ZONE START PERCENTAGE
What it is: Like Corsi, this is an easy stat to understand. It measures the usage, or the situations when a coach chooses to play a particular player. The only time a coach can give players a distinct advantage is after a whistle. In other words, when there's a draw in a particular zone, the coach chooses the best guys for the situation. OZstart% looks at how often the coach chooses to give a player a shift that will start in the offensive end of the ice. The implication is that guys who have defensive roles will see lower OZstart% and guys who emphasize offense will have a higher %.
Limits: The two major limitations are fatigue and circumstance. For example, if the Ducks third line this year, arguably their best defensive line, has a long shift that ends in the defensive zone, Boudreau might have to send out a line he wouldn't normally want to take that draw, and that might affect this stat. Another possibility could be the end of a special teams situation where the PP earns an offensive zone start after time on the penalty expires and then a defensive line has to take the next shift because the offensive players are tired from the power play.
Another issue is that this stat favors defensive players. Offensive players who are strong defensively might start in the offensive zone a larger percentage of the time, simply because the team doesn't have many other offensive options.
Advantages: Any statistician will tell you that with a large enough sample size the trend becomes clear. While the situations I listed in the limits are possible, I'm sure they aren't as common as the coach's decision. It gives us a clear idea of the situations in which a player is most often playing.
What it means: While the decision does lie with the coach, it's still difficult to say that offensive zone start percentage is a clear indication of whether or not a player is fulfilling a defensive role. This is a mistake I have sometimes made, but I'm trying to see it more as the type of minutes a player has to play. Starting in your defensive zone is tougher than starting in your offensive zone, especially in terms of corsi. I think the best service that is performed by OZone% is its ability to help us better interpret other numbers.
Players' corsi can be affected by where they start on the ice. How significantly it affects corsi is probably subject to question though. I think that corsi's sample size wouldn't allow for a major shift, unless there is a drastic shift in OZone%. Similarly, I think you can only use OZone% as an interpretation of a defensive role, when the numbers are largely skewed towards defensive starts.