A look at the ups and downs of the Isles' even-strength play

It's been much debated of course that the 10-11 season has been a roller coaster ride for the Islanders. Most people including the media seem to believe that they played awful hockey early on, that pretty much every aspect of their game has improved in the 2nd half and that they've been a very competitive team as of late. While I agree that they've been a better team overall in the 2nd half and I do think certain mistakes were made early in the season, in terms of coaching or also how the team was put together, I'm not sure some of the predominant narratives withstand a closer look. To examine that I collected some data and tried to find out how much their actual play at even strength really has fluctuated or what the most important reasons were for the highs and lows resultwise in their season.
By the way, don't read all this as an approach to evalute the coaching. Also, the numbers provided here obviously don't tell the whole story. So, for those who've read my comments in the various threads, this might more be a follow-up with some more indications to what I think has been happnening this season. Or of course a starting point for more discussions.

However, it turns out that their EV play, measured in shot/Corsi ratios, actually has been on a pretty consistent and quite mediocre level all season. For some I guess it's nothing new the save percentage and the shooting percentage have been extremely volatile on the other hand and driven the outcome of their results indeed. However, I think some charts you'll find after the jump provide a quite interesting look at how extreme a season it's been in that regard.

First of all, as always a big thanks to Vic Ferrari, who does a fantastic job indeed providing all the data. And to garik16 who found out how Vic's scripts work this year around. Some of the usual tricks don't seem to be in action, yet, meaning that we'll have to be patient to do some WOWY analyses, but on the team level all the data's thankfully available indeed. Also, I'm no experts when it comes to statistics or graphs apparently. So, if you find anything suspicious or would like to know something in greater detail please leave a comment.

As mentioned, I'm looking at even-strength data here and EV data here means all EV situations with the score close excluding empty-net situations, and actually excluding the Isles-Atlanta game from mid-December. Somehow the NHL's failed to provide the usual shot data for that game. Let's for this discussion here just assume that game never actually happened and the Isles have played 75 instead of 76 games as of now. And, very imporant, I restrict my analysis to EV play with the score close, meaning that only events occured with the score within one goal in the 1st and 2nd periods and tied in the 3rd period come into play. That's to assure that score effects don't skew the results.

Consistent Corsi

What I did for the following few graphs is calculate 10 game rolling averages for all EV situations with the score close. The graphs therefore always start with game number 10 - obviously the first game a 10 game average is available for. The following data point will then consist of games 2-11 and so on until the last point which includes games 66 to 75. Again, one game against Atlanta in December is missing.

In the first graph you can see the rolling averages for the Isles' goal ratio (gf%, dark blue) and Corsi ratio (cf%, red) as well as the respective season averages, goal ratio in light blue, Corsi in orange. The calculation is straight forward of course - for example, the goal ratio is 40% for a given stretch if the Isles scored 4 goals in that stretch, but conceded 6. The thin brown line is the 10 game average for the Isles' shots ratio. As you can see, it moves very closely to Corsi, but is a bit more volatile.

via assets.sbnation.com

I guess the goal ratio (dark blue) is always likely to look rather volatile because the sample is small with only a few EV goals happening per game. But in most cases it certainly will be more like in the 2nd half this season and definitely not as extreme as early this season. And the interesting thing there is certainly that their hockey seemed to be of similar level during the lows as well as the highs in scoring. We can see that from looking at their Corsi (or shot) ratio. Both were floating quite consistently around the season average in the first 30 games - yet the Isles were of course in their biggest hole in that part of the season. Later on, they improved for a while, but actually played their worst hockey of the season then before picking it up again and sitting at or above the season average over the last 30 games. That average is still rather disappointing, though. At currently 45.4% their Corsi with the score close is much lower than last season (48.2%) and even slightly below 08-09 (46.0%). 

A closer look at some stretches

Just to remember, here are the game numbers for certain interesting points of the season and a look at what was going on around that time.

Game 17: Last game of the Gordon era. The data point includes obviously games 8-17 and thus exactly the final winless stretch of Gordon (0-9-1). As you can see their Corsi was consistent up to that point, but they simply couldn't score and were at a low in terms of goal ratio indeed. It got better quite quickly once they changed the coach, but the problem in these first few games was that they couldn't score on the PP.

Game 28: The famous game against Anaheim on 12/16/10 also referred to as the turning point of the season. Just from looking at the goals ratio we can see where this narrative is coming from. Corsi doesn't really confirm it, though. The high was after game 31, but let's remember that point includes games 22 to 31 and thus indicates that they played good hockey around 12/16/10 indeed, but mostly before that date because their numbers declined quite seriously from game 31 on.
I'd like to mention at this point that nothing controls for schedule strength or injuries here of course. So, a considerable decrease could also come from the fact that lots of tough games had to be played in those stretches. And I guess it was a bit of a factor with the road-trip at the beginning of 2011 and difficult games against Pittsburgh, Detroit, Chicago, Vancouver, etc. However, the fact that they were relatively successful around that time had nothing to do with their EV hockey.

Game 49: First game after the All-Star break. We can see that they had recovered already in late January, but just hadn't got the results there and still not fully right after the break. Their game did pick up a little and reached highs in late February, early March. Have to add, though, that they played a lot of home games in that stretch.

Very volatile goaltending and shooting

Now, as we can see their hockey has been fairly consistent, yet their results have been so streaky. The following graph shows their save (black, left) and shooting (pink, right) percentages and it gets pretty clear quickly why their results have been so unsteady indeed.

