Advanced Stats: Using VUKOTA To Predict Stats and Finish This Season

In the last post we introduced Hockey Prospectus’ VUKOTA projection system and scrutinized its forecasting skills against the actual results of last season’s Blueshirts squad. While VUKOTA proved somewhat underwhelming in a limited sample I thought it would be interesting to see how VUKOTA pegged this year’s club. Here are the results.

Player

Pos

GP

G

A

Pts

PPG

PIM

GVT

Brad Richards

F

72

28

48

76

9

22

14.0

Marian Gaborik

F

69

27

33

60

8

28

11.1

Derek Stepan

F

79

25

30

55

6

26

8.8

Brandon Dubinsky

F

75

24

30

54

6

78

9.3

Ryan Callahan

F

66

23

26

49

8

47

8.3

Wojtek Wolski

F

67

15

27

42

3

19

6.7

Artem Anisimov

F

72

18

23

41

4

22

6.7

Mats Zuccarello

F

51

10

21

32

2

19

4.8

Brian Boyle

F

70

16

13

29

3

58

5.1

Marc Staal

D

73

6

22

28

2

45

7.3

Dan Girardi

D

71

5

21

26

2

36

7.0

Ruslan Fedotenko

F

62

11

16

26

2

29

3.9

Sean Avery

F

62

8

17

25

2

130

3.3

Brandon Prust

F

68

9

13

22

1

131

3.1

Michael Del Zotto

D

59

5

17

22

2

28

3.5

Mike Rupp

F

64

8

7

15

1

86

1.5

John Mitchell

F

44

6

9

15

1

27

1.4

Michael Sauer

D

66

3

11

14

0

62

4.7

Ryan McDonagh

D

51

2

10

12

0

32

3.4

Dale Weise

F

30

5

5

10

1

42

0.8

Chad Kolarik

F

25

4

5

9

1

24

1.0

Steve Eminger

D

55

2

7

9

0

24

1.6

Kris Newbury

F

28

3

5

8

1

43

0.8

Totals:

246

393

639

60

Name

P

GP

Save %

 

 

 

 

GVT

Henrik Lundqvist

G

59

0.919

 

 

 

 

17.9

Martin Biron

G

21

0.908

 

 

 

 

0.8

Team GVT:

137.6

Let’s ignore the individual stats for a moment; instead we’ll focus on the team GVT value of 137.6. If we recall the post in which we extrapolated GVT into a WAR-type stat, Tom Awad indicated that a GVT value of three was equivalent to one point in the team standings. Dividing 137.6 by three will leave us with 45.9. Also, a threshold team is expected to finish a full slate of games with 50 points. If the Rangers team GVT value of 137.6 equals an additional 45.9 points in the standings over threshold then VUKOTA is projecting the Blueshirts to finish the year with 95.9 points. Keep that in mind for a bit.

As far as individual stats are concerned I am somewhat surprised Gaborik wasn’t projected higher. A 27-goal output would likely be considered a disappointment given his talents and the addition of Richards. Richards meanwhile was forecast to be slightly better than a point-per-game player which makes sense.

Brandon Dubinsky is expected by VUKOTA to virtually repeat his production from a year ago. Depending on how the Richards acquisition impacts his ice time and whether or not Dubinsky skates on the top line with Richards and Gaborik much, a case could be made that VUKOTA’s projection is a baseline for the Rangers winger.

After potting 23 goals in just 60 games, a figure which would prorate to better than 31 tallies in a full schedule, it’s surprising VUKOTA only pegged Ryan Callahan to produce the same 23 goals this season. That might have something to do with the 16 games Hockey Prospectus’ projection system expects Callahan to miss this upcoming year. Of course Callahan plays a physical game, isn’t the biggest of guys and will lie out to block a shot from literally anyone in the game with no fear and yes. All of that can lead to injuries like we saw last year. However, in his first two full NHL seasons, Callahan combined to miss just 6 games. I think it’s just as likely he plays around 75 games and finishes with somewhere between 25 and 28 goals next season as he is to meet VUKOTA’s estimate. We’ll check back later to see who was closer in their predictions.

