The Puck Stops Here

One interest of mine is sabermetrics and hockey. I am interested in how much can be learned by applying statistical methods and common sense to the hockey statistical record. I think that the most important thing that must be kept in mind during these calculations is the uncertainties and errors that go into these calculations. Certainly, some can be learned, but we must always keep in mind that this is not an exact theory and if a calculation says player X was better than player Y, it may be because player Y was underrated by the formula or player X was overrated.

With that disclaimer, lets look at what has been attempted to rate players. This is something that is often done to shed light on arguments about which player was better or which players belong in the Hockey Hall of Fame. One example attempting this was put on hfboards by a poster called pnep. He has different formulas to rank forwards, defencemen and goalies. As a result, it leaves open the question of how to compare between these different positions and how to rank players (such as Mark Howe, Phil Housley or King Clancy) who spent some time playing in more than one position.

Nevertheless, let's look at his forward formula. Here it is:
"HHOF Monitor" formula for LW,C,RW =
Adjusted PTS/2 +
PO PTS +
Top 10 in Goals after Season - 5 pts +
# 1 in Goals after Season - 75 pts +
# 2 in Goals after Season - 50 pts +
Top in 10 Assists after Season - 5 pts +
# 1 in Assists after Season - 75 pts +
# 2 in Assists after Season - 50 pts +
Top 10 in PTS after Season - 5 pts +
# 1 in PTS after Season - 75 pts +
# 2 in PTS after Season - 50 pts +
# 1 in "+/-" after Season - 35 pts +
# 2 in "+/-" after Season - 25 pts +
Cup - 50 pts +
Final - 25 pts +
All Star Game - 20 pts +
HART - 150 pts +
HART Runner Up - 100 pts +
BYNG - 75 pts +
BYNG Runner Up - 50 pts +
1 ALL STAR TEAM - 75 pts +
2 ALL STAR TEAM - 50 pts +
CALDER - 35 pts +
CALDER Runner Up - 20 pts +
SELKE - 30 pts +
SELKE Runner Up - 20 pts +
CONN SMYTHE - 150 pts +
LESTER PEARSON - 100 pts

Adjusted points are calculated by the Total Hockey method. Notice that playoffs are very strongly emphasized. Playoff points (which are not adjusted in this formula) are worth double regular season points and team success such as winning the Stanley Cup or going to the finals are worth points. This will bias this formula to players who played with good teammates and thus had playoff success (example: Glenn Anderson) and against players who played on poor teams (example: Dale Hawerchuk). Further, this formula is biased toward modern day players because they play more games in the playoffs (and thus have a better chance of scoring playoff points) and also because the awards listed have not existed throughout the entire NHL (Pearson and Selke are the newest ones). This formula is also highly dependant upon other people's opinions. Much of a top player's point total will come from winning awards and making all star teams. The best player in the league may not win these awards. Sometimes the selection processes fail and come up with less deserving winners. An ideal formula would not require this. An ideal formula would be able to determine who should have won awards in a given season and not need any input from who the voters for fdifferent awards chose. Another problem is that any contribution made on the ice (aside from scoring) is very hard to measure and for the most part ignored in this formula. As a result, players with high defensive value will be underrated and players who scored well but had little value beyond their offense will be overrated. My final complaint is that totals are cumulative. If a player hangs around with marginal value for his team, he will continue gathering points, when a similarly able player retires and gets left behind. With all those complaints, this is an erxample of the "state of the art" in hockey sabermetrics at this point.

Here ared the top 10 forwards of all time by this method:
1. Wayne Gretzky 8827.02
2. Gordie Howe 6171.61
3. Mario Lemieux 4586.35
4. Jean Beliveau 4253.44
5. Phil Esposito 4096.48
6. Maurice Richard 3829.98
7. Bobby Hull 3671.88
8. Stan Mikita 3583.61
9. Jaromir Jagr 3237.64
10. Guy Lafleur 3060.04

Only NHL stats are used in these calculations (even though Gretzky, Howe and Hull played WHA).

Are these the top ten forwards in NHL history? Probably not. But it is one guess. There are many other ways to attempt to answer this question. Some of them may be more meaningful than this one and others may be less meaningful. If the same names keep on appearing at the top they are likely the best forwards in NHL history. When discrepancies occur between various methods, we can look further to investigate these discrepancies. Of course, the possibility exists that all methods haved the same bias and will overrate the same players and underrate the same others.

In the future, I will look at defencemen and goalies and the problem of calibrating their different totals and other attempts to answer this question (that others have tried). Finally, I will lay out what I would use as a method (were it possible) and explain why it is not possible at this time. Hopefully, these topics will fill most of the dog days of summer when there is little else to report upon in the hockey scene.