Again, some volatility is certainly natural because the samples are rather small with the data restricted to EV with the score close and 10 games per data point. And it's clear the shooters will somehow "get hot" in one stretch and then be "unlucky" or run into some hot goaltending or whatever. The same for the goalies who'll have some good nights in a row and then maybe some bad ones. But it's quite incredible how extreme the ups and downs have been. The goaltending was at a low of around 88% after game 17, meaning of course that this was the average goaltending they got from game 8 to 17. That's just terrible. Towards the end of the 1st half they then had around 15 games with average goaltending of 96%.

The shooting's been even more volatile. And what's really rare I guess is the fact that the goaltending as well as the shooting went hot and were at a low respectively at exactly the same time. Normally, these two things are totally independent and we'd expect a picture similar to the one in the 2nd half. Obviously, if the shooting is hot, but the goaltending not so much, or vice versa, your ups and downs will never be as extreme as they were here in the 1st half. Around game 17, paired with the ugly goaltending was a shooting of under 2%. On the other hand their shooting was incredibly hot from the All-Star break on with numbers north of 10 percent for a 20-game stretch and compensating for the most awful goaltending they got all season.

To make things more clear we'll bring together the previous graphs. You can see the goals ratio again (dark blue as in the 1st graph), then the Corsi ratio (red), which as mentioned doesn't explain the volatile goals ratio. But you can now also see PDO (green, right), which is just the sum of the save percentage and the shooting percentage.

The conclusion is of course straight forward from the previous graphs and actually natural anyway. Given the relatively consistent hockey they played (Corsi) it's the PDO, thus the goaltending and the shooting, explaining a large part of the volatility of the Isles' season. Again, that's natural and could be found for many other teams, but it just shows nicely again how extreme the goaltending and the shooting trended in the same direction in the 1st half of the season and what that did with the results they got.

Quick look at the individual level

There'll be more time later to take a closer look at how the guys performed individually over the course of the season. I just wanted to add one little thing here because it illustrates a point I've tried to make in a few comments. It's about how much the Isles have depended on the Nielsen line and how much the 4th, but also the 3rd liners have struggled lately. It's not all to assess the coaching, though, just to illustrate who's had what kind of Corsi numbers at which point in the season. You can see two points for every player, one is always the average Corsi in the first 17 games of the season (under Gordon, red), the other is the average Corsi in the rest of the season (under Capuano, blue). Again, tells us nothing about the coaching, just shows who's delivered what for whom. To be clear, the sample is very small for Gordon's tenure of course and these numbers generally don't tell us anything about the talent of the player or which coach got more out of the respective player. Also if someone was used in a much more defensive role under either coach he'd naturally have a lower Corsi for that coach. As always with individual Corsi we need to look at the context.

Grabner-Nielsen-Okposo: They've done all the work for Capuano, outshooting the opposition quite clearly despite playing against tough competition and starting in the own zone quite a bit. We don't have the exact numbers for these measures just for the Capuano tenure, but since Okposo has only played for Capuano this season and pretty much exclusively on this line indeed, his individual figures represent a good guess for what this line has faced in terms of competition and zone starts. We can also see that Grabner's improved a lot, thanks to various factors of course, and that Okposo has been the best player for Capuano, whereas Nino and/or his line was a liability early on.

Moulson-JT-PA: They've played pretty consistent all year and also have had pretty much the same roles all year indeed. Actually, it looks like PA's been a bit better away from the other two in terms of Corsi. Again, that could depend on usage and team mates for example and will be difficult to track, but it's however wrong to say PA was off to a slow start.

Comeau-Bailey: These are the disappointing numbers. Both have regressed considerably and especially for Bailey this is a concern. Comeau played a lot with JT early on and benefited a bit there I guess and his numbers would be expected to be down a bit later on. Still, it's disappointing he's been below 45%. And since paired with Bailey on the 3rd line they might not always have had a good 3rd guy, but certainly didn't face tough competition, either, and just haven't done very well indeed.

Martin-Konopka: Martin's tough to get a read on due to a small sample in the first part and then different usage. Konopka's been better lately mainly due to the fact that he's had easier minutes. Early on, he played against tougher competition and most of all had much less favourable zone starts. He still starts in the own zone a lot, but the 4th line to be close to 40% has certainly been a disappointment given the fact they played against 4th lines most of the time indeed.

Conclusions

That last part about the individual performances is just a side note really. Corsi is not easy to read on an individual level. Then again, it's pretty obvious Nielsen's line has been huge for them and the 3rd and the 4th lines haven't been good enough. Appartently, that's the main reason why their team Corsi has still been quite mediocre even since the All-Star break. We can see in the first graph it's been relatively steady above the season average, but as mentioned still a bit away from 50%. And that's really where you need to get to in order to be a competitive team. Obviously, there's other factors, such as goaltending and special teams, but Corsi really does mean a lot on the team level. For example, from the 16 teams likely to go to the playoffs this season only Anaheim and the Rangers (narrowly) will be below 50% in Corsi. Both will make it thanks to great special teams and great goaltending. Not sure the Isles will be able to rely on both these factors next season... Of course, a more healthy D will help, but I still think they need to improve their team during the summer.

Anyway, this is as said not meant to explain everything. But it does offer a different perspective and is also for me a starting point to maybe look at these things for EV in general or 20 game rolling averages. Also, I should be able to update once the season is over.