View Single Post
07-22-2011, 01:05 PM
#8
DAChampion
Registered User

Join Date: May 2011
Location: Canberra, Australia
Country:
Posts: 4,816
vCash: 500
Quote:
 Originally Posted by MathMan With p = points/games? And n = games? That gives you a n under 100. That's really tiny. Not to mention that points/game is not a binary probability. Actually it would seem to me that p = goals/shots would make more sense.
If he has 42 points, the 68% confidence interval is 6.4 points.
Points per game is just something converges faster. 42/87 = 0.48 ppg with an error of 0.07 ppg.
However, if he had 420 points over 870 games, the error on 420 would be 20.4, and the ppg would be 0.48 with an error of 0.02 ppg.

The probability I'm assuming is time to number of points. Gomez took about 1.8 games to get a point. Every 1.8 games, he gets 1 point, that's poisson.

Quote:
 Originally Posted by MathMan Trying to find some, but most of the stats sites are either blogs or chart sites, and they don't have a convenient reference or indeed much of an index. And this is such a settled question that there's a cash prize contest for someone who can demonstrate that teams have a talent to influence shot quality: http://www.arcticicehockey.com/2011/...-quality-prize This one is pretty good; it compares first half to the season to second half at the team level, but the same applies to individual players (especially their team's shooting percentage while they are on the ice): http://www.arcticicehockey.com/2009/...229/dont-panic I'll see if I can find more.