February 17, 2010
From Daigle To Datsyuk
The Tilted-Ice Effect

by Corey Pronman

Throughout the last few years, hockey has made some significant advancements in the field of statistical analysis. Bloggers and experts alike have combined to create new and creative metrics to help us better evaluate the game. From Corsi to On/Off Ice +/-, from Quality of Competition to GVT, and from Player Contribution to Shot Quality, this last decade has brought some significant breakthroughs. All of these stats have helped us look at the offensive and defensive side of a player’s contributions differently. However, with all of these new metrics, when we look at a player’s statistics and when we analyze the offensive or defensive side of the ice, are we missing a key factor?

The game of hockey is so much different than a sport such as baseball. Baseball is a series of 1 on 1 matchups, whereas hockey is continuous, the game flows and is much more open. The most common criticism of hockey metrics in comparison to baseball is that it doesn’t isolate everything, which to some degree is true, but statisticians have made great strides in isolating what is important. However, there is still one glaring error in hockey metrics that needs to be addressed. Let's use baseball for comparisons sake. The game of Baseball has two phases to its game: hitting/baserunning and pitching/defense. No matter what any of the pundits will tell you or anyone else for that matter, what you produce when you’re hitting has a very, very minute effect on your pitching and fielding production. In hockey, this is not the case. Your offensive production is directly tied to your defensive production and your defensive production is directly tied to your offensive production.

What does this mean? Well think of it this way. If you look at a player’s Goals For On Ice/60 or Shots For On Ice/60 statistics, you will notice that there is a decrease in value for every second that you either aren't putting a shot on net or scoring a goal. Knowing this, as long as the puck is in your defensive zone, those rate statistics will continue to decrease. So the longer the puck stays in your defensive zone, the longer your offensive metrics continue to take a hit. If you are an extremely talented offensive player, but you screw up on a breakout or as a winger, or miss your assignment as a defenseman at the point, which results in the puck being held in, your offensive metrics will decrease. You can be an amazing offensive player, but if you never get the puck into the offensive zone what good does that do? On the flipside, if you are a good offensive player, your defensive metrics will most likely increase. So if you're able to keep the puck in the offensive zone, your Goals Against On Ice/60 and Shots Against On Ice/60 will both decrease. I call this the Tilted-Ice Effect, which is the result of increasing or decreasing your contributions at one end by keeping the puck in the other. The Tilted-Ice Effect essentially states:

• A good offensive player, one who can continue to apply offensive pressure onto an opponent, is in fact a good defensive contributor.
• A bad offensive player, who cannot apply offensive pressure onto an opponent, is in fact a poor defensive contributor.
• A good defensive player, who can keep the opponent from applying offensive pressure is in fact, a good offensive contributor.
• A bad defensive player, who cannot keep the opponent from applying offensive pressure is in fact, a poor offensive contributor.

This does not mean that a good offensive player is necessarily a good defensive player, but rather that his offensive contributions create positive contributions on the defensive end. He could be a poor defensive player too, which in turn may lower his offensive contributions, which is why statistics like On/Off Ice +/- or Corsi Relative are more reliable statistics because they encompass a player’s total contributions. However, the Tilted-Ice Effect can only be applied when analyzing one side of the ice, be it a player’s offensive or defensive metrics.

Also keep in mind that the key word in the bullet points above is pressure. Simply going into the zone, and chucking the puck at the net for an easy save that results in a turnover, thus raising your Shots/60, is not applying offensive pressure. This is why it’s hard to completely isolate and understand the Tilted-Ice Effect because we don’t have a Time on Attack metric available to us.

To try and demonstrate this effect, let's first take a look at the shot differentials for every team in the NHL:

Team 	S/G 	SA/G 		Diff	Abs Diff
WSH	32.6	31.1		1.5	1.5
CHI	33.6	24.3		9.3	9.3
SJS	32.0	31.4		0.6	0.6
LAK	28.9	27.6		1.3	1.3
NJD	30.3	27.5		2.8	2.8
PHX	29.6	29.4		0.2	0.2
PIT	32.4	29.3		3.1	3.1
VAN	30.2	28.9		1.3	1.3
OTT	29.7	28.4		1.3	1.3
BUF	31.3	31.6	       -0.3	0.3
COL	27.0	31.8	       -4.8	4.8
NSH	30.4	29.0		1.4	1.4
CGY	28.3	29.2	       -0.9	0.9
MIN	28.7	28.6		0.1	0.1
MTL	27.8	32.8	       -5.0	5.0
PHI	31.7	28.3		3.4	3.4
DET	32.7	30.2		2.5	2.5
ANA	29.0	33.0	       -4.0	4.0
NYR	29.9	29.7		0.2	0.2
DAL	31.4	30.3		1.1	1.1
STL	28.7	30.1	       -1.4	1.4
ATL	29.8	33.1	       -3.3	3.3
TBL	28.9	30.9	       -2.0	2.0
CBJ	29.2	29.7	       -0.5	0.5
FLA	28.4	34.2	       -5.8	5.8
BOS	31.0	29.7		1.3	1.3
NYI	30.4	31.2	       -0.8	0.8
CAR	29.0	31.4	       -2.4	2.4
TOR	33.1	29.9		3.2	3.2
EDM	28.5	31.9	       -3.4	3.4

The average here for the Absolute Differential is 2.31, which means that every team averages a 2.31 shot slant in some direction per game. It’s not huge, but it means that the game is always tilted in one direction. So what? A game has to be tilted in a direction for a team to win. Well applying this to baseball, I took the American League teams Hits/9 and Hits Against/9 and did the same calculations (mind you I only did it for the American League) and the average difference was 0.33. As discussed before, baseball is a game without a tilt, though it is a game with phases and it seems when you average the hits differential between phases, the spread is very small and almost non-existent. Also while shots don’t indicate the full tilted-ice effect, it shows something, that we’re on the right track at the very least.

