The Islanders vs. Hockey Prospectus projections, and what it means for the future

One thing I've noticed a lot in the comments of posts on this site is that there is a majorly optimistic idea of how well this season would have turned out for the Isles if not for certain injuries.  The feeling is that the Isles could've been a possible playoff contender.  In reality, that was probably untrue, as Hockey Prospectus' projections, which are not subjective in any way (unlike those of ESPN),  had the Isles projected to finish 28th in the NHL.

Hockey Prospectus (formerly "Puck Prospectus") has created a projection system, like their more well-known owner Baseball Prospectus' PECOTA, called VUKOTA (yes, after the ex-Isle) to predict how well players in the NHL would do in the future.  Before the 2010-2011 season, in the Hockey Prospectus 2010-2011 Annual, HP used the system to project how every player in the NHL (no prospects) would perform and how well each team would do in the standings.

The Islanders were projected the Isles to end up with 79 points in 28th place, ahead of Montreal (30th)* and Toronto (29th).  Hockey Prospectus projected the Isles to allow the most goals in the league while only scoring the 21st amount in the league.  In reality, the Isles are 19th in goals scored and 28th in goals allowed.  Again, pretty close.

*While it would seem that HP goofed on Montreal, that's not really the case: In fact if you look, they've come pretty close to the amount of goals actually scored by Montreal  (Projected: 169, Actual 165).  Instead, they've whiffed on the goals allowed....which is due entirely (no seriously, entirely) by a miss on Carey Price bouncing back to an extreme.  Goaltenders are seriously hard to predict. 

In this post, I'm going to go through the individual projections of VUKOTA at the beginning of the year and see how certain players are exceeding/failing-to-meet expectations and to see really where the Isles would be if not for injuries.

Hockey Prospectus' projects the basic statistics of players as well as its own specialty statistic: Goals Versus Threshold (GVT).  GVT is a value statistic: it attempts to take all of the offensive and defensive (and shootout!) and goaltending (where applicable) contributions of a player individually and put an overall value on that performance. 

GVT is measured in goals (created) above a replacement player.  This is a confusing definition, so let me explain each part separately:

Goals Created:  When GVT talks about goals, it's not simply talking about scoring goals, but how the players overall play contributed to that team scoring AND preventing opposing teams from scoring.  A player who provides no contribution offensively but whose play PREVENTS an opponent from scoring 4 goals will be said to have created 4 goals.  Essentially this part measures goal differential. 

Above a Replacement Player:  In THEORY, every team can easily find in the AHL or on some other team's waiver wire or free agency a replacement player who they can call up if one of the team's regulars are injured.  These players are poor players, such that they're freely available to each team.  Thus, in theory, a player who is AT or below "replacement level" can be replaced by a team without a problem.  To explain this concept more easily, lets take examples from the Isles:

For Defensemen, Bruno Gervais is a replacement player (essentially).  In other words, every team in the NHL to start the season, barring injuries, can call up from the AHL defensemen able to give the same amount of contribution as Bruno Gervais has this season. 

For Forwards, Michael Haley is a replacement player.  Once again, this means that every team can easily call up a player who can give the same performance as Michael Haley without any problem. 

Now, when we talk about Goals above a replacement player (aka GVT), what we mean is that if you gave both a certain defenseman (or forward) and Bruno Gervais (or Michael Haley) a specific amount of ice time, GVT measures how many more goals would be created by the certain defenseman than the amount created by Bruno.   Essentially, GVT basically measures player performance relative to a certain threshold, which is set at a pretty poor level. 

Every team overall consists on average of players who are a good bit better than replacement.  A team full of replacement level players would be outscored by 123 goals.  For comparison sakes, #30 Edmonton last year only was outscored by 70 goals.  Replacement players are bad. 

NOTE:  Those who follow baseball will note that GVT is essentially a hockey version of WAR.  That said, there's a key thing to remember when making this comparison: GVT is NOT measured in wins, but in GOALS.  If you want to figure out the amount of wins that have been created by a player, divide their total GVT by 6. 

On an individual level, GVT has some clear issues, as you'd expect since it's kind of hard to categorize hockey.  A goon or fighter for example should theoretically be making his teammates play better, and thus the positive impact he has on a team will probably not be seen in that player's own GVT but in the numbers of his teammates.  Similarly, while GVT DOES try to account for in-what-zone players start their shifts, as well as PK and PP minutes, it doesn't take into account how much a player is hindered or helped by having good teammates or by facing tough opponents.  So, a D-man who faces the opponents' top lines may have worse defensive GVT numbers simply due to who he has to face.  On a team level, these problems go away, but on an individual level they are clear. 

