Monday, November 16, 2020

Viktor Tikhonov vs Art Ross: Is pulling the goalie a sound investment?

 


Viktor Tikhonov vs Art Ross:
Is pulling the goalie a sound investment?

Scott A Hibbs

22 September 2020


Viktor Tikhonov vs Art Ross

Viktor Tikhonov and Art Ross were both multiple championship defenceman players. They were both multiple championship winning coaches. It is unknown if they ever met and neither would suspect they held opposing views on Hockey strategy. Art Ross is most known today for his donation of the Art Ross Trophy which is awarded to the leading scorer of the NHL regular season. Art Ross retired from the Boston Bruins in 1954 and died near Boston in 1964. (“Art Ross”, 2020)

That same year in 1964, Viktor Tikhonov started his coaching career winning multiple World Championships and Olympic gold medals and died in Moscow in 2014. Viktor Tikhonov is most known today as the dictatorial style, unbeatable coach of the Soviet Union for the game known as the “Miracle on Ice” in 1980. (“Viktor Tikhonov”, 2020)

When we look at the idea of pulling the goalie, it is Art Ross who is given the credit for this strategy being introduced into the NHL. It had previously been experimentally done in amateur and minor leagues. In 1931, as coach of the Boston Bruins this strategy was employed at the last minute during a playoff game against the Montreal Canadiens. Coach Ross was unsuccessful at this tactic, but the tradition was born. Most coaches still wait to the last minute or so to pull their goalie. In contrast, Russian and Soviet teams, most notably Viktor Tikhonov, never pulled their goalie. (“Extra Attacker”, 2019) In fact, one of coach Tikhonov’s players, Sergei Starikov is quoted as saying “Tikhonov just didn’t believe in it.” (Coffey, 2005)

It is unknown why Tikhonov didn’t believe in it, but a possibility could be his dominance of Russian and Soviet hockey didn’t require a losing endgame strategy. Maybe, he thought it was unprofitable. We have two champion coaches and two viewpoints on success, and both lost memorial games using their own respective strategy.

To be more successful, teams should seek out strategies that predict success and avoid those predictors that do not lead to success. Therefor the importance of pulling the goalie has serious real world business implications. This is why multiple studies have been done focusing on pulling the goalie. All have unique perspectives and various times and optimal situations to do so. If this strategy wins more games or yields a team to have a higher rank in the standings, then pulling the goalie should not be debated and progress can be made on how to best implement this strategy for greater success.

Each of this research is focused on how to pull the goalie successfully. One study by Clifford Asness and Aaron Brown titled “Pulling the Goalie: Hockey and Investment Implications” used their chosen fields of business, investing and risk management to formulate when coaches should pull the goalie in the 2015/16 season. They relate this strategy to an insurance move and a proper risk measure, yet suggest that the optimal time to pull the goalie is with 6:10 left to play the game. (Asness & Brown, 2018)

All teams pull the goalie as the only strategy when behind during regular play. This is usually not seen in overtime. The NHL rules discourage pulling the goalie by not rewarding the losing team with any game points if they had the goalie pulled.
Asness and Brown agree with prior research that pulling the goalie should be done earlier (not the last minute), however their conclusion to play without a goalie for the last six minutes and ten seconds of a loosing game seems overkill. Restated simply, their conclusion is that games can be tied if the losing team will play without a goalie for over 30% of the last period. Double this time if the losing team is down two points. If this were true, maybe they have just proven teams don’t need goalies
. (Asness & Brown, 2018)

Instead, maybe Coach Tikhonov is right after all. However, no research has been done on the effectiveness of this strategy. Validating this business practice based upon profitability (ranking or wins) will allow a team can be more innovative and improve it’s endgame approach. A team can increase championship winnings and improve performance if the decision to pull or not pull the goalie can be determined to be a significant predictor of overall success.

The question is how to do we go about proving a strategy where there is no direct statistics given for when pulling the goalie was able to tie the game? Then which of these tied games actually led to a victory? Is there a success rate for this? It isn’t unusual for the losing team to have an across ice long goal scored against them from the far end into the empty net. Thus, scoring by the winning team can be done without any skating at all. The only risk for the winning team (other than defending against an extra attacker) is icing if they miss the shot. So without direct numbers on games where pulling the goalie was successful, (and yet led to a victory) we will have to find other predictors.

