Devils Goaltending Will Improve, But It’s Still a Problem

Sometimes, in hockey, there are things that go poorly in the beginning of the season that require intervention to improve: a line doesn’t gel, a schematic component doesn’t fit, the special teams isn’t working, etc. Other times, there are deficiencies that are due either exclusively to the bad luck of random variation: funny hops, bad calls, the existence of the shootout, etc. Other times still, there are things that fall into both categories. The Devils’ goaltending is in that last category.

First let’s look at what is most definitely problematic about the Devils’ goaltending. Our net is minded by two equally uncertain quantities: a wholly unproven youngster with an enigmatic statistical history as a prospect, and an aging vet whose deterioration and health have made it 3 years since he put together a healthy and effective season. I’ll start with the latter — Cory Schneider.

I’ve been beating this drum since 2017 when I wrote this article, and followed it up in 2018 by suggesting he should be sent to the AHL for Blackwood. I was elated when it began to look like I was wrong at the end of last season as Cory looked healthy for the first time since 2016. In the last two months of the season, he registered a .921 save percentage over 17 games. Most people, myself included, agreed that he looked healthy and locked in to close that season — much like the Cory Schneider of old. Therefore, many of us were cautiously optimistic that he’d give us something in the vicinity of the .920s again this season. He looked strong in the preseason and to start his first game, saving 19 of his first 20 shots. Then he got a cramp on the 21st shot and would not return to the game. I have no evidence that this was something more pernicious. I’ll just say that, as a fan, it concerned me that goalie whose health has been, perhaps, the most important factor in making him the 46th ranked goalie out of 50 in Sv% vs expectation, was removed from game 1 and subsequently saved just 87% of shots over the next 3 games.

While Cory Schneider’s recent checkered statistical history is problematic, Mackenzie Blackwood is far from an analytics darling himself. As recently as 2018, Blackwood held the worst Sv% in the AHL and had to be demoted to the ECHL. He was back up to a .902 in the AHL last season, but it’s fair to say the .918 he posted in New Jersey was a shockingly pleasant surprise. It’s also worth pointing out, though, that after an amazing start that saw him post a .947 Sv% and a 4-2 record over his first 9 appearances (6 starts), he ended up with a negative dSv% (Sv% vs expectation) for the season. He’s sporting a n .871 Sv% so far this season, which is not far from what you’d expect of a sub-900 AHL goaltender, but very far from what you want from your top goaltending prospect coming off an encouraging rookie campaign. The charitable view of his season is that he came in cold to game 1 against Winnipeg, then played in a back-to-back next against Buffalo, and started his 3rd consecutive game despite being 1 day removed from the flu against Arizona. In the games he started with normal rest with a normal body temperature, he’s 2-0-1 with a .936 Sv%. A less charitable interpretation is that he’s been a highly inconsistent goalie throughout his entire career, and that’s bound to continue at the NHL-level.

So, Cory’s bad because he’s old and broken. Blackwood’s bad because we don’t have enough evidence to see he’ll ever be consistently good. Why am I saying the Devils goaltending will improve? Basically — no team is THIS bad for a full season. In the analytics era, no team has ever posted below a 0.885 Sv%. Now, you may rightly point out that Sv%s are down across the NHL this year, but they’re only down 0.002 from last season, when the worst team, San Jose, was a .889 team. The Devils stand this season at an .870, so given the pessimistic assumption that we are equivalent to the worst goaltending team in the NHL last year, and that this year’s Sv%s are .002 worse than that, we STILL would stand to see an improvement of 0.017 through nothing but random variation. That may not seem like much, but over the course of 261 shots against, that translates to roughly 4 and a half goals, or over a third of the Devils current goal differential (-12). If we focus on only even-strength, that calculus would equate to a 2+ goal improvement, erasing half of our 5v5 goal differential (-4).

The Devils have a problem at the goalie position. It’s not clear that Schneider has any good years left ahead of him, and it’s not clear if Blackwood will ever be consistently good. But even with this positional deficiency, it’s not unreasonable to expect the goaltending to improve a little. The exact anatomy of the improvement is a mystery. Perhaps my charitable interpretation of Blackwood was right and he actually is a good NHL goalie when put in a position to succeed. Perhaps my concern with Schneider’s potential injury is not well-founded and he’ll do just find after having had an extended rest. All I know is that the Devils are either on course for a fairly notable goaltending improvement, or a historically bad season in that regard. I suppose we can never rule out the latter, but it’s not where I’d put my money.

What do you think of the Devils goaltending woes so far? How much is real, how much is just a statistical artifact? How much do you blame the team in front of them versus the goalies themselves? Feel free to leave your thoughts on these and any other goaltending issues in the comments below, and as always, thank you for reading.

*Note on metrics used

The best goaltender metrics are ones that account for the expected goals versus actual goals (GSAA, dSv%, GSAx, Goalie_GAR, etc.). Unfortunately, as has been pointed out by EvolvingWild (scroll up through thread here) and reported on by Greg Wyshynski here, the NHL changed the way in which they record shot location data, and there was a glitch through the first 91 games of the season that has since been corrected. However, those first 91 games have yet to be retroactively adjusted, so all xG models that include those games will be inaccurate. As such, I opted to use straight Sv% and 5v5 Sv% numbers when discussing this season.