That said, it's easiest to compare the VUKOTA projections in GVT, so I'll be doing so from here on in. 

OKAY ENOUGH OF THAT, On to the projections.  Two notes:

First, due to the incredible difficulty in projecting shootouts, I removed shootout numbers from the projections and actual results.  For the most part this doesn't change anything, though it helps the goalies generally and hurts Frans and Rob Schremp.

Second, when a player was not projected by VUKOTA, I assumed he was projected to be a replacement player (GVT of 0).  You'll notice this a lot, given how many substitutes the Islanders have called up.

FORWARDS:

Table 1:  The Current GVT Numbers of Islander Forwards, the Amount expected by VUKOTA in preseason and the difference between the actual numbers and the projections. 
TABLE LEGEND:
Offensive GVT:  Amount of Goals (above replacement) created on Offense, by a player's play resulting in more goals scored.
Defensive GVT:  Amount of Goals (above replacement) created on Defense, by a player's play preventing opponents from scoring goals. 
Total GVT:  Amount of Goals Created (above replacement) in total (OGVT + DGVT).
Expected GVT:  The GVT expected of each Islander by VUKOTA in this amount of games played. 
Extra GVT Above Expectations: The amount (in GVT) that each Islander has surpassed (or by how much the player failed to equal) expectations.

I've taken the liberty of highlighting the players who have really by a good amount exceeded expectations in green, while those in red have been disappointments. 

First things first, the most valuable Islander forward (and in fact, most valuable Islander overall) is not John Tavares or Matt Moulson, but is Michael Grabner.  This may confuse you, as Grabner is fourth on the team in points.  The answer lies in two factors: defense and ice time.  Grabner gets a good amount of defensive ice time, a large amount of PK time, and very little power play time (though that's increasing lately).  Meanwhile, the Tavares line (of MM-JT-PAP) gets the most offensive ice time at even strength and the most power play time, with basically no PK time.  Thus, GVT states that Grabner's offensive contributions are almost as valuable as Tavares' contributions on offense, because Grabner is scoring with much less opportunities than Tavares (thus a replacement player playing JT's ice time is expected to score more than a replacement player playing Grabner's).  Meanwhile, Grabner has been a revelation on Defense, easily the team's 2nd best defensive forward after Frans Nielsen (though it helps to be playing WITH Frans).  These two factors make Grabner the team's MVP so far. 

Aside from Grabner up top, the GVT rankings should pose little surprise.  Grabner is #1, then Tavares, then Moulson, then Nielsen (though Nielsen jumps Moulson if you add shootout numbers), then Blake Comeau (who gets a boost because of the tougher ice time he faces than PAP), then PAP, and then Kyle Okposo above Josh Bailey and Rob Schremp, despite playing only 19 games.  Meanwhile, Jon Sim was truly dreadful this year at hockey and the team is WELL rid of him. 

So who exceeded their projections?  Four Isles have exceeded expectations this year:  Michael Grabner (duh!), John Taveres (duh), P.A. Parenteau (again, duh), and Frans Nielsen.  Nielsen is the only sort-of surprise, as he's the oldest of the group and unlike PAP, has played a number of years in the NHL before.  We might have expected Frans to have peaked already last year.  That said, his defensive numbers have been terrific this year AGAIN despite playing harder minutes, and his offense has been solid as well.  In fact, I think GVT is underrated Frans' defense by a good bit since GVT doesn't take into account the tough opponents that Frans faces. 

Who simply has met their projections (give or take a little)?  Well Matt Moulson and Blake Comeau both simply seem to be meeting their projections, though MM is ahead of them by a little bit.  This suggests that VUKOTA has them pegged correctly...which means that we really shouldn't expect too much growth for either player in the future.  Comeau is better than most people here believe...but he's not going to get better.  Moulson is terrific, but he's at his peak (he's not young obviously), so any better numbers from him are likely to come from better teammates (JT as he grows) and better ice time (caused by having a better team) rather than any growth of MM on his own. 

