Estimated Ice Time

[Iain Fyffe]

This is what I mean by inspection. As I understand it, you're looking at a few numbers, taking what you know about the team, and assigning estimates that you think make sense. This is precisely why I asked you to do other examples, such as: the 1985/86 Hartford Whalers and the 1977/78 Cleveland Barons? Or, indeed, all NHL teams for 1992/93? If you use different criteria for different team-seasons, and even different segments within a team-season, I'd be hard-pressed to call it a "method". It appears to be an ad hoc designation of estimated ice time, rather than a calculation of estimated ice time.

You arrive at pretty good estimates for two reasons:

1. Being faimiliar with the game in general. As I said there are no big suprises revealed by these estimates, so anyone familiar with the game can produce decent estimates just by thinking about it for a bit.

2. Familiarity with the specific team. This is a key point, and how you can get estimates for a particular team-season beyond "decent" and into "pretty good."

But no one is familiar with all team-seasons. Any method that relies upon significant familiarity with the team in question will fail in comparison to the goals method, because the goals method can be applied to every NHL team from 1967/68 to 1997/98. You've done the defencemen for one team-season you're very familiar with. Do the 1985/86 Whalers and 1977/78 Barons and the entire 1992/93 NHL.

In terms of validity, the questions remains: how can you show your method is equally valid to the existing method, for all team-seasons under consideration?

[Canadiens1958]

Originally Posted by Iain Fyffe

This is what I mean by inspection. As I understand it, you're looking at a few numbers, taking what you know about the team, and assigning estimates that you think make sense. This is precisely why I asked you to do other examples, such as: the 1985/86 Hartford Whalers and the 1977/78 Cleveland Barons? Or, indeed, all NHL teams for 1992/93? If you use different criteria for different team-seasons, and even different segments within a team-season, I'd be hard-pressed to call it a "method". It appears to be an ad hoc designation of estimated ice time, rather than a calculation of estimated ice time.

You arrive at pretty good estimates for two reasons:

1. Being faimiliar with the game in general. As I said there are no big suprises revealed by these estimates, so anyone familiar with the game can produce decent estimates just by thinking about it for a bit.

2. Familiarity with the specific team. This is a key point, and how you can get estimates for a particular team-season beyond "decent" and into "pretty good."

But no one is familiar with all team-seasons. Any method that relies upon significant familiarity with the team in question will fail in comparison to the goals method, because the goals method can be applied to every NHL team from 1967/68 to 1997/98. You've done the defencemen for one team-season you're very familiar with. Do the 1985/86 Whalers and 1977/78 Barons and the entire 1992/93 NHL.

In terms of validity, the questions remains: how can you show your method is equally valid to the existing method, for all team-seasons under consideration?

This brings the discussion to the key question. If you could find a person for each NHL team that had my background and used my approach would the resulting TOI estimates hold for each team / each season, as well as a global league method like yours?

Both approaches are stalled by a common problem, the time and resources to run the data.

[Iain Fyffe]

Originally Posted by Canadiens1958

This brings the discussion to the key question. If you could find a person for each NHL team that had my background and used my approach would the resulting TOI estimates hold for each team / each season, as well as a global league method like yours?

Both approaches are stalled by a common problem, the time and resources to run the data.

Considering that the thread began with a discussion about an existing spreadsheet with the estimates you're talking about, claiming that the existing approach is stalled by time and resources seems fallacious, since it's already been used to calculate all the estimates that it is designed to calculate.

My own current variant of it hasn't been run for every team-season yet, but considering I have every NHL season from 1967/68 to 1978/79 complete, and then a few from the 80s and 90s as well, and the 1978/79 WHA season, and you have the defencemen for the 1970/71 Habs, we see which one is really constrained by time and resources.