Quoting: Xqb15a
It’s just to painful to point out what statistical modeling is and truncated mean, and that’s why I tried simplifying it for you by eliminating the top result instead of the top and bottom so you get a truer mean of production, but hey I’m the one that needs to shake my head. Literally you stated you make assumptions that are not factually based, MacKinnon is not (yet) a 100 pt player you don’t think for a minute Marchand doesn’t let him know that every time they see each other. But yeah you go with just a points that I have used, ignore driving play, or his limited xGD success. I’m sorry my man, I should have never started a discussion with somebody who isn’t interested in facts but is blinded by a bias of the team he roots for. Good luck with that.
Only one of us has presented facts that a reasonable person could see and say yeah, that makes sense. Your justification for Lindholm being a 50 point player is not one of them lol, any casual hockey fan would be able to find something wrong with that, and if you don’t think Mack is a 100 point player then great, we’ll get back to that Gaudreau for Mack trade, they have the same career high in points after all. Disproving the points made by the other party and defending your own is how a debate works, but you have yet to disprove any of my points or successfully defend yours so y’know, any sort of evidence would be appreciated. The problem with your “statistical model” is you’re including meaningless data points. A data point from 8 years ago may be useful for something like weather forecasting. A data point from 8 years ago has literally no affect on the offensive production of an NHL player, and a 3 year sample size provides a significantly more accurate model. You can truncate whatever you want, but it doesn’t change the fact that the size of your model is too large. For example, if you wanted to predict average life expectancy for a human, would you go back 500 years to collect data? Probably not, because you’ll get a flawed, lower number, as nowadays thanks to modern medicine and other factors, life expectancy is significantly longer, but using a flawed, large dataset with now meaningless data points will give you an inaccurate number (and as the average life expectancy was below 30 until the late 1800s, and is now over 70, clearly those data points are meaningless but would significantly affect the average, even if you cut out the singular highest and lowest data points). Same kind of deal for the NHL, as a player matures and develops and works their way up the lineup to play with more skilled players, their old data becomes meaningless. I really don’t care about Lindholm’s production on Carolina’s 3rd line 8 years ago when he’s on Calgary’s 1st line and PP1 now. If you ever took a statistics class I’d advise going back and giving your old textbook another read, it doesn’t seem like you retained much. Defining your dataset accurately is a pretty early step to mess up on. And as for his possession numbers, I’m not even sure where you pulled stuff like 56 ozone starts from (looking at the wrong year?), but following that up with a statement like “he’s bad” doesn’t help. He had a 52.85 CF%, a 56.72 GF%, a 52.11 xG%, with a 49.63% ozone start rate at 5v5 this year. So he has a positive expected goal difference, and his actual goal difference was higher, so I’m dying to know how that’s a bad thing, especially when he started in his own end more often than the offensive zone. And making it all strengths for this year lowers the ozone start % and raises the other numbers lol.