Quote:
Originally Posted by Calgary4LIfe
Well, I don't mind someone trying to figure something out using stats at all. I can't say I agree with the results, but kudos on starting a new thread where stats can be discussed and dissected, and people can gain some insight into this side of things rather than get shouted-down at times (and yes, I am fully aware I have been one of the people shouting down - PDO IS a joke, and no explanation will ever convince me otherwise).
My opinion of the advanced stats we currently are exposed to - most of it is junk based on loose correlations that are indirectly measuring something.
How about this for a novel idea:
Devise stats that DIRECTLY measure events and name them something that tries to explain what you are measuring. Crazy, right?
I will believe in the existing advanced stats when teams OTHER than Edmonton employ people to work with them. I think Burke's analogies of the existing stats is apt - something something drunk guy lampost something... I believe Treliving stated matter-of-factly that the Flames use their own in-house developed stats and do NOT use CORSI and the other stats out there.
So, while I don't agree exactly with FingerCooking's work here, at least he is trying to apply different things to see if there is correlation or if it is a predictor of any sort.
Columbus to me has been a team on the verge of doing a lot of great things, or falling on their faces yet again. One year they were on the verge of really making noise until injuries decimated them, and then they sucked again. This year I think they are playing to their potential. From watching them play, they are fairly dominant on the ice, and are backed by one of the league's best goalies for when they make a mistake. From that standpoint, I would infer that they should be near the top of the list, not the bottom.
However, at least someone is trying to think differently than trying to say how many shot attempts were directed at the net - regardless of placement and circumstance - and say this is possession. Call me crazy, but a hail-mary shot from the redline is NOT as important as 30 seconds of complete domination in the offensive zone, even though a few high-chance cross-crease passes just didn't connect as they just took a bad bounce on a patch of poor ice.
How about we time possession - just like they do in soccer - instead of trying to indirectly measure it? Makes more sense, no?
|
Well the problem with stats is that it doesn't account for everything. My personal favorite is the player usage chart, even though it only accounts for shot differential. I would like to see a more broad metric based on quality-of-competition and zone starts, therefore for players are subdivided into the following categories:
1st liner
2nd liner
3rd liner
4th liner
1st pairing
2nd pairing
3rd pairing
Once a player is in a given category, there should be a differential of their performance based on positive and negative events. These events would be:
Positive/Forwards
Successful shot block (D)
Body check (D)
Winning board battle (O)
Causing board battle turnover (D)
Successful pass (O)
Successful shot on net (O)
Contributed to faceoff win (O)
Did not take a penalty
Stick checked puck (D)
The negatives would be the opposite of these categories (ie, failed shot block/didn't try). The differential would yeild a metric that indicates their positive contribution to victory, assuming coaching system endorses these events. Then you would have a metric (ie, +1.5) that shows a player is a contributor and not a burden. Since that metric could be placed on a usage chart, you can then see if they're "A contributing/burden 1st liner with mostly offensive zone starts" or whatever. Then analystics could compare apples with apples.
This would be different that shot differentials because not every player is a shooter. Some guys just block shots and body check and cause turn overs (ie, Stajan). Assumingly a guy like Stajan would yeild a "4th line contributor with mostly defensive starts while facing an average of 3rd line competition", whereas a guy like Monahan would yeild a "1st line burden with all zone starts facing an average of 2nd line competition".
While it's not perfect, because theoretically you could have a coach who wants a player to stretch pass (ie, yield more turnovers) therefore the negative events are biased against them, it at least encompasses a player usage chart with more events than shot differentials