Now let's take a look at several players to try and look for the effect:

From the New York Rangers, here are the top five forwards in Shots For on Ice/60 at even strength.

Forward                 SFON/60
Sean Avery		30.4
Brandon Dubinsky	30.1
Marian Gaborik		29.7
Vinny Prospal		28.6
Ryan Callahan		27.7

And now the bottom five in Shots Against on Ice/60:

Forward                 SAON/60
Sean Avery		23.9
Vinny Prospal		25.6
Artem Anisimov		25.8
Enver Lisin		26.1
Ryan Callahan		27.2

*Dubinsky came in at 6th

*Restrictions are 20 games played with at least some substantial ice time

There is a bit of a correlation here, with Sean Avery being the top performer at both ends of the ice, while Callahan and Dubinsky come around the 5 spot for Shots Against on ice per 60 minutes at even strength.

Doing this again for Vancouver, here are the top five in SFON/60:

Forward                 SFON/60
Mason Raymond	        32.5
Ryan Kesler		31.4
Daniel Sedin		31.1
Alex Burrows		30.2
Mikael Samuelsson 	29.6

And now SAON/60:

Forward                 SAON/60
Mikael Samuelsson	25.0
Ryan Kesler		25.2
Mason Raymond	        25.3
Daniel Sedin		25.9
Kyle Wellwood		26.1

*Burrows was 6th

Obviously a very large correlation here, all of the top five in shots for appear in the top six of shots against.

Again for Boston, SFON/60:

Forward                 SFON/60
Patrice Bergeron	30.8
Mark Recchi		29.8
Michael Ryder		28.9
Marco Sturm		28.9
David Krejci		28.9\

SAON/60:

Forward                 SAON/60
Patrice Bergeron	22.1
Mark Recchi		22.2
Vladimir Sobotka	24.3
Daniel Paille		24.9
Shawn Thorton		25.1

Some correlation here again. The top two in shots for are the top two in shots against. However, while some teams show a correlation, there are others that don’t, case and point the Montreal Canadiens, SFON/60:

Forward                 SFON/60
Brian Gionta		28.9
Scott Gomez		27.7
Benoit Pouliot		27.6
Michael Cammalleri	26.5
Tomas Plekanec		25.3

And SAON/60:

Forward                 SAON/60
Andrei Kostitsyn	27.0
Glen Metropolit		28.0
Matt D’Agostini		28.6
Max Pacioretty		28.8
Tomas Plekanec		28.9

Woah, asides from Plekanec, the top five in Shots For on ice per 60 minutes at even strength are not in the top five for shots against on ice per 60 minutes at even strength. Why is this and where iss the tilted-ice effect? Well having watched extensive coverage of the Habs, Gomez specifically tends to be a chucker. He enters the zone and throws the puck on net, thus raising his SFON/60 as well as his linemates SFON/60, like Gionta, to look much better than it actually is. His line also doesn’t gain the zone well and turns it over quite a lot which is the main reason both Scott and Brian are the bottom two in SAON/60.

The theory isn’t perfect, because there are things you need to remember:

1. Using shots is an imperfect way of evaluating tilted-ice, time on attack if it ever becomes tracked, would be much more preferable.
2. It’s not a huge impact, though the impact is there. However, as illustrated above, an average of roughly 30 shots per game are taken in either direction, and the slant in either direction is a little over two. In exact numbers, the slant is 13.05% of the average shot totals in either direction.
3. Using any per 60 minutes metric puts a huge reliance on the players around you. They are all imperfect metrics that fail to isolate the individual and thus raises the margin for error.

Due to these points, some teams have a more minute tilted-ice effect or none at all using the methods I illustrate above.

So while there are errors in the theory, the effect is there, and just using basic hockey logic can uncover this and there is some empirical evidence to point to it as well. This entire concept is just to give you some food for thought for the next time you take a look at certain metrics and just another step in trying to develop the highest-end statistics in a sport that makes it very hard on those that attempt to do so.

Special thanks to Gabriel Desjardins and Behind the Net for some of the statistics used in this column.

Follow Corey on Twitter at @coreypronman.

Corey Pronman is a contributor to Puck Prospectus, an Associate Scout for the USHL Sioux Falls Stampede and runs the statistical hockey site The Hock Project. You can contact him at CPronman@fau.edu.

Corey Pronman is an author of Hockey Prospectus. You can contact Corey by clicking here or click here to see Corey's other articles.