Timely Saves exist: But it’s not being evaluated correctly

The Winnipeg Jets have won 67 percent of games with Ondrej Pavelec in net despite the netminder only stoping a paltry 90.1 percent of shots on net. This has brought up some interesting conversations on timely saves and “just win” over online forums.

There are quite a few logical holes behind these ideas and I’m going to point them out.

Let’s take a step to the side and talk about not caring how a team wins.

It is not necessarily bad for a fan to care more about the end result than the reasons behind the result. Sports fans are emotionally invested individuals who aspire for their team to conquer the opponent to a “fanatical” extreme. Fans can subjectively like or care about whatever they please.

Some of us do care about how a team wins though.

For me, my job is to critique and improve decision making of the decision makers in hockey. How a team wins matters. Let us say some gameshow contains three options where one gives you a sixty-percent chance at winning a prize while the others each garners you a twenty-percent chance. The correct process that will allow you to win most often is choosing the sixty-percent option, even though you may still win with the other option. Selecting one of the twenty percent options and winning still means you used the wrong process.

This is where data analysis in hockey comes into play.

There will always be unknown and intangible parts of the game, but there are also known and quantifiable parts as well. In a game where the winner is literally the team with the best stats, or most goals, wins, the quantifiable parts play the driving role in what makes a team win more often than they lose.

This season the Jets won more often than they lost with Pavelec in net, but all evidence points in the same direction: the Jets do not win because of Pavelec being in net.

The team has not performed better defensively. They have allowed a similar number of shots against and, when including shot quality variables into the shots, we see the team’s expected goals against also remains similar.

The team has allowed a greater percentage of these shots score, though, despite the quality being similar.

The team has performed better offensively by scoring a greater number of goals, and this has lead to more wins. All evidence points out to this offense not being sustainable or driven by a boost of confidence, as both the Jets generation in shots and expected goals remains similar.

It is just a greater percentage of shots have happened to score relative to normal.

Timely saves do exist; they are more typically just called saves.

The truth is that the vast majority of saves are timely. Some may be slightly and marginally more timely than others, due to falling in higher leverage moments, but these are minor degrees of timeliness and not at all appropriately weighed or processed by the very well-known faulty memories of human beings.

To be blunt, nobody has the ability to appropriately combine all the variables and probabilities that go into a game to confidently and accurately depict the true timeliness of a goaltenders overall performance throughout a game. There are strengths and weaknesses to quantitative analysis. There are strengths and weaknesses to the traditional “eye-test.” Combining the probability of impact of each potential save has on winning over 30+ shots does not fall in the eye-test’s strengths column.

This is not to insult the eye-test; it is merely the way our brains work.

Another issue is that hockey being a dynamic flow sport makes the leverage of a potential goal at the time unknown. Any save or goal at any point in time may be the turning point in the game or the winning save or goal. Hockey is a low scoring sport where most chances are created from errors and most of these chances are stopped.

Making a big save when the game is tied 3-3 is nice, but not making two more save earlier in the game created that environment. How often in other areas of life may someone be praised for fixing a mistake they themselves helped create?

Goalies know this. They are literally under pressure essentially all game except maybe in the very extreme cases of “garbage time.” And just like far more saves are timely than people give credit for, far fewer saves are in true low-pressure, garbage time than people give credit for.

The truth of the matter is when people look at and talk about timely saves, they are really just arbitrarily grabbing a few saves that stood out to them and so they are placing absurd amounts of value over that individual event. Again, this is perfectly natural as humans since this is part of basic human psychology. We cannot remember everything, so our memory emphasizes a few moments that stand out.

There is some irony that those who extol these late game timely saves due to “”the importance”” of late goals are the ones who stress the importance of scoring first and setting the pace. Again, we find that people misunderstand probability.

The first goal is important and those that score first win 67 percent of their games, but the second goal is also as important as those that score second win 67 percent of their games. In addition, those that that score third win 68 percent of their games, and the difference from 67 percent is likely just due to sample.

