In addition to my previous post on expected Zone Shift yesterday, I just want to point out that I put the zone starts and finishes of forwards and defensemen together to develop the expected Zone Shift, and that to do so requires a bit of justification. So, I split up the forwards and defensemen over this same period just to confirm that we weren't losing something by mixing them.
More graphs...
You'll remember that I divided the performances into buckets of 50 to get a bit more complexity into the data; in this case (thankfully), the forwards-to-defensemen performances are almost a perfect 2:1 split, so I continued to split the forwards into buckets of 50 but split defensemen into buckets of 25. This keeps my buckets in-line with each other for comparison's sake. (click on any of the graphs to enlarge)
The main difference I recognized was that forwards had more Zone Starts at the extremes of the distribution, which actually shifted the ranges of the individual buckets out of line with each other. To make sure that wasn't masking a key divergence in the forward/defensemen data, for the final graph I locked down my x-axis by working with the average of each bucket rather than its range.
So as you can see, nothing to get too excited about, but I figured I'd put it out there...