[Editor's Note:] This is a great stat-centric look at which players do the best at "flipping the ice," using good possession skills to move the puck from a defensive start position to an offensive stop position -- basically, taking a defensive faceoff and having the puck stopped in the offensive zone. This discussion came from a comment about Trevor Daley, but I think it also highlights why the Stars may have felt better about the return of a certain player than most fans.
Also, this is a perfect example of great reader-generated content. We are always willing to promote stuff like this to the front page. The title has been edited for clarity. [End Note]
Recently, I compared Trevor Daley to a very good punt returner by comparing his O-Zone starts to his O-Zone finishes. While it helped make my point, the observation was badly flawed. I caught this flaw while examining elite offensive players. Players with high O-Zone starts nearly always had negative Zone differentials. Players with low O-Zone starts had positive differentials. Something’s amiss. I decided to take a peek at the data. My initial hypothesis was that the greater the distance for 50% O-zone the more inaccurate O-Zone Start/Finish differentials were. I pulled O-Zone starts and finishes from behindthenet.ca for players with 50 or more games in the 2011-2012 season. While doing the basic data preparation I realized my hypothesis didn’t go far enough.
My Initial run at the data showed mostly expected numbers, except standard deviation. The standard deviation between O-Zone starts and O-Zone finishes was too large for direct comparisons. (Kids new to statistics should goggle standard deviation at this point). The standard deviation would produce very different graphs. We don’t like different shaped graphs for comparisons. Also notable is the count of outliers.
NOTE: A sample is subset of a population. We use samples for various reasons. Normally, we work with samples in hockey analytics. I am working these data as a population.
|Off Zone Start %||Off Zone Finish %|
|population variance||49.186||population variance||8.315|
|population standard deviation||7.013||population standard deviation||2.883|
|low extremes||3||low extremes||0|
|low outliers||15||low outliers||5|
|high outliers||4||high outliers||3|
|high extremes||2||high extremes||0|
NOTE: More than 50% of plays appear to finish in the offensive zone. That percentage isn’t a statistical flaw. If you can’t drive possession, you won’t make 50 NHL games.
We need to normalize these data. Because we’re using a population the standard score, or z-score, is a well accepted choice. The z-score quantifies standard deviations from mean.
This is the tricky bit. I almost got a headache from over-thinking it. The farther a negative number is from zero (smaller) the more frequently a player generated a favorable zone shift. The farther a positive number from zero (larger) the higher the frequency of undesirable zone shifts. I zone shift must exceed one standard deviation from mean to achieve statistical significance. Thus, statistically, a zone shift of .01 is equal to .09.
For my final sort I rejected all outliers. Something else drives those numbers (Alain Vingeault).
The best zone shift guy by z-score wasn’t Trevor Daley. It was Tom Wandell. That's in the league.
|NAME||Off Zone Start %||Off Zone Finish %||Start Z-score||Finish Z-score||ZZ-score Shift|
Where was Trevor? Trevor, or Jamie Benn, isn’t significantly above average.
Stars new and old…
It might be time for the AHL.
Phillip Larsen and Adam Burish were in the -1 group (minus is good). I didn’t see a single Star one or more zone diff deviations below mean.
Multiplying the z-score diff by negative was tempting. I left it so others could over think the data and find any flaws in the method. Of great importance is that these data are not ceteris paribus. Jarret Stoll’s -2 zone diff is much more impressive than Tom Wandell’s. Patrice Bergeron in -1 group is more impressive than Tom Wandell. Quality of competition and quality of teammates affects the real value of these numbers. Normalizing these data opens room for more normalized comparisons. If I do another, my other favorite differential will be taken to task, +/- on/off ice.
Trevor Daley isn’t analogous to a great punt returner. If anything he is a solid possession receiver or very good nickel-back. You still can’t trade him, because you can’t get fair return.