Sab(re)metrics: VUKOTA Projections for the 2011-2012 Buffalo Sabres

Welcome to the first weekly installment of Sab(re)metrics, a column fully dedicated to the advanced statistical analysis of the Buffalo Sabres. Throughout the season, I will write about both the seasonal and career advanced statistics of the team and their players. Some of the advanced statistics I will examine include:


2. Game Versus Threshold (GVT)

3. Corsi Number

4. GAA in Even Strength and Special Teams Play

Check after the jump for the meat and potatoes to today's article: VUKOTA Projections for the 2011-2012 Buffalo Sabres.

First of all, lets define the VUKOTA Projection System. Developed by the people at Hockey Prospectus, the VUKOTA Projection System uses statistical data to project what an NHL player would do in their next season by comparing them to similar, post-1967 NHL players (1967 being the year that the NHL first officially recorded statistics to calculate GVT data). This projection system, similar to the PECOTA and SCHOENE projection systems of Baseball and Basketball, was named after Mick Vukota, a fringe NHL player who amassed 17 goals, 29 assists, and 2071 penalty minutes over an 11 year NHL career. Overall, VUKOTA only projects individual statistics (G, A, Save %, GVT) and not statistics that are team oriented, such as +/- or goaltender wins.

An example of using VUKOTA to project a player's 2011-2012 statistics will be done with our own young stud-of-a-Defenseman, Tyler Myers. To project his stats for the upcoming season, VUKOTA will compare Myers to all comparable 20 and 21 year old defensemen throughout hockey history and examine how they performed in their next season. Myers' 2011-2012 stats would then be fit to a curve based on all this player data throughout NHL history. Listed below is Tyler Myers' upcoming statistical projections for the 2011-2012 season:


2009-2010 19 82 11 37 48 13 23:44 2:58 3:04 8.8 3.2 0.0 12.1
2010-2011 20 80 10 27 37 0 22:27 2:47 2:41 5.2 4.0 -0.3 9.0
2011-2012 21 76.3 10.0 30.7 40.7 5.5 3.3 0.0 8.7

VUKOTA of course has its disadvantages, such as the difficulty of managing variables such as ice time and special teams (statistics that do not occur in Baseball or Basketball) and the lack of historical data within the tool since it is in its infancy (Hockey Prospectus has only calculated VUKOTA projects for two seasons). Also, it would definitely have problems projecting a player's output who jumps from Juniors or the AHL to the NHL level (i.e. Luke Adam). Some of the disadvantages was discussed quite well in the most recent publication of the Hockey Prospectus 2011-2012 by Tom Awad:

Mathematically, [VUKOTA] is a combination of comparable player selection and multi-variable linear regression. This makes VUKOTA rougher than its baseball or basketball cousins, but that is both because of the limited nature of the statistics and the nature of the sport. Player profiles in basketball are far more defined: big men get a lot of rebounds, have poor free throw rates, and take almost no three-pointers. Point guards who run the offense record many assists and many turnovers. In hockey, the best information we have on player types is ice time and special teams usage: after all, based on traditional statistics, what is the difference between Alex Goligoski and Shea Weber? Yet we know there is a world of difference between these two players, and they are far from having the same value.

Overall, the full two years of VUKOTA projections have seen some promising results. For example, it correctly predicted the top two teams in the NHL last season would be Washington and Vancouver for the regular season, that Henrik Sedin could not repeat his Hart Memorial Trophy Season from 2009-2010 (he dropped from 112 to 94 pts, with VUKOTA projecting 89 pts), and that Ilya Kovalchuk would not be in the top 10 of scoring last season (he finished with 31 goals, tied for 21st). With more time and historical data, the VUKOTA Projection System could become a very safe projector of a NHL player's future season statistics.