Now who has disappointed?  A whole bunch:  Rob Schremp, Josh Bailey, Trent Hunter, Jesse Joensuu, Trevor Gillies, and Jon Sim.  VUKOTA was very bullish on Rob Schremp, who had a better end of last year than most people realize.  Josh Bailey also has disappointed, though part of this is GVT undervaluing Bailey's defense.  Regardless, this is quite a disappointment....VUKOTA thinks that Bailey should've contributed at least as much as PAP and he's not even contributing as much as he did last year...by a good bit.  That's not good for Bailey's development long term. 

Trent Hunter, though now injured, was pretty poor when he was on the ice.  His run as an Islander is coming near an end.  Jon Sim's time as an Islander HAS come to an end.

VUKOTA was oddly positive of Gillies (from a tiny sample size) and Joensuu (same), and has not been rewarded.  Gillies is barely on the ice but he has managed to be just atrocious when he has been on the ice in the offensive end.  Joensuu similarly has been just bad on O, which has been sad.

And then there's Kyle Okposo.  In the 19 games KO has played, KO has underperformed his expectations.  Oh, yeah, and in addition he's been out for most of the season.  Kyle Okposo was projected to put up 6.8 GVT if he played all 63 games this season.  Instead, he's just put up 1.7.  Quite a bit of a drop there. 

So how have Islander forwards overall fared? 

Overall, ignoring the Okposo injury, Islander forwards have failed to live up to expectations by 1.5 GVT.  That's not what you look for in a growing team, though to be fair, if you count only the core 6, only Josh Bailey has failed to exceed expectations. 

DEFENSEMEN

Table 2:  The Current GVT Numbers of Islander Defensemen, the Amount expected by VUKOTA in preseason and the difference between the actual numbers and the projections. 

The Isles started this season way shorthanded without Mark Streit.  I've included in the table for emphasis what VUKOTA projected for Streit had he been healthy for all 63 games so far, a mark that would easily surpass all of our current defensemen by a good margin (of at least 2 goals over AMac). Meanwhile, Aside from Streit, the Isles have used TWELVE DEFENSEMEN (Mark Katic is not in the chart since his data wasn't available)!  Needless to say, one would be very surprised if the Isles had done much with this D unit, and VUKOTA agrees, not projecting high performance from anyone but Mark Streit.

Despite that, the Isles actually have had 3 players greatly exceed VUKOTA's admittedly not-that-high expectations.  Andrew Macdonald has put up GVT numbers at a rate that ranks as #48 among all D men who have played at least 20 games.  Essentially, when you consider that there are 30 teams in the league, this puts him in the middle of the pack of who should be considered #2 Defensemen...which is pretty impressive when you think about it. 

Milan Jurcina has also apparently produced tremendously: his production rate puts him #49 among all D men in the NHL.  Oddly this is because GVT likes his defense.  It's my (known) opinion that this happens to be a large amount of luck, in that Jurcina has gotten some lucky breaks with his goaltending, but GVT thinks he's been pretty good and that he's already produced more value in 36 games than he was projected to during the entire season. 

But the biggest, and most pleasant, surprise has been Travis Hamonic.  His rate of production puts him at #52 in the NHL, meaning that at age 20, he's already performing at the level of a number 2 defenseman.  That's really really impressive and bodes great things for Islanders fans.  The vast majority of this great performance, unsurprisingly, comes on the offensive end, where Hamonic leads all of our D-Men in offensive value, now that James Wisniewski is gone.  That said, Hamonic's defensive numbers, while nothing special, are at the least solid.  If he improves his D, which is likely, he's going to be a big part of this rebuild. 

Aside from the big 3, some mention needs to be given to Ty Wishart.  I think that GVT is overrating Wishart's meh defense here, but Wishart's offensive contributions in limited offensive time have been enough to give him a pretty good rate of creating goals.  This bears watching.  That said, the same could have been said of Bruno Gervais and Dylan Reese in the past. 

Disappointment wise, there's a lot to speak of here.  VUKOTA was bullish on Dylan Reese after some solid play last year....and was rewarded with you know what. VUKOTA also liked Bruno Gervais and was disappointed in his play (VUKOTA includes in its projection forecast Bruno's play 2 and 3 years ago with Mark Streit as his D-pairing, which made him look good).