Using numbers from the “More Hockey Stats” web page we can gather how many positive goals were scored when the goalie was pulled for each team. “More Hockey Stats” also has the number of times each team pulled their goalie. (“Empty Net Performance stats for NHL Teams”, 2020) With these two numbers we will determine a success rate for pulling the goalie. To determine this success rate, we will divide the number of positive goals when the goalie was pulled by the count of times the goalie was pulled for each of the thirty teams. Since rankings are more suggestive, wins should provide a more accurate predictor. The 2015/16 regular season stats for each of the thirty teams in the NHL will be used such as team rankings and the number of wins. (“Standings”, n.d.)

We should see a positive correlation between the pulling goalie success rate to wining hockey games or higher rankings, and show that Coach Ross has the right idea. Without this correlation, Coach Tikhonov will be more correct and the success rate of pulling the goalie is not a predictor of wins or rankings. So which winning coach is more correct? Is the success rate of pulling the goalie (empty net positive goals / Count of pulling goalie) a significant predictor of successful team ranking or total number of games won?

This is the data we need for the 2015/16 season.

Ranking

Team

# Wins

Scored w/EN

Count w/EN

Success Rate

1

Washington

56

4

18

0.2222

2

Dallas

50

5

19

0.2632

3

St. Louis

49

3

21

0.1429

4

Pittsburgh

48

4

26

0.1538

5

Chicago

48

6

25

0.2400

6

Anaheim

47

3

21

0.1429

7

Florida

47

4

24

0.1667

8

Los Angeles

46

7

28

0.2500

9

NY Rangers

46

5

26

0.1923

10

NY Islanders

46

2

23

0.0870

11

San Jose

46

1

28

0.0357

12

Tampa Bay

45

5

35

0.1429

13

Philadelphia

42

4

22

0.1818

14

Nashville

41

6

28

0.2143

15

Detroit

41

10

31

0.3226

16

Boston

41

1

28

0.0357

17

Minnesota

39

4

35

0.1143

18

Carolina

38

5

22

0.2273

19

Ottawa

38

7

37

0.1892

20

New Jersey

38

8

36

0.2222

21

Colorado

38

8

39

0.2051

22

Montreal

35

2

33

0.0606

23

Buffalo

35

5

30

0.1667

24

Arizona

35

6

33

0.1818

25

Winnipeg

35

2

37

0.0541

26

Calgary

35

5

35

0.1429

27

Columbus

34

7

38

0.1842

28

Vancouver

31

5

31

0.1613

29

Edmonton

31

5

31

0.1613

30

Toronto

29

12

39

0.3077

Summery Output for Rank (using linear regression)

R^2

0.005011629750701

Standard Error

0.073052255769934

Count of X variables

1

Observations

30

Adjusted R^2

-0.030523669186774

Analysis of Variance (ANOVA)








df

SS

MS

F

Significance F


Regression

1

0.000753

0.000753

0.141032

0.710087


Residual

28

0.149426

0.005337




Total

29

0.150178













Coefficients

Standard Error

t-Statistic

P-value

Lower 95%

Upper 95%

Intercept

0.181384

0.027356

6.630475

3.41474E-07

0.125347

0.237420

Rank

0.181384

0.027356

6.630475

3.41474E-07

0.125347

0.237420





Summery Output for Wins (using linear regression)

R^2

0.000983969205835

Standard Error

0.073199962286726

Count of X variables

1

Observations

30

Adjusted R^2

-0.0346951747511

Analysis of Variance (ANOVA)








df

SS

MS

F

Significance F


Regression

1

0.000148

0.000148

0.027578

0.869298


Residual

28

0.150031

0.005358




Total

29

0.150178













Coefficients

Standard Error

t-Statistic

P-value

Lower 95%

Upper 95%

Intercept

0.158429

0.085269

1.857977

0.073717

-0.016238

0.333095

Wins

0.000341

0.002054

0.166067

0.869298

-0.003866

0.004549


Linear regression is the best way to find a correlation and therefor a predictor. It was used to show the success rate for pulling the goalie for each team (our independent variable) to predict the final ranking of each team for the year (our dependent variable). For comparison and to be sure of any correlation, we used linear regression to show the success rate for pulling the goalie for each team (our independent variable) to predict the total number of games won for the season for each team.

The data did not show a significant correlation. The R squared is barely above zero. There was hardly any correlation at all. The p value for each is different. When comparing to team rankings we got a low r^2 and a low p value which means we have enough data and still don’t have a correlation. When comparing to the number of wins we get a low r^2 value and a high p value which indicates there is no correlation, but there may not be sufficient data to support the lack of correlation.