Another player I think VUKOTA is missing the boat completely on is Artem Anisimov. In two NHL campaigns he has appeared in every game and improved his rookie year offensive production by 16 points in his sophomore year. Yet for some reason VUKOTA presumes he will miss 10 games and record fewer points this year. I actually expect the talented center to further boost his scoring totals and potentially threaten the 60 point plateau. I don’t have any advanced stats to back that supposition up but I do recognize the skill level and his incremental scoring improvement from year one to year two of his career.

I would also expect McDonagh to make more of an offensive contribution than predicted. He will see a lot of ice time on the Rangers second pair; he is a smooth skater; and is a solid passer. I wouldn’t be surprised if he approached 20 points this season given a little PP time.

Now back to the expected 96-point campaign that VUKOTA projects the Rangers to finish with based on GVT. Adam linked to a ratings system devised by ESPN.com and originally explained in articles by Craig Custance and Alvin Chang. In essence, they’ve researched different characteristics of Cup winning teams to find common themes amongst successful playoff clubs. Apparently, at least within the volume of data researched, a team’s record against clubs defined as elite (top eight finishers in the league’s regular season) was most important in achieving playoff longevity. Therefore it is assigned the highest relative value in their system and if you finish with a record of or better than .500 against the “elite eight” you are awarded eight points.

Other stats that matter are goals for (if a team averages better than 2.8 they are awarded another four points), shots for (averaging better than 31 per contest is worth four points in this system) and save percentage (a team SV% of 0.910 or better earns that club two points). Simply tallying all of the points accrued by a team is supposed to help tell how well a team may do in the playoffs. The Rangers finished fourth in the East in this exercise on the “Playoff Power Meter Score.” That can be interpreted to mean the Rangers statistically have a great chance to at least make it past round one and have a fighting chance to make it as far as the Conference Finals if this research (and the projections) is spot on.

Of course none of that matters if the Rangers don’t make it to the playoffs in the first place, right? Well the initial step this system took was to utilize some combination of VUKOTA and AccuScore game simulations to decide the final regular season standings and thus where the Rangers placed both in the East and in the NHL overall. It was also used to calculate how many of the thresholds identified as keys for playoff success a team met and thus was used in the “Playoff Power Meter Score.” I couldn’t find where the math was explained and even if I had I’m not sure I’d understand it well enough to explain it. Regardless this system pegged our beloved Rangers to win the Atlantic, finish third in the East and fourth overall in the league. That projection doesn’t seem to fit the 96-point team though as we determined VUKOTA forecasting for the Rangers.

In actuality it doesn’t. Just last year 13 NHL clubs posted more than the 96 points VUKOTA forecasts the Rangers to finish with this year according to the accumulative projected GVT ratings. That would seem to indicate either Vukota is underestimating New York or AccuScore is overestimating the squad. Either way one of them has to be mistaken.

As for ESPN’s system, I appreciate the intent and can respect some of the research done. It’s probably as good a guess as any right now. But the reality of it is that it is just a guess; an educated one yes but still a guess. We have no way of knowing for sure how well all of the new pieces in Philadelphia will mesh or if Buffalo is really talented enough to finish second in the conference as this system projects or if even Brad Richards can coax another 40+ goal season out of Marian Gaborik.

Advanced stats and projection systems can suggest certain things but not guarantee anything. That’s why they play the games afterall.

If anything I expect this system to be more accurate predicting which clubs will fare well in the postseason when utilizing the actual regular season stats recorded by the clubs this year.

Advanced stats were never meant to do anything other than provide useful new information to the folks running NHL clubs and for the fans that follow those teams. I don’t recommend any manager to run an organization using a spreadsheet. However, advanced stats can help to make important roster decisions, assist with assembling line combinations and in identifying weaknesses of your opponent. They do have their place in this great game of ours.

Remember, the use and development of these metrics is relatively new to hockey. They can and will only get better as we refine them and apply more data to the formulas. Who didn’t like reading the backs of hockey trading cards as a kid? Everyone loves stats and it’s only a matter of time before even the doubters start liking some of the advanced metrics too.

Thanks as always for reading and forgive the rant. I just wanted to clarify my position; I have never advocated for advanced stats to replace traditional methods of player evaluation. I simply happen to know a little about them, find them to be interesting blog fodder and believe they can be a tremendous resource for NHL clubs if used correctly along with being a topic that can generate interesting dialogue among fans.

Remember to give me a follow on Twitter: @Gkmkiller and if you have any suggestions for further topics please email me at gkmkiller@cox.net.