluding this season is .908 (21009 saves on 23126 shots).

If they had played the whole season at .908, the Lightning would have been worse off. The team's record drops to 28-33-13 and they would have had 69 points in the standings. This is because we're trading all of the very good games for average ones.

B. The goalies had performed at league average (.914, right now)?

Again, trading all very bad games and very good games for league average hurts the Lightning. They end up at 30-32-12, with 72 points.

C. The team had very, very good to elite goaltending (.920)?

This is the only way that goaltending alone makes a difference across the season. The Bolts end up at 32-27-15, with 77 points. Still not quite in the playoff race, though.

D. The team had minimal puck possession?

Taking only 50% of the shots in each game (an increase of 2.4%) led to a record of 31-33-10 and 72 points.

E. The team was pretty good at puck possession?

A shot ratio of .510 led to 33-33-8 and 74 points, whereas a shot ratio of .520 led to a record of 34-32-8 and 76 points.

What does all of this mean?

Well a few conclusions can be drawn. First, neither goaltending alone nor puck possession alone would have really cured the Lightning's ills and saved the season. The season was as bad as it was because both of those things failed at the same time.

Second, the goalies had more good games than it seems like they did. There were enough games where the goalies performed better than either their career averages or the league average to make a difference in how the season turned out.

Looking at how the averages affected the outcomes of games helps to isolate those games where save percentage made an important difference. With the lower two save percentages, a handful of losses became outright wins and another handful became close, but at the same time, this change highlighted the fact that at least as many wins were affected by the changed save percentage as losses. Only if the goaltending had been elite (above .920) all season would goaltending alone have been enough to save the season.

@ .908 @ .914 @ .920
Losses > Wins 4 5 7
Losses > Close 5 6 8
Wins > Losses 5 4 2
Wins > Close 7 6 6

Third, puck possession also began to show changes only when somewhat above league averages. The league average for shot ratio is .500 on the dot because every shot for is also a shot against. That average doesn't really tell us much. But the average for the Eastern Conference is .499, and the average for the Eastern Conference playoff teams is .510. Neither one of those figures made a significant change in the Lightning's record. Not until puck possession was .520 did the Lightning gain standings points.

@ .500 @ .510 @ .520
Losses > Wins 2 4 4
Losses > Close 6 4 4
Wins > Losses 1 1 0
Wins > Close 4 4 4

Again only a handful of games changed outright, not nearly enough to be considered saving the season. [Note that the very best puck possession teams, like the Detroit Red Wings, take about 55-60% of the shots in their games.]

One final experiment. What if both puck possession and goaltending were average? That, my friends, made a huge difference. With an average shot ratio of .500 and an average save percentage of .914--while assuming the same number of shots and the same shooting percentage--the Lightning's record becomes 34-24-16, giving them 84 points and puts them firmly in the race for that last playoff spot.

Of course all of this is speculative and assumes quite a bit. However, it does help to isolate exactly how many games were changed directly by either poor goaltending or poor puck possession. The answer is not that many. The weak link wasn't either the goaltending or the defense. It was both.

*Yes we know that shooting more than your opponent correlates more strongly with winning than any other measurable on-ice tactic and that it's one of the few correlations that's in any real way repeatable, but we also know that "luck" or randomness or whatever you want to call it accounts for a very, very large part of winning, too.