Although outside the scope of this paper, and given that we might not have enough data, a quick look at combining the 2014/15, 2015/16, 2016/17 seasons together for 90 data points was completed.

RANK

# OF WINS

R^2 = 0.00254

R^2 = 0.0365

P = 0.134

P= 0.071



Thus, we get better p values when we add more data. However, there is still no correlation between our success rate when compared to rank or wins, even over 3 years together. We can safely say that from season to season, pulling the goalie is not a predictor of winning games or a predictor of final NHL team rankings.

There is an obvious correlation to winning teams having fewer number of times that they pull their goalies because they tend to be winning at the end of the game. This is the most likely reason Coach Tikhonov didn’t believe in the strategy. However, if Coach Ross’s strategy helps, at all, there should be some correlation and we found none.

For example, the Toronto Maple Leafs finished last in the rankings and had the fewest winning games. Their success rate at pulling the goalie was 30%. (The second place team, Dallas Stars had 26%.) Toronto scored the most goals with an empty net and they pulled the goalie the most. Over all, it didn’t help them avoid last place.

Our data may have power plays counted as times with an empty net. When a team has a delayed penalty, that is they fouled another player with the puck, the play does not stop until the penalized team touches the puck. When the referee holds up his hand, he’s waiting for the other team to touch the puck. It is these moments (unpredictable time) when the goalie’s value is zero. It is only then that the goalie is not important since the other team can not score. Savvy teams pull their goalie in this instance for an extra attacker. Although our analysis includes possible counts playing and possible goals scored with an empty net for delayed penalties, there was still no correlation. With this into account, playing with an empty net is not a predictor of wins.

For a comparison, we can look at a few stats of playing with an extra attacker, such as power plays. Once a penalty is called, the penalized team sends a skater to the penalty box where they will count as playing. Thus providing the other team with an extra attacker for 2 minutes or more or until a goal is scored. The average power play percentage success rate is on average is 20% (thus the penalty kill percentage averages about 80%), it seems only slightly easier to score with an extra attacker in regular play.

Goalies have a tough role to play. The rule book states there are two positions for a team. 18 skaters and 2 goalies. They have the only role on the ice where their failure is the most evident. If a skater is beaten that results a score against the goalie, it’s usually the goalie and not the skater that everyone sees. Almost every goalie’s save percentages are above 90%, meaning less than 10% of the shots on goal get past them on average. With them sharing an 80% success rate against an extra attacker, it makes goalies very important.

There must be other strategies to try other than pulling the goalie. It could be suggested to pull both defenceman instead. Play only one defenceman and have 4 forwards is another option to try. There are many system strategies like the torpedo system. However, until we stop pulling the goalie, we can not try other strategies. These other strategies might become a predictor of winning games and success.

In conclusion, unless it is a delayed penalty when the goalies value is zero, it seems pulling the goalie must be a bad investment, a horrible insurance policy, and not worth the risk.


References


Art Ross, (2020) In Wikipedia. Retrieved September 23, 2020, from https://en.wikipedia.org/wiki/Art_Ross

Viktor Tikhonov, (2020) In Wikipedia. Retrieved September 23, 2020, from

https://en.wikipedia.org/wiki/Viktor_Tikhonov_(ice_hockey,_born_1930)

Extra Attacker, (2019) In Wikipedia. Retrieved September 23,2020, from

https://en.wikipedia.org/wiki/Extra_attacker

Coffey, Wayne (2005) The Boys of Winter (E ed.). New York: Crown Publishers.

ISBN 0-307-23731-1. page 384

Asness, C.S., & Brown, A. (2018). Pulling the Goalie: Hockey and Investment Implications. Behavioral & Experimental Finance eJournal.

Empty Net Performance stats for NHL Teams, (2020) In More Hockey Stats. Retrieved, September 23, 2020, from https://morehockeystats.com/teams/en

Standings, (n.d.) In NHL.com. Retrieved, September 23, 2020, from

https://www.nhl.com/standings/2015/league


Biography

Scott A Hibbs

Scott fell in love with Hockey while attending the University Of Kentucky and became a fan of the club team. A life long computer geek, he is currently attending WGU to achieve his MBA in IT Management. He enjoys hockey, chess and amateur radio. He is married, lives in Indiana and has two grown kids.

No comments: