thoughts (some even make sense).
If you like this article, check out Adjusted Winning Percentage.
Why .500 Isn't .500 Anymore
It used to be so simple. Two points for a win, one point for a tie, none for a loss. For a hundred years, that's how it worked. Then, the NHL changed the rules. If you lose in overtime, you get a point. Aside from the aspects of how a game is played, this has radically altered the concept of winning percentage.
Let's start with a basic understanding of what a "winning percentage" is (and why it matters). To calculate the winning percentage for a team, simply take the number of points they have earned and divide by the number of possible points they could have earned. (And you thought algebra would never come in handy.)
For example: in 1974-75, the Canucks finished with 86 points in 80 games. Back then, this was a big accomplishment. It marked the first time in their brief history that the Canucks hit the .500 mark. It demonstrated that the Canucks were average. After all, they had won more games than they lost.
For many years, the Canucks did not finish above .500. Between 1976-77 and 1990-91, the Canucks were below average in the standings. It was a dubious streak which put them in the same company as some of the worst franchises in professional sports. When the Canucks finished with 96 points in 1991-92, it was cause for celebration. After all, the team had finally risen above the .500 mark.
Now, we play in an era of "overtime losses" (referred to as "regulation ties" during their inaugural season). In 1999-2000, the Canucks finished with a record of 30-37-15-8, good for 83 points in 82 games and a .506 winning percentage (at least according to The Hockey News). But, wait a minute. The Canucks lost seven more games than they won. Those eight "regulation ties" added eight points to their total and pushed them over the .500 mark.
My theory is that you need to take the 83 points the Canucks earned and divide it by the maximum number of points an average team could have earned. (Here comes some more complicated math.) In most games, two points are awarded (two points for a win or one point each for a tie). In some games, however, three points are awarded to the teams involved (two points for an overtime win, one point for an overtime loss). This skews the .500 mark away from the 82 points in 82 games concept. (It's like when your teacher gave you a test and someone scored 105% because there was a "bonus question" at the end.)
During the 1999-2000 season, a total of 2,410 points were awarded to the teams (including 114 points for overtime losses). If you divide the 2,410 points into 28 teams, you get about 86 points (86.07 to be slightly more accurate). Therefore, the .500 mark must be raised and the Canucks finished with a winning percentage of .482.
There's one more reason why the .500 mark is important. Prior to the 1999-2000 season, a .500 record would put you in the middle of the pack. In any given season, half the teams will finish above .500 and half will finish below .500. Although a really good or really bad team can shift this balance, it's usually very close. For example: in the 1998-99 season, 14 teams finished with more than 82 points and 13 teams finished with less than 82 points.
Compare that to the 1999-2000 season when 20 of 28 teams finished with 83 points or more. How can you consider 82 points to be the .500 mark? Remembering my calculation, you would find that 15 of the 28 teams finished with more than 86 points. That's pretty close to half (it probably would have worked out if Atlanta hadn't lost so many games).
One final thought: I can understand why the NHL would not want to calculate winning percentages back in pre-calculator days. Have you ever tried to divide 86 / 160 by hand? (Go ahead and try ... I'll wait.) But in this era of computers, surely someone (other than me) can figure out how to calculate these numbers on a regular basis. Then again, this is the same league that can tell you a certain forward played 20 minutes and 13 seconds per game, but still rounds goaltender times to the nearest minute.
copyright © 2001-2010 David Marchak
This page last updated July 28, 2010