Not team shots, just shots that are taken when the individual player is on the ice. Then the possession statistic is adjusted for score effects (a team behind on the scoreboard tends to take more but lower quality shots), team, etc (that's why you see us using relative statistics instead of just raw shot attempt numbers).
Yes, but shots that the team takes while the individual is on the ice. It's all the shots taken by all five Leafs on the ice, not just Leivo, who notoriously couldn't hit the net. Measured against all the shots the opposing team takes while he's on the ice. The adjustments you mention are good, to the extent they are reliable. I don't know the math behind assuming shots are lower quality just because a team is behind - that would seem to depend on the team, and isn't just an across the board fact. Actually, when a team is behind by a goal and they abandon the defensive game somewhat in the last five minutes or so, the shots tend to be pretty high quality and dangerous because everyone is on the attack. But let's say they've got the math mastered and the adjustments work, it still doesn't move me that the stat rolls up the contributions of all five guys on the shift and assigns a number. The dmen could have a monster shift and let nothing through, while taking it to the other team offensively. Leivo gets the credit, but doesn't do any of the work. No different than if he's on the ice with Matthews and Marner and they pop a goal while Leivo never even gets near the puck, but they all get a +1 just the same.
If a player is consistently better than average in relative possession statistics (which again, is measured against their team mates) then they're very likely doing things on the ice to promote the puck going towards the right net. It doesn't mean that they're doing it themselves, just that they're doing things better than their team mates are to drive that play.
But that's the thing. Presumably all five guys on the ice are earning themselves the same number for that shift, right (ignoring seconds gained or lost for getting on and off at different times)? If so, it doesn't differentiate between what they are each individually actually doing on the ice. It's musketeer style - all for one and one for all. When you say it's measured against their teammates, it's really measured against just the teammates not playing with them at the same time. So, if the Matthews line is going up against the other team's best players usually - say Crosby - maybe Matty is winning that advanced stats battle or not, and Leivo is going up the opposing team's worst players, what's the value of even comparing the two against each other. How is Matty's number relevant to Leivo's number? It's apples and oranges, isn't it?
Nah, it's a lot different. Goals are a really infrequent events (so not a lot of available data, if there are only a couple data points per game to work with) and +/- has no adjustments done. When you see a player go from a good team to bad, or vice versa, their +/- swings wildly. Their possession numbers though, don't.
I'll take your word for it on the possession numbers. Why would they not change though? They're playing with four different guys. So if Tavares goes from playing with lesser players on NYI to playing with Marner and Rielly having career seasons in Toronto, his numbers don't see a jump? Now the Leafs are maybe a bad example because I think we give up a lot of shots, but still, where's the logic? It doesn't matter who the other four guys on the ice are, even if they're way worse, your shift is still resulting in more shots for than against, and you're the one responsible for it?
Well first is that the stats track shot attempts, not actual shots on net. So the posts count as shot attempts, blocked shots count as shot attempts. So in your hypothetical, those players on the ice in scenario A would be rewarded in the possession stats for their shot attempts.
Okay, thanks for clarifying. To the extent I didn't account for this above, my bad, but I don't have the time to go back. So say in the same scenario then, modified a bit, Leivo gives the puck away at the beginning of the shift, the opposing team is cycling the **** out of the Leivo line and is in our zone for 45 seconds but doesn't get a shot off, and then Leivo takes the puck and fires it down the ice directly on the goalie. Leivo line's possession stats went up, opposing line's possession stats went down, even though the opposing line actually "possessed" the puck the entire time?
Even that though is kind of the wrong way to think about it, because one isolated event kind of doesn't matter. What matters is the collection of a pile of data. For example, if 1000 of your scenario A's happened, and 1000 of your scenario B's happened, I'm comfortable saying that scenario A would generate for more shot attempts than scenario B. The players who found themselves in scenario A more often, would be superior possession players to scenario B. To further clarify the matter though, statistics like expected goals actually tracks the shot location and gives the shot a % chance of going in from there (based on real shot data, how many pucks have gone in from that location versus shots taken from there) and folds that into it's number. So if scenario A's players were taking shot attempts from areas with a high chance of going in, they would be rewarded in the xGF% statistics for those shot attempts from dangerous areas, and that 150 foot clearing attempt just happening to land on the opponents net, as it has a very nearly 0% chance of going in the net, isn't rewarded to nearly, nearly the same degree.
