Copyright Iain Fyffe, 2002
One of the most widely-held convictions (which is to say, myths) in hockey is that "defence wins championships". For instance, during the 2001 NHL All-Star Game broadcast on CBC on February 4, Pat LaFontaine said that "defence wins championships", and noted that "everybody" knows this. It may seem obvious to anyone who actually thinks about this statement that it is probably false. But let's see if we can demonstrate this with some analysis.
The contention is apparently that defence is more important than offence in winning championships (does no one remember the Oilers and Penguins?). Just thinking about this assertion leads to some conclusions: defence shouldn't be any more important than offence; the important thing would seem to be the difference between the two. That is, the most important thing in winning a hockey game is scoring more goals than your opponents do. This seems fairly obvious, since it is the factor by which the winner of a game is actually determined.
So this is my prediction: goal differential (GD) is more important than both goals for (GF) and goals against (GA) in predicting champions. To test this, I examined every NHL season from 1926/27 to 1999/2000; the starting point was chosen because it is when the NHL took exclusive control of the Cup. We therefore will not have to compare teams from different leagues in doing this analysis. For each of these 74 years, I noted the following: the Cup-winning team, the number of teams in the league, the winner's rank in GF, the winner's rank in GA (lowest GA being best), and the winner's rank in GD.
Using this data, we can make some observations simply by inspection: The team with the league's best offence won the Cup 24 of 74 years; the best defence won 26 of 74, and the best GD won 29 of 74. This supports my prediction. For Cup winners, the most important factor (that is, the best-ranked factor) is distributed as follows: GF 40, GA 44, and GD 46. This adds to greater than 74 due to ties. This is more support, albeit slim, for my prediction.
Now I hear you all yelling, "Come on, Iain! Get to the real numbers!" Well here you go. Before describing what I did, I'll cover what I didn't do. One thought was to compare the champions' GF, GA, and GD to the league averages. This won't work, of course, because the average GD for any year must be zero. We could compare the champion's GF, GA and GD to the league's best GF, GA and GD that year. This cannot be done either, because while GF and GA have a downward limit (zero), GD can be negative, with no absolute limit. This prevents any direct comparison of the numbers. So, we must use rankings to compare them. I did this using percentiles.
For each Cup winner, I calculated their percentile scores for each of GF, GA, and GD using their ranks in each category. For example, a team's percentile in GF is calculated as follows:
GFRnk / TMS x 100
Where GFRnk is the team's rank in GF, and TMS is the number of teams in the league.
Thus, if a team ranked first out of 10 teams, it was in the top ten percent, while if it ranked tenth, it was only in the top 100 percent. The lower the percentile, the more important the factor is in predicting championships. I then calculated the average percentiles for each category. The results:
That is, the average champion ranks in the top 28.7% in GF, the top 28.2% in GA, and the top 22.2% in GD. Notice that while GF and GA are virtually identical, GD is significantly better-ranked, and is therefore more important.
In conclusion, no matter how you look at it, defence by itself does not win championships. Nor does offence. It's the ability to outscore your opponents by the greatest degree that matters. This may seem so obvious that it's not even worth noting, but given the comments of LaFontaine and the majority of sportscasters, it needs to be said, and said often.