Now that we have a little background information, I want to analyze how well the VUKOTA projections will fit a Sabres player's real statistics throughout the season. I will re-examine these projections after 41 and 82 games respectively, and we will just see how accurate VUKOTA is for this year's Buffalo Sabres. Below this paragraph are tables that list the VUKOTA projections for Sabres Forwards, Defenseman and Goaltenders for the upcoming season. For forwards and defensemen, the GVT data listed is offensive GVT (OGVT), defensive GVT (DGVT), shootout GVT (SGVT) and total GVT (GVT). For goaltenders, the OGVT is replaced with goal-tending GVT (GGVT). I will analyze and discuss these types of statistics later in the year, but for now, you can think of GVT for forwards as a stat the shows how many more goals and assists a forward can score over a replacement-level player, while GVT for defensemen and goaltenders analyze how well they prevent shots on goal and prevent goals over a replacement-player, respectively.



Luke Adam 21 35.4 7.4 6.7 14.2 0.9 0.4 0.0 1.3
Brad Boyes 29 67.6 13.2 27.9 41.1 3.9 2.2 0.2 6.3
Matt Ellis 30 43.4 4.6 5.7 10.2 -0.3 1.2 0.0 0.9
Tyler Ennis 22 73.4 23.2 31.7 54.9 7.3 1.9 0.1 9.3
Paul Gaustad 29 66.8 11.5 15.8 27.3 1.9 2.2 0.0 4.1
Nathan Gerbe 24 61.5 18.0 17.4 35.4 4.6 2.0 0.0 6.6
Jochen Hecht 34 66.8 14.4 18.3 32.7 2.8 2.7 0.1 5.6
Patrick Kaleta 25 52.0 7.7 8.0 15.7 0.3 1.2 0.0 1.5
Ville Leino 28 67.4 16.2 26.9 43.1 5.0 1.9 0.0 6.9
Cody McCormick 28 63.7 8.4 10.0 18.4 0.5 1.3 0.0 1.8
Jason Pominville 29 70.9 23.3 31.7 54.9 7.3 2.3 0.0 9.6
Derek Roy 28 57.8 18.4 32.4 50.9 7.1 1.6 0.0 8.7
Drew Stafford 26 63.8 24.3 23.6 47.8 7.0 2.0 0.0 9.0
Thomas Vanek 27 70.0 29.4 34.5 63.9 9.8 1.4 0.2 11.4


Christian Ehrhoff 29 71.4 11.4 32.9 44.3 7.2 3.4 0.0 10.6
Marc-Andre Gragnani 24 30.7 2.9 9.6 12.6 1.7 1.0 0.0 2.6
Jordan Leopold 31 65.2 7.2 20.6 28.0 3.8 2.6 0.0 6.3
Tyler Myers 21 76.3 10.0 30.7 40.7 5.3 3.3 0.0 8.7
Robyn Regehr 31 65.4 2.6 12.5 15.1 0.3 4.2 0.0 4.5
Andrej Sekera 25 68.2 5.2 19.4 24.6 2.6 3.7 0.0 6.3
Mike Weber 24 56.9 3.8 12.5 16.3 1.7 2.7 0.0 4.4


Jhonas Enroth 23 19 .908 1.5 0.0 0.2 1.6
Ryan Miller 31 54.5 .916 14.0 -0.3 -0.1 13.7

Some initial observations from the above projections:

  • VUKOTA projects that both Brad Boyes and Jason Pominville will continue their downward trends in offensive output that have started after their peak offensive years a few years ago
  • Ryan Miller well have a slightly better year at the shootout (last year he had a -1.4 SGVT), and play more than 12 games less than he has in the past two seasons. Jhonas Enroth on the other hand will almost play 20 games and help lighten the goaltending load. It should be noted that when added their projected game totals together, only 73.4 of the games this season is predicted to be played by these two.
  • Derek Roy will struggle coming back from injury and will not be able to play a full season, yet will keep a similar career PPG average for the games he does play in
  • Luke Adam and Marc-Andre Gragnani will only play a portion of the season this year
  • Most of the other projections are similar to the player's career averages for seasonal play

I want to thank the people of Hockey Prospectus for their publication and work on the VUKOTA projection system. I highly recommend their publication, which has wonderful statistical analysis and interesting articles that will expand your view on the game and management of the NHL.

As for the DBTB readers, if you would like to suggest a stat for me to examine for this season, please let me know either by email or in the comment section below. For now though, I must conclude column #1 of Sab(re)metrics. Please tune in next week for my next piece: Sab(re)metrics: How Good Is Thomas Vanek?