Okay, more comforting. And I agree, cycle the puck for a full shift in the offensive zone, and more often than not, you'll have your shot attempts. But it still doesn't make me feel any better about my biggest gripe about the stat - that it doesn't differentiate between the individuals and what they're doing on the shift. If Leivo, Ennis, and Gauthier were all on at the same time, and say they all played the exact amount of time (on when the puck dropped, and then the play stops later on a penalty call), no matter what happens, they all earn the same possession number for that shift, correct? Leivo can give the puck away five times, while Ennis rushes the puck up off the ice and takes two or three shot attempts. But they're treated as though they contributed the exact same to the shift.
Well that's the thing, if Leivo was turning the puck over a lot, he would be spending a lot of his time in the defensive zone and would see a lot of shot attempts against. But we see, over his Leaf career, with a number of different team and line mates, Leivo has generated a higher percentage of shot attempts going the right way than his team mates have. Like I've repeated a bunch of times, if Leivo was playing the way you've claimed, he simply wouldn't have the metrics he has. They're specifically tuned to look for the events in hockey games that lead to shot attempts for and against.
How does this work though? So he had some time with Kadri, I think? And he had some stints with Lindholm and Gauthier. So it doesn't matter who the other four guys are or what the other four guys are doing on the ice, somehow Leivo being on the ice results in a net positive on team shot attempts? How does that make sense? And again, how do we account for the fact that he's not playing against the other team's best players? Why compare him to Matthews or Tavares, for instance, when they are basically playing a different game? And if we're comparing him to his own linemates, why would his number be any different than theirs? But say the third and fourth lines get juggled. It was Leivo-Kadri-Brown and Lindholm-Gauthier-Ennis, and then Leivo and Lindholm get flipped. So now what happened, you're saying, is that Lindholm's number would go down while Leivo's would stay up, even though Lindholm would be playing with better players? What about the dmen that aren't glued to the forward lines and change? Isn't the possession number like plus/minus too in the sense that it doesn't track who you're playing with, just what your individual number is? What I'm getting at is that it seems like there are so many variables that can't possibly be accounted for. Who you're playing with, who you're playing against, what the players on the ice are actually doing, random circumstances (player from one team breaks his stick and is playing without it for half the shift), dmen playing with you (if Rielly is playing behind Leivo and Ozy is playing behind Lindholm, isn't that significant?), etc. And most importantly, it doesn't qualitatively account for individual play. So when I'm seeing Leivo playing like crap, why can't it be the case that Rielly and Hainsey are on the back end and preventing shot attempts against, and that Ennis is dangling while Leivo is watching? And that Kadri is getting the shutdown matchup against the other's team's top line and is maybe having his ass handed to him by McDavid or whomever? Why do I trust that number assigned to Leivo when all these other things can be happening?
One final little bit on some of the old school "intuitive" statistics that are extremely misleading. My favourite two are blocked shots and turnovers. One has been praised and the other vilified over the years but the old consensus on them is largely wrong. A blocked shot is a bad thing. Now, it's not the act of blocking the shot that's bad, it's that the other team had possession of the puck in your end and attempted a shot on net. At some point, someone made a bad play with the puck and gave up possession of it. For example, it can be as simple as a dump in (again, a play that is almost universally praised) that leads to a change of possession.
A turnover though, specifically an offensive zone turnover, aren't nearly as bad as they seem. If a turnover is made while trying to attempt to make a play (like a pass into a dangerous shooting area for example), that was a play with a chance of turning into a goal. That's a play worth trying over and over again because a significant amount of the time, you're going to generate a good shot off of it, and those lead to goals. The other part of that is the act of holding the puck in the NHL is a skill, full stop. The single hardest thing to do in that league is to hold possession of the puck. If you're good at that supremely difficult skill, you're going to turn the puck over a lot because you're going to have the puck a lot. If you're shit at holding/carrying the puck, you never have it to actually get a "turnover"...but of course the league doesn't consider a player willingly giving up possession (a dump in) as a turnover, only when a player is trying to make a play with it.
Not sure the direct relevance, but interesting nonetheless. What I would say to this though is how isn't it even a misnomer to call this "possession"? It would seem you could have the puck for 7 seconds and get a quick flurry of three shots on the goalie, and then the other team could be actually passing the puck back and forth and making nice plays that ultimately don't result in a shot attempt. You tend to see it happen frequently on PPs. So the team with the PP is moving the puck around in the offensive zone for 45 seconds, waiting to set up a perfect play that never comes, with no shot attempts at all, but literally "possessing" the puck the entire time, and then the dmen turn it over at the point, leading to a breakaway for say Brown who gets two quick shots on the goalie at the other end. All four Leafs just "outpossessed" the team with the PP? (I assume they count the possession stats just the same during PP and SH time).