Meanwhile, Mike Mottau was a catastrophe this year.  He was one of our starting 6 D-men...and yet was 1.3 goals WORSE than replacement in his limited time.  Had we started the season with even BRUNO GERVAIS in the lineup in his stead, the Isles would have been better off.  Oh yeah, and we have another season of him still on the books.  He's probably someone who should be waived early next year if Streit is healthy and he can't show any clear positive impact.  He should be on a short leash. 

Meanwhile Mark Eaton, while a tiny bit above replacement in his short playing time this year, was still over a goal worse than he was projected to be.  So it seems that while Jurcina was a good signing by Garth this offseason, Mottau and Eaton are clearly busts. 

Radek Martinek and Jack Hillen both put up solid numbers, but neither are particularly impressive.  Given Radek's injury issues, his future on the Island is probably numbered.  Hillen's actually put up enough numbers to be worthy of 3rd pair ice time, or at the very least to be the #7 D man that's scratched most games....unfortunately he's been getting 2nd pair time.

Overall, thanks to Travis Hamonic, the Islander D has actually surpassed VUKOTA's projections as a whole (by 1.4 goals), if you ignore Mark Streit's injury.  If Streit can return at a similar level to before and Hamonic continues to improve, the Isles next year would have a #1 D Man (Streit) and 3 guys who could fit the role of #2 D men.  That should be a massive improvement to the blue line, and could be a big factor in a better Islander season next year. 

And now....onto the mess....that is:

GOALTENDING.

Table 3:  The Current GVT Numbers of Islander Goaltenders, the Amount expected by VUKOTA in preseason and the difference between the actual numbers and the projections. 

Yikes that's a lot of color!  One of GVT's major issues is that it overrates goaltending and that it does nothing to try and remove luck from the goaltending results.  Thus Al Montoya, Nathan Lawson, and Kevin Poulin, and Mikko Koskinen despite playing less than 9 games each, all have GVT numbers that are pretty large in one direction or the other. 

Regardless, interestingly enough, The Islanders' four replacement Goaltenders (Mikko, Poulin, Lawson, Montoya) have, in their combined 22 games, managed to play at EXACTLY replacement level.  That's probably the best the Isles could have hoped for from goalies numbered 3-6 on their depth chart (well, 3-7 if we include Nabakov's -100000 GVT). 

Where the Islanders have truly suffered was in the 20.5 games manned by former goalie of the future, Rick Dipietro.  DiPi was one of the worst players in the entire NHL this year, costing the Islanders 6.4 goals more than a replacement goaltender would in his games.  Yeah that's bad.  VUKOTA was hardly bullish on DiPi, expecting only for him to be barely above replacement, and he failed to even meet that standard.  I don't care if he was coming back from injury...it's clear that DiPi is not worthy of net time any more. 

Meanwhile, the Isles did benefit from 20 games of Dwayne Roloson in which Rolo returned to form after a poor last year (which is why VUKOTA didn't expect much from Rolo).  Alas, the Isles traded Roloson for Wishart..though with Rolo heading to Free Agency and then retirement, this was not a bad move.  Oh and Rolo has failed to keep up the pace in Tampa, with his numbers sinking closer to replacement level.  So the Isles did manage to sell Rolo high just as he peaked this season.  Rock on Garth. 

=====================

Conclusion:

The Islanders were projected to finish 28th.  But the core offensive and defensive players in general outperformed their projections by a decent amount, meaning had Streit, Okposo, Jurcina, and to a lesser extent AMac, stayed healthy, this team might be closer to a playoff spot.  But even then they likely would have fallen short. 

We can tell this with a tiny bit of arithmetic:  Every 6 GVT translates into an additional win created.  The loss of Streit will end up costing the team a projected 10 goals, so 1.5  wins.  Okposo's loss will cost the team about 3 goals, so that's another .5 wins (2 total).  But 4 points, or even 8 points if we doubled that, would not put the Isles into a playoff spot.  And remember, if Streit was healthy, odds are it would have been a good bit longer before the Isles risked bringing up Travis Hamonic, meaning that such addition is probably overvaluing the Isles' playoff hopes. 

That said, there's good news here for the future.  If we can get healthy, we have 4 quality D men, 5 quality forwards who have at least met and (most of them) beat their projected numbers, with Okposo performing only a tiny bit below his projection himself.  Bailey's regression is disappointing, but of the core, it all seems to be burning quite brightly, so perhaps next year could result in the return to the playoffs we've been hoping for.