April 2, 2009
Hockey's Different Games

by Gabriel Desjardins

Let’s start with a simple hypothetical. Imagine you have never been to a hockey game before. The referee drops the puck and five on five play begins with both teams forechecking, playing defense and passing the puck with a few shots per minute on each netminder. Suddenly a penalty is drawn, completely changing the dynamics of the game. Each team cherry picks its own players for the power play and the penalty. The team with the power-play controls the puck and sets up in the offensive zone. The penalty-killing team plays zone defense, hoping to block or deflect shots in order to clear the puck out of their territory.

My take is that hockey is actually a number of different games where individual play and team play need to be analyzed separately. Like in other sports, NHL strategy is heavily dependent upon the score of a game and the time remaining in the particular contest. However, hockey is distinctively unique from the other major North American professional sports for forcing teams to play short-handed when a player draws a penalty. Here are the league-wide offensive and defensive statistics for each strength during the 2008-09 season:

POT: Percentage of Time

GF/60: Goals For/60

GA/60: Goals Against/60

SF/60: Shots For/60

SA/60: Shots Against/60

S%A: Save Percentage Against

S%F: Save Percentage For

        POT	GF/60	GA/60	SF/60	SA/60	S%A	S%F

5v5	74.7	2.45	2.45	29.3	29.3	916	916
5v4	10.5	6.50	0.91	51.1	8.4	873	892
4v5	10.5	0.91	6.50	8.4	51.1	892	873
4v4	3.1	3.08	3.08	31.9	31.9	903	903
							
5v3	0.38	20.25	0.12	98.9	1.5	795	917
4v3	0.18	10.96	0.25	74.4	4.8	853	947
3v5	0.38	0.12	20.25	1.5	98.8	917	795
3v4	0.18	0.25	10.96	4.9	74.4	947	853
3v3	0.014	10.11	10.11	37.1	37.2	727	728

The graph above illustrates the notion of hockey being vastly different depending on how many skaters are in the penalty box. The number of Shots For per 60 minutes and Shooting Percentage ascend rapidly as there are fewer defenders on the ice, particularly with only three on the short-handed squad. Incidentally, I am often asked which teams do well at 5-on-3 play. I am left with responding that at 12 seconds per game, it’s difficult to determine if it’s a separate skill from a 5-on-4 play.

Let’s take a look at the two most common situations: the scenarios of even strength play, 5-on-5, and a one man advantage, 5-on-4. It’s rather conspicuous to any hockey fan that it’s easier to score on the power-play than at even-strength: the ratio of goals for to goals against climbs from 1:1 to 7:1. A player who is on the ice for 18 minutes per game with five minutes on the power-play unit will score 45% more than a fellow teammate who receives the same amount of ice time without getting the opportunity to play on the power-play. We know this implicitly, yet we evaluate players all too often on their so-called “boxcar” statistics, goals-assists-points (G-A-P). Let’s take a look at an example from this season:

2008-09 Stats	        GP	G	A	PTS

Martin St. Louis	74	28	46	74
Louie Eriksson	        74	33	24	57

Clearly, Marty is the more productive player of the two. When we dissect the players based on Time-On-Ice (TOI) in different games we see a different picture:

P: Points

PP60: Points Points Per 60 minutes

ESP/60: Even Strength (5v5 Points/60 minutes)

ESTOI: Even Strength (5v5) Time On Ice

PPP/60: Power Play (5v4) Points/60 minutes

PPTOI: Power Play (5v4) Time On Ice

                         P	PP60	ESP/60	ESTOI  PPP/60	PPTOI

Martin St. Louis	74	2.85	2.44	14:01	3.82	4:08
Louie Eriksson	        57	2.19	2.38	13:57	2.26	2:57

At 5-on-5, even strength play, Eriksson and St. Louis have very similar offensive production. St. Louis’ extra points come on power-play opportunities. He’s better on the power-play, ergo more minutes and time on the first-line power-play. What is not evident yet is that the gap between the two of them is significantly smaller than basic statistics suggest. To be fair, we’ve only had data like this for the past few years. It is certainly not surprising that most people defer to G-A-P as a statistical preference. As the result of the dearth of statistics available in the past, this is the only way to view eras in context. It also delineates a lot about a player’s career progression, his offensive abilities and whether he received plenty of power-play time.

Now that we have better data provided, we can start analyzing the last handful of NHL seasons. Not only can we break down a player’s offense by his opportunities for the first time, but we can also analyze variables such as individual defense that have eluded us for some time. Next time, I’ll take a look at some basic improvements to the plus-minus rating with an explanation of what this statistic can tell us about hockey activity on both sides of the puck.