Here's a few ideas I have to drive analysis. I'm not sure if I'll have a chance to look at the numbers, but I thought I'd throw it on here for others to digest and comment on.
First we'll start with the meta analysis. Two statistically average teams play against each other. On average they each score 50% of the goals that occur (GF% and GA% = 50). Suddenly team A decides "hey maybe if I stood in front of Team Bs shots they'd score less" and team B copies them, but isn't as good. Team A is able to block 30% of team Bs shots, but team B only blocks 10% of team As
Now, logic would tell us that **if all else is equal** Team A should now be scoring a higher percentage of the goals. However the numbers, as provided here by Cunneen, shows us that they don't, therefore either we have faulty data (it doesn't appear we do) or our assumption ("all else is equal") is incorrect. Since I trust the data, lets try and delve into what might not be correct.
I look at it this way. Lets look at two stats I'll call the Corsi Shooting Percentage (I have no idea if this stat has a real name, I didn't see it anywhere) which is simply Goals Scored / Corsi Events for an individual player and the Fenwick Shooting Percentage (ditto on the name) If I have 5 shots on goal, 3 shots of mine blocked and fire another 2 wide, and out of that net 1 goal I have a CSh of 10% and a FSh of 12.5%. Goals scored would simply be CSh * corsi or FSh * fenwick
Now from our setup, both teams are statistically average meaning they have the same number of corsi and fenwick events occurring. In this case
CSh(a) * corsi(a) = CSh(b) * corsi(b) = X
and
FSh(a) * fenwick(a) = FSh(b) * fenwick(b) = X
Where X is goals scored by each team. This also means that for a given team
CSh * corsi = FSh * fenwick
In our initial setup there are no blocked shots so corsi = fenwick and CSh = FSh
But now we look at block percentages. Lets say team A is blocking 30% of shots against and team B is blocking 10% against. Now fenwick(b) = .7 * corsi(b) and fenwick(a) = .9 * corsi(a). But here's where the interesting bit comes in. According to the numbers Cunneen compiled the original equations still hold true.
CSh(a) * corsi(a) = CSh(b) * corsi(b) = X
and
FSh(a) * fenwick(a) = FSh(b) * fenwick(b) = X
Now that we have blocks however our substitutions can become fun by substituting fenwick(a) and fenwick(b) for their relative corsi.
FSh(a) * .9 * corsi(a) = FSh(b) * .7 * corsi(b) = X
or simplified
FSh(a) * corsi(a) = .777 FSh(b) * corsi(b)
In order for that to happen either corsi(b) must be lower than corsi(a) or FSh(b) must be lower than FSh(a), or both.
In plain terms it seems to me that it means either.
1.) Blocked shot percentage has an inverse correlation with corsi against. The higher percentage of shots I block, the more get thrown towards me.
or
2.) Blocked shot percentage has an inverse correlation with FSh against. The higher percentage of shots I block, the higher percentage of goals an opponent scores on a given number of fenwick events
Or
3.) Both
From an anecdotal standpoint I can defend these. If I'm blocking shots I'm also in the goalie's way, so he's less likely to see the puck and more likely to let it in (so says common sense). Similarly, if I'm blocking shots, I'm more often than not giving the pick back to the opponent so he can simply take another one (again, according to common sense). However, I have no statistical proof of either of these.
Does that make sense to anyone? Did I screw up my analysis somewhere?