Agreed with the bolded. IMO and as you mentioned CF% can be very misleading without looking at a wide range of other variables that impact it. And again IMO straight CF% is thrown around way too much as the best measure of a players worth. A few weeks back I was curious about CF% over a long period compared to the almost universally agreed upon value of a player.
An 8 year (2007-15) CF% ranking of NHL players comes with some pretty unexpected results. Screened for a min of 1000 mins to weed out players with less than 1 season.
Crosby the greatest player of our generation comes in at 114. Right between Colaiacovo and Trevor Lewis.
Well behind Johnathon Cheechoo (19), Zhedorov (28) and our good friend Poni (39).
What I found particularly interesting was one of our favorite whipping boys Adam Pardy was ranked 132 (a 314 game sample size). A former whipping boy Derek Meech was ranked 144 (140 game sample size). While 2 of the best d-man in the game Shea Weber ranked 384 and McDonagh ranked 395 were well behind them.
Some other notables including top Jets players:
Toews - 12 (my favorite non Jet from a team I like)
Doughty - 47 (my favorite non Jet from a team I hate)
Perreault - 91
Ladd - 96
Ovi - 98
Buff - 138
Wheeler 194
Giroux - 198
Malkin - 199
Getzlaf - 211
Perry - 223
Little - 309
Stamkos - 430
Iginla - 435
Selanne - 649
Info from Stats.HockeyAnalysis
Well you shouldn't be that surprised. Corsi, like I say below, is not a holistic statistic. It never was one.
Future refinements in Corsi will tweak the above list, but do not be surprised if the best hockey player is not the top since Corsi - and its refinements - is not about who is the best player. Plus the whole no number will ever be perfect thing.
In the context of the current discussion above, it's not saying that "those two players having the worst Corsi = worst players". It's "those two players having the worst Corsi = an issue".
I mean it's not like we don't already know the reasons why those two suffered in shot differentials. I've already shown how they were both poor in zone transitions. Hopefully they will get better as they get more acquired to the system.
This type of work is a great start but is likely just starting to scratch the surface, yet we throw around short term raw corsi numbers like they are a true measure rather than a small piece of data in a very large puzzle.
Yes and no.
Corsi will likely never become irrelevant, but rather be built upon and refined.
You will know better and earlier who is driving. There will be other factors added to this as well. We already have this in part with things like expected goals and WOI's scoring chances.
In terms of a puzzle, we always had that. Corsi was never designed as a holistic statistic and none of hockey statistical community used it as that. It was used much like baseball uses OBP and is a fairly similar statistic in concept. It was that shot drivers were under paid relative to sh% drivers in terms of impact on wins, just like how players getting on base used to be undervalued.
The puzzle picture is already are starting to be filled in with Goals Above Replacement. Shot driving is a huge part of WAR and will also always likely be a pretty large piece.
The reason why I can be confident in saying shot metrics will always remain a fairly large part of the picture is due to the relationships it has and testing with goals. The future is in refining shot metrics (like exG and SC) and also earlier detection with transitional statistics (aka knowing how and why so players can be coached better to improve their numbers).
One day likely to be seen as a small step above giving players pluses and minus if they were on the ice when goals are scored. There was a day when that was seen as a big step forward in measuring a players value.
Goals, as in differentials, have always been important and always will be, just like Corsi.
The only changes have been that now it's numbers people looking at numbers as opposed to hockey people looking at numbers.
GD is still important, we just know that it takes a certain sample to become significant, which is stats 101 not some knew knowledge in technology or hockey.
Corsi didn't make GD obsolete, but rather was a refinement of one shot metric to another.
Plus/minus though never would have been had statistically trained individuals ever were part of the process and never should have been. There is an inherent flaw in its stupid selections.
Ex:
Mark Stuart had the Jets worst ES GD (GF-GA) for dmen.
Mark Stuart had the Jets worst PK GD (GF-GA) for dmen.
Mark Stuart did not have the worst plus/minus ES+PK factors (ES GF - ES GA + PK GF) though because plus/minus arbitrarily ignores PK GA.
Plus/minus has become obsolete not because of new knowledge but just numbers being looked at and tested instead of being assumed as useful because it represents what a coach assumes is important. We still have problems like these today. Coaches always want me to split data into super refined situations where SSS makes things ridiculous.