Hockey Outsider
Registered User
- Jan 16, 2005
- 9,428
- 15,558
On the topic of goals versus assists, I did a deep dive into this a few years ago.
My general conclusion is primary assists have nearly the same value as goals, and although secondary assists have less value than either, points are still the best "simple" method to evaluate a player's offensive contribution. (You can start assigning fractional weights to goals, primary assists and secondary assists, but you're making things much more complex and gaining fairly little value).
The article went like this - generally, something has value if it can be repeated from year to year. Using 1,200+ data points, I found that goals have virtually the same predictive value, year over year, as primary assists. And although secondary assists have relatively little predictive value on their own, total assists are more predictive than just goals. And total points is more predictive than anything else. This is statistical evidence that points (by virtue of being more predictive from year to year) are more valuable than either goals or assists in isolation. Link
I also did another study where I looked at what stats correlate with a forward's salary. Total points produced a higher correlation than either goals or assists in isolation - meaning that, all other things being equal, points are a bigger drive of a forward's compensation than either goal-scoring or playmaking in isolation. Granted, we don't know if the GM's always make good decisions. But they're paid a lot of money, and are under intense scrutiny from owners and from fans. So if points correlates more strongly than goals (or assists) in isolation, that's a pretty good argument for it's value. It's possible that every GM is wrong and they're just following tradition - but in this age of advanced analytics, if GM's were overpaying for assists, surely a few teams would realize this and stock up on goal-scorers, who would be comparatively undervalued - and there's no evidence this is happening.
Obviously, most fans look at points because that's what the NHL has prioritized for more than a century. But there's statistical evidence that, in general, points have more informational value than either goals or assists in isolation.
My general conclusion is primary assists have nearly the same value as goals, and although secondary assists have less value than either, points are still the best "simple" method to evaluate a player's offensive contribution. (You can start assigning fractional weights to goals, primary assists and secondary assists, but you're making things much more complex and gaining fairly little value).
The article went like this - generally, something has value if it can be repeated from year to year. Using 1,200+ data points, I found that goals have virtually the same predictive value, year over year, as primary assists. And although secondary assists have relatively little predictive value on their own, total assists are more predictive than just goals. And total points is more predictive than anything else. This is statistical evidence that points (by virtue of being more predictive from year to year) are more valuable than either goals or assists in isolation. Link
I also did another study where I looked at what stats correlate with a forward's salary. Total points produced a higher correlation than either goals or assists in isolation - meaning that, all other things being equal, points are a bigger drive of a forward's compensation than either goal-scoring or playmaking in isolation. Granted, we don't know if the GM's always make good decisions. But they're paid a lot of money, and are under intense scrutiny from owners and from fans. So if points correlates more strongly than goals (or assists) in isolation, that's a pretty good argument for it's value. It's possible that every GM is wrong and they're just following tradition - but in this age of advanced analytics, if GM's were overpaying for assists, surely a few teams would realize this and stock up on goal-scorers, who would be comparatively undervalued - and there's no evidence this is happening.
Obviously, most fans look at points because that's what the NHL has prioritized for more than a century. But there's statistical evidence that, in general, points have more informational value than either goals or assists in isolation.
