Playoff Push: Winnipeg Jets should go all in on Michael Hutchinson

Usually when someone calls for a team to increase the workload of a goaltender, it extends from excellent performance from said goaltender. For the Winnipeg Jets though this is not the case. Hutchinson has not played very well over the past ten or more games.

Despite Hutchinson’s slumping play, the Jets should start leaning on him more.

Here is why.

The Winnipeg Jets’ current choices for goaltending are Michael Hutchinson and Ondrej Pavelec. That’s it. Kevin Cheveldayoff could sell assets for a goaltender, but that would be a dangerous gamble. Unless a trade happens, it’s either Hutchinson or Pavelec and neither have been playing well lately.

Over each of their past 10 games with full 3 periods played, Hutchinson and Pavelec have posted a 0.901 and 0.902 save percentage. Pavelec has only won three of these, with two being in regulation. Hutchinson’s wins over that span looks better on the surface with seven wins, but only three have been in regulation and much of it has been due to large goal support by the Jets’ skaters.

The Jets’ have been outscoring their goaltenders’ failures; however, they will not be able to keep this pace forever. Eventually the Jets goal scoring will slow down. If the Jets cannot improve their goaltending, the teams chasing them will catch up.

Why change the current 50/50 rotation?

Screen shot 2015-02-18 at 2.09.47 AM

At this point in his career, Ondrej Pavelec is a known commodity. His final save percentage each season has been relatively consistent. His struggles in stopping the puck at average rates suggest that the Jets’ poor win differentials under Pavelec are sustainable. Even with the current Jets performing as a top ten shot metric team, Pavelec has struggled to push the Jets over 50 percent win percentage.

According to Sports Club Statistics, 14 wins in their final 23 games essentially guarantees them a spot in the playoffs. Thirteen wins gives them a 99.5 percent chance, while twelve, eleven, and ten wins gives them a 91, 82, and 52 percent chance respectively. At nine wins, the Jets become long shots with only a 14 percent chance.

If Pavelec plays the full remaining 23 games, the average expectation would be ten wins. While he could end up with more or less, on average you would expect ten. If Pavelec were to play eleven of the twenty-three, you would expect somewhere between four to five wins on average.

This is not to beat on Pavelec. Statistical analysis is not for throwing certain players under the bus. Rather, statistical analysis is for making better informed decisions. It’s all about optimizing wins, nothing more.

Why then give the greater workload to Hutchinson?

Screen shot 2015-02-18 at 1.19.50 AM

While the two netminders’ performance has been similar over their last recent span of games, Hutchinson has performed superior over the span of the season. In fact, Hutchinson’s pace in stopping pucks from going into the net is well above any season Pavelec has had yet as a Jet.

The Jets have won significantly more often with Hutchinson starting than Pavelec. Of course, this is a bit unfair to Pavelec. The Jets have scored on a greater percentage of shots with Hutchinson on net, which is something that goaltenders have very little direct control over.

The Jets shooting percentage has been 1.22 percentage points higher with Hutchinson than with Pavelec’s combined four seasons as a Jet. Still, Hutchinson has been the larger impact, with his save percentage being 1.66 percentage points higher than Pavelec’s. It is interesting to note that Hutchinson on average has faced teams with better goal differentials and higher scoring rates, so the dichotomy is unlikely due to quality of competition difference.

Now it is not completely true that goaltenders have no impact on shooting percentage. Shooting percentage is impacted by score effects. Teams leading in the game tend to experience a boost in shooting percentage as their opponents get caught taking risks when trying to catch up. There is also empty net opportunities as well.

This means that a portion of Jets’ shooting percentage boost with Hutchinson could have been impacted by his superior performance in stopping the puck.

What about Hutchinson’s falling numbers?

Untitled2

Even Henrik Lundqvist has stretches of ten games where he performs poorly.

Hutchinson’s numbers falling has been expected, even predicted. No goaltender consistently posts good numbers all the time; slumps and streaks are expected. This is why numbers regress towards the mean over the long run. What separates goaltenders is how often they post good or poor numbers; how they perform on average.

There have been some that have assumed Hutchinson has struggled due to the increase workload. While it is possible, I don’t find it the most rational explanation. It is making a correlation equates to causation assumption typical of confirmation bias. During the first half of the 50/50 rotation, Hutchinson was posting far superior results than Pavelec. It is far more likely that the difference between Hutchinson’s hot start and cold recent play is natural variance and these players being human.

The real question is: which is the real Hutchinson?

The more probable answer is that the real Hutchinson is not as bad as his worst moments, but not as good as his best. He is likely somewhere in the middle.

This is no guarantee however. “Goaltenders are voodoo” is a common saying amongst the hockey statistical blogsophere. Hutchinson could very well be struggling due to the workload. It is even possible that Hutchinson is worse than Pavelec.

Possible and probable are not the same though.

While past performance does not guarantee what future performance will be, Hutchinson’s history does suggest Hutchinson is unlikely worse than or equal to Pavelec. For one, Hutchinson’s performance this season has been better than any of Pavelec’s other seasons as a Jet thus far. Also, Hutchinson’s save percentage performance in the AHL was far better than Pavelec’s. 

In fact, Hutchinson’s career numbers in the AHL were quite dominant:

hutch

Over the regular season, Hutchinson has posted a 0.917 save percentage where the league average is 0.912. What is even more exceptional is his 0.938 performance in the playoffs, with the league average being 0.914 over the same period. Combined this puts his AHL career at 0.921. Adjusting for the differences between average in regular season and playoffs estimates him as a 0.919 talent in the AHL.

Side note: The adjusted number may actually be less accurate than the combined number. There is evidence that the rise in save percentage in the post season is almost exclusively driven due to survivorship of talent. In other words, goalies don’t tend to post higher numbers in the playoffs, but rather less back-ups and terrible netminders play in the post-season. It used to be thought that there was a decrease in PPTOI in the playoffs, but this has not been true over the past 7 seasons. I plan on investigating this further in the future.

Could this end up rushing Hutchinson’s development?

Indexed_NHL_Index_medium

Image source.

Again, anything is possible, but not everything is probable. Less than a month shy of twenty-five, Hutchinson is at the age goaltenders typically peak in performance. The fact that this is his rookie season is unlikely to shift his developmental curve.

Currently evidence indicates that developmental curves are predominately impacted by age, not years of experience in the NHL.

Historically it was once thought that players of different positions tended to peak at different ages. This erroneous opinion was fortified by the fact that forwards tended to enter the league at a younger age on average, followed by defenders, and then goaltenders.

Most recent studies though have shown all three positions peaking in performance at the same point despite the difference in average rookie age. Whether save percentage, Corsi, or point production per minute, all three positions tend to perform best around 25. Not unrelated, this is also the same age most academic studies have found to be the peak in human physical and cognitive performance.

The pattern in average rookie ages entering the NHL likely arises from two factors. There are more forward positions available than defenders, and only sixty spots for netminders. In addition, a forward’s error is less likely to end up a goal against then either a defender or a goaltender. Coaches -and humans in general- are risk averse and will avoid the unknown even if the known is sometimes below replacement level.

What is the right split?

Screen shot 2015-02-18 at 2.50.46 AM

The first expected wins column uses the Jets win percentage with each goaltender. I used the slightly higher 43.6% four season sample for Pavelec over his current 40.6% season. The second column reduces Hutchinson’s win percentage distance from Pavelec’s to about half, to account for the inflated shooting percentage.

According to past performances, the Jets would probably do well in giving somewhere between 15 and 18 of the remaining 23 starts to Hutchinson. Historically speaking, this gives the Jets their best chance in making the playoffs.