View Single Post
Old 12-11-2017, 12:05 PM   #36
united
First Line Centre
 
united's Avatar
 
Join Date: Mar 2007
Exp:
Default

Quote:
Originally Posted by mrkajz44 View Post
The reason for the spread is all based on method used.

I know Sports Club Stats uses goal differential as a predictor for future match-ups. The Flames are currently -6 in that category, so the simulator does not think Calgary is very good and will lose more games int he future.

I would guess the Athletic and Corsica hockey are likely using some form of advanced stats, which show the Flames are underperforming a bit. Hence, these guys give the Flames a much better chance as they are winning more in their future simulated games.

Hockey Viz and Money Puck, I don't know, so I can't say for sure.

Long story short - all these numbers are "good" so long as you think their forecasting method is accurate. Think goal difference will predict future performance? Use Sports Club stats. Think advanced stats can predict future results? Use Corsica hockey. Think all this prediction stuff is hooey and you just need a set amount of points at season's end? Use the snake.
This is exactly it.

The Athletic is player advanced stats-based. Player x has advanced stats of x and he plays x% of the team's ice time. Combine all players on a team for team strength of x. This also accounts for injuries.

Corsica and HockeyViz use team-level shot-based advanced stats.

MoneyPuck uses team-level expected goals-based advanced stats, discounting the first 20 games of the season and weighing all overtimes as 50/50. So their model is using an extremely small sample size at this point of the season.

Hockey Reference and Sports Club Stats use goal differential-based models.

Or:
Advanced stats models
63%
59%
55%
42% (first 20 games discounted)

Non-advanced stats models
40%
33%

This early in the season, there is far too much volatility in goal-based models thus I wouldn't put too much weight into them, personally. But some hate all things advanced stats so if that applies, discount those models completely. In the long run - several hundred team level games - the shot-based and goal-based models will closely match...but obviously seasons are far shorter than that.
__________________
"I think the eye test is still good, but analytics can sure give you confirmation: what you see...is that what you really believe?"
Scotty Bowman, 0 NHL games played

"You ain't gotta like me. You're just mad 'cause I tell it how it is and you tell it how it might be."

Last edited by united; 12-11-2017 at 12:10 PM.
united is offline   Reply With Quote