I have noticed a lot of advanced stats guys put way too much stock into Corsi and Fenwick numbers, sometimes to the exclusion of more direct and conclusive metrics. The newly annointed 'advanced stats' are a useful metric, but they are not the be-all-and-end-all of hockey analysis and in my opinion are like +/- in some ways.
Although in previous use, the +/- stat came to the fore in the 80s as a valuable metric. If you are on the ice for more goals for, than against, it seemed reasonable to deduce you are a more valuable asset than someone for who the reverse is true. It was once considered such an important evaluator that the league determined there should be an award for the best +/- player in the league and 'The Emery Edge Award' est. 1981-82 was born.
GMs rushed to trade for, and sign players(defensemen in particular) based on their positive +/- rating. Agents used a positive +/- rating to gouge more money from stingy owners when negotiating their client's contracts and it all seemed reasonable. Thirty odd years down the road, +/- has been exposed as a somewhat useful stat, but certainly not worthy of an NHL award, which has since been scrapped.
Many advanced stats users seem to be caught up in this same mindset. They smugly toss around Corsi, Fenwick and PDO as sure fire predictors and are so caught up in these stats(as were the principles in the 80s) they dismiss many previously esteemed metrics as of secondary importance or barely credible. I call these people advanced stats snobs.
They think they know more about hockey than they do and they value advanced stats more than they should...in my humble opinion.
To me these stats are only advanced compared to previously documented hockey stats. Compared to stats that I have had to use in database development, which are based on differential calculus and linear algebra they are pretty rudimentary. But I digress.
An example of the blind faith ASSes(Advanced Stats Snobs) put into these popular new metrics is in evidence when they are using them to predict a future collapse of the Calgary Flames.
The Flames have horrible 'possession stats' which Corsi and Fenwick supposedly represent. The truth however is the shots for/against aren't possession stats, they are used as a proxy for possession, which isn't the same thing. The best possession stat is not based on shots at all, it is the seemingly forgotten faceoff winning % stat which is not a proxy, it has a direct correlation to possession.
Since the Flames have lost three centers, their FO% has dipped to around 33% which is horrendous, but understandable since centers take the majority of faceoffs. That means they give up possession 66% of the time after a faceoff. Does this correlate to more shots for the opposition and less for the Flames? Of course it does, how could it not?
Laughably ASSes don't seem to realize this is a big part of why the Flames 'possession stats' are so shabby, and are still convinced they are a valid predictor of the Flames future fall in the standings. They are so oblivious to the obvious they don't recognize when the Flames get their centers back in the lineup, their FO% will trend closer to 50%, which will improve their possession stats. Will this translate to more wins, which is really the only determining factor of how a team performs? Maybe...maybe not.
Still I imagine this will go unnoticed when the Flames Corsi trends upward and the ASSes attribute it to better play rather than an improved FO% which will be staring them right in the face.
Cheers and thanks for reading.
All my comments were lost but someone suggested faceoff % didn't effect Corsi:
My comments are lost check the bottom of my blog for a link to the affect faceoffs have on Corsi
Yeah, I think my comments got deleted. But no worries, the conclusion remains unchanged :)
Faceoff% is a largely a non-factor for TEAM Corsi. That is why zone start adjusted Corsi (subtraction of all shot attempts within X seconds after OZ/DZ faceoffs) are very similar to the non-adjusted numbers, as I pointed out yesterday.
Calgary 5v5 ajdusted: 43.7Calgary 5v5 non-ajdusted: 43.2 (for X=10s)
The results you linked to indicated that 36% of individual Corsi variability was accounted for by faceoffs and zone starts. Have you any idea what those numbers would be for team Corsi?
I ran the regression yesterday: a whopping 0.6%. Adjusting for that would get you nowhere. Calgary would still rank as 29th.
I am NOT saying that Flames are "bad" or anything like that. They have their strengths in other areas than puck possession, namely shooting.
Deleted comments again....still a shitty blog lol.
Plus/minus isnt a good stat because of its limited usage. Where as you are listing several different stats which, when used together paint a pretty clear picture of what can reasonably be predicted for a team. They are legitimate because they are being consistently proven to be true. Teams may defy the odds for a season but as the Colorado avalanche are currently doing a fine job of proving, the regression monster can only be silenced for so long. He is the same monster that is eventually gonna eat the flames.