Quoting: TheEarthmaster
Ah, you're looking at the Standard charts. I'm looking at the Gar/xGar charts that is on the tab next to it, that's why they look different.
I tend to like GAR/WAR because it distills everything down to one number- Goals, or Wins - and it takes into account stuff like usage, level of competition, with/without. They have all the math explained on hockey-graphs, kinda dull but I've gone through it a few times. Kind of a moneyball approach, though that's not a perfect comparison. There's a legitimate criticism that that distillation loses things, but I think, based on who their model says is good, it's a good evaluation of value and effectiveness.
Quoting: BeastModeUnknown
GAR metrics are the best tho.
ok this is responding to two people so please be patient with the "I didn't say that, that was [the other person I QTd here]"
Anyway, yeah I get that but I come from a baseball background looking at analytics (and form what I read on hockeygraphs they look to Tango et all) and transferring lessons from there it's best IMO to use an all encompassing stat as a shorthand most of the time, but if you remotely have the time at all look at the breakdown of that stat and put that ahead, like way way ahead, of the all encompassing stat, unless you are getting to like career levels of big enough sample.
For example looking at baseball and WAR lets say for hitters, a guy has a 7 WAR season, maybe because his UZR, which takes about 3600 innings or nearly 3 years to become a useful sample, is crazy high. Is that real? what if we adjust his defense for a 3 year rolling average by adding the UZR for the year before and year after (or 2 years before if you have no future data yet) and divide it by 3 then look at the WAR? but nobody does that& the player did in fact create those outs even if they are not indicative of that players talent going forward. What about a player whose hitting, their wRC+, jumped up? Was it a flukey BABIP? are they a high BABIP guy? was it more ISO? was that ISO more BABIP because stuff was landing in gaps instead of gloves, was it an elevated HR/FB%? Will that go down? or maybe it was an elevated HR/FB% but their LD% spiked up because they are hitting the snot out of the ball? Maybe their LD% was down but their HR/FB% are both up because they are hitting the ball hard and LDs are being misclassified as FB? Maybe it was just a few flukey FBs down the line? If you just look at WAR you don't see any of these things
additionally in baseball there is a lot of easy filtering of results from individual battles, hockey there is a lot of moving parts so we need to loot at globs of data and adjust for our controls, but even our controls (like a player changing teams ) have a lot of unquantifiable variables (worse linemates, coaching schemes that are good but don't fit the player, coaching schemes that are bad, coaching schemes that are good and make a player look better than before etc.). We've drawn some good conclusions (primary points are indicative of scoring prowess but secondary assists are mostly noise, 5v5 better shows a players prowess unless we are looking at goalies in which case we should be looking at all shots faces with 5 defenders in front - i.e. include PK, etc.) but we don't know say how to quantify primary points vs possession for total value
Also some attempts at adjusting this stuff away (level of competition,,etc) can actually add back in SSS noise instead of alleviate it. I don't see any specific weight calculations in GAR, I see that they encpass loosely defined things like EVO & EVD which sound right but I have no idea how much pull say primary assists or shot location vs shot volume factor in - or more importantly why what weight was used, maybe I haven't found the magic link with the formula. I love how available the individual stats (like xGF/60 for example) are and how well they explain stuff in general but I haven't seen taht last link adn I wonder if I will agree with it.
This is probably why I think I value Sanford far more and Blais far less looking over Corsi% (which tbf should probably be Fenwick because how valuable are blocked shots to telling us possession but then again I'm weighing that in tandem with...)xG%, primary points, and realize that guys sustaining it over 80 games is going to be a bit more valuable than a guy who has done it for 40 that we've extrapolated. IN that sense, sure Blais's possession is fine, but Sanford's is a bit better 5v5 from what I see, he's sustained it for longer, and he has more scoring to go with it. That's also why I think Barbie with noticeably worse possession stats than Blais last year but better possession stats over his career and last year was a bit of a down year, who also scores more and plays center, is more valuable.
All in all we are asking which cheap 4th line/throw to the left of O'Reilly guy is better or more worth keeping in the 1.5-3M range 2 years from now and my answer like both you, start with "hopefully all of them" but then moves on to a radically different Sanford (who I think of as more of a middle 6 guy) > Barbashev > Blais and in this exercise of trying to realistically manage a roster & keep Pietrangelo (who moves the needle in a way I think we all agree on) and with Seattle taking Perron (because IMO that's the realistic take as ideal as them taking Faulk would be), I ran up against the cap, over by <600k and had to get rid of 1 of them to stay under the min.
Just team fit alone I think having another 4C behind Sundqvist in the event he goes down, let alone if he moves up so Thomas or Schenn can play wing, as well as having the better player and fit playing to teh left of O'Reilly if Thomas doesn't move up or Schwartz does walk or again someone gets hurt, all means more than Blais, and he's 3rd on my list so he went. that's all.