Have the Jets fixed Ondrej Pavelec?

The Winnipeg Jets start off the season 3-1-0, thanks in part by a superb 0.942 save percentage.

The Jets’ netminder Ondrej Pavelec stopped just over 93 per cent of shots during the road trip, continuing from his strong performance from last season.

It was not long ago though that Pavelec was the figure of scrutiny, with many calling for a buyout.

How sure can we be of this “new Pavelec” is indeed new?

Sometimes past success is indicative of future success; sometimes it is not.

The whole point to statistical analysis is to test and study when you can best say success is indicative of the future and how much confidence you can have in that.

“Goaltenders are voodoo”

It’s a common saying due to the difficulty in evaluating goaltenders. Scouts have been notoriously bad in drafting them. Teams have been notoriously bad in evaluating which ones to use and keep. And numbers take ridiculously large samples to balance out all the noise.

During his time as a Jet, Pavelec has posted a 0.9085 save percentage under all-situations, well below the 0.915 league average and 0.917 starter average during the same time frame.

Pavelec has been better though. He has stopped more pucks recently that before.

There is a noticeable shift between his time under Claude Noel and Paul Maurice:

Not only has Pavelec stopped more pucks, but he’s stopped more pucks from all three of the high, medium, and low scoring probability areas, suggesting the improvement is not due to better team defensive play.

How confident can we be of the numbers under Maurice?

The industry standard for analyzing goaltenders with these numbers is 3000 shots against. This is the minimal sample size we have decided upon with balancing confidence and utility (having to wait the length of a full career would be great confidence wise, but would take too long).

We still see improvement, but not to the same degree.

Interestingly, Pavelec has recently performed quite a bit better in the low slot area, and slightly better in the high slot area, but slightly worse in shots from the outside. This suggests that the difference in save percentage from the two samples may be under exaggerating the difference between the two samples.

In the end, the improvement is minor, although still an improvement.

The difference between the 3K splits as opposed to the coaching splits is due to Pavelec having one of his worst streaks ever just at the end of Noel’s tenure, and likely in part why True North removed the coach.

There is one other method for predicting future success with goaltenders while being open to the idea of improvement.

“Garik16” at Hockey-Graphs constructed Goaltender Marcels. Marcels is simply a weighted average, so the more recent results are weighed as more important than the distant past results. It is a way to increase sample size while being understanding that change is possible in talent.

Pavelec’s Marcel save percentage for 2009-2013 is 0.9037.

When we look at his most recent Marcels though we get a 0.9095.

What’s the take home?

By the eyes and the numbers Pavelec has been a better goaltender. He’s stopped more pucks, which is the central role of goaltending.

Still, while Pavelec has been better, numbers suggest he is only just a better version of the previous Pavelec.

The underlying numbers suggest that the newer Pavelec is somewhere between 0.0015 to 0.0058 better goaltender in talent level. This should lead to more wins.

However, with Connor Hellebuyck in the wings, and Eric Comrie starting in pro, the Jets should not tie themselves overly to Pavelec.

Pavelec’s contract ends after the 2016-2017 season.

The other thing to note is the status of Michael Hutchinson. Hutchinson’s contract ends one year earlier, at the end of this season.

It is possible the Jets would drop Hutchinson as Hellebuyck pushes for a spot. Hopefully not though, as every number places Hutchinson as the better performer thus far:

Pavelec may be better than what he was, but the Jets may still be best suited moving him at the end of the season if possible.