Hockey is a goal scoring contest where the most goals wins; therefore, it is important to score and prevent as many as possible. All goals and saves are important.

Players are always trying to score and prevent the opposition from scoring, because they know every goal may be imperative to the team winning. The job of a goalie is to prevent as many shots and chances as possible being goals to help improve the team’s chance at winning each game.

The Winnipeg Jets have won more than they have loss with Pavelec in net. This is not the same as the team winning more than they have loss because of Pavelec being in net.

We cannot say that Pavelec’s play has been the reason for the Jets winning; however, we can say that the Pavelec has played well enough for the Jets to win the games that they did play, given all the things that Pavelec does not control or impact.

The issue many have is that this “good enough to win” is in part due to the environment that Pavelec is in place and has no control over. The Jets have scored far more goals with Pavelec in net this season.

It is not due to his stability or creating confidence in the team either. As I noted earlier, the Jets have not been performing better in shot quantity nor the many variables we can measure that impacts shot quality. It is extremely unlikely that confidence would somehow significantly change the team to shoot well over 13 percent while also not impacting any of the known and measurable factors that impact shot quality.

Pavelec has been “good enough” to win, but that bar has been extremely low. If the bar of “good enough” was lowered to Michael Hutchinson or Connor Hellebuyck to the same level as it has been for Pavelec, they would have a great deal more wins as well.

The Jets have scored 3.8 goals with Pavelec in net. They have one game where they scored two goals, and they lost. They have one game where they scored three goals, and they lost. They have three games where they have scored four or more, and they won all three.

In games where Hutchinson has played at least two periods, the team has averaged 2.6 goals. They have seven games where they have scored less than three goals, and Hutchinson is 1-4-2 in those games. They have five games where they have scored three times, which Hutchinson is 1-3-1 in those games. They have only scored four goals or more in three games for Hutchinson, where he is 1-1-1 in those games.

If the Jets were able to produce three goals in regulation in those games Hutchinson lost (0-8-3), the team would have added three more wins and pushed one regulation loss into overtime. If the Jets were able to produce four goals in regulation in those games Hutchinson lost, the team would have added five more wins and three regulation losses would have gone into overtime.

In games where Hellebuyck has played at least two periods, the team has averaged 3.0 goals. The team has scored fewer than three goals 11 times, with a record of 2-9-0. The team has scored three goals eight times, with a record of 5-2-1. The team has scored more than three goals 10 times, with a perfect 10-0 record.

If the Jets were able to produce three goals in regulation in those games Hellebuyck lost (0-11-1), five of the regulation losses would have been pushed into overtime. If they scored four goals, the team would have added six wins and pushed three regulation losses into overtime.

Interestingly it seems like the team is scoring goals more often in games where Pavelec stopped a low percentage of pucks than when the other goaltenders stopped a similar percentage of pucks.

Of course, the previous exercise is holding all things equal in a vacuum and hockey does not work like that. Every goal impacts future events in the game just like how every save impacts the game. However, it does help demonstrate why the Jets have won so much more with Pavelec in the net.

Quality starts are one way where we can look at a goaltender’s impact on wins outside from their overall save percentage. A quality start is simply any game a netminder either a) posts above a league average save percentage or b) posts above a replacement level save percentage while allowing two goals against or fewer.

The formula essentially gives you the number of starts where a goaltender stopped enough pucks that a league average performing team or better would have won the game.

Hellebuyck has led the Jets netminders with sixteen quality starts of his thirty-three games started, with a percentage of 48.5. Hutchinson and Pavelec have only four and two each respectively, for percentages of 26.7 and 33.3. It should be noted that typically a good goaltender posts a 60 percent Quality Start Percentage or better. None of the three Jets’ netminders have been above average, but all indications are that Pavelec has not been the goalie deserving the most starts.

Quality Starts are imperfect, but they allow one to look just beyond save percentage and separate how often a goaltender was good enough for the win with how often a goaltender was good enough that it should have been a win.