# Building a NHL Fantasy Predictor: Do NHL Coaches Have Zone Start Trends?

Ben Wendorf
November 20 2012 07:51AM

Tavares, Sedin of the East
Photo by Michael Miller

Whether coaches have zone start trends is another question I find that I'm asking myself as I build my fantasy prediction model. We know, for instance, that Alain Vigneault uses a drastic zone start deployment, where his 3rd liners would yield virtually no value for fantasy hockey (though they certainly have real-life value) because they start so much of their time in the defensive zone. Why is this important for me? Well, having identified a pretty stable indicator of the amount of on-ice shots a player will take in %AttSh, and knowing that 1st, 2nd, and 3rd lines have typically different amounts of shots-for per 60 minutes, we also need to remember that a coach might affect those shots-for per 60 minutes by having drastically different zone deployments. After the jump, I'm going to look at this question and see what I can find (and potentially use).

## Who gets the plum assignments?

Okay, so we have to keep in mind that the amount of zone starts a player takes in the offensive zone is typically measured as a percentage of offensive zone starts divided by the sum of offensive plus defensive zone starts (OZ Starts / [OZ Starts+DZ Starts]). Stronger possession teams will typically have more offensive zone starts, period, so to say that one team's top line sees 52% of their starts in the offensive zone and another sees 56% doesn't necessarily mean that the two coaches are applying different deployment schemes. What I like to do is to control to the first lines OZ St%, so that the first line is always considered 1, and the other two lines are identified by their values different from 1 (the first line's OZ St%). Thus, the typical 1st/2nd/3rd line deployment, instead of being 52.5%, 50%, and 48.5%, would be 1.000, 0.951, and 0.922. By using these data, I can do cross-team comparisons. Then, I took the cumulative data of coaches with 2+ seasons of coaching the last five years and took a look to see who was trending towards drastic Zone Start deployment. So yes, you know about Alain Vigneault, but did you know about Jack Capuano?

The furthest right column adds up the difference from 1st line over to 3rd line (Line 1 minus Line 2, then minus Line 3). Worth noting is that the differences across all the single-year performances had zero correlation with Fenwick Close, which is considered one of the best indicators of team quality. Intuitively this should make sense, as you see a smattering of good coaches/teams at the top and bottom of the list, right alongside the crappy ones. A few things are becoming clearer with all this information: for one, there's little doubt that team depth can have an effect on this. Even the least stats-inclined coach would give favorable zone starts to a top line considerably better than the remaining lines on his team (**cough** Islanders **cough**), in the same way that you wouldn't put Daniel Carcillo on your powerplay unit. Oh wait, somebody did do that? Nevermind. What's that? Gretzky did that? Well that explains things.

## When coaches change teams

Another thing that becomes clearer, and is somewhat related to the above reason, is that you'd ideally like to have more than, say, three or so years to determine whether a coach is really committing to a deployment scheme. Three years should be sufficient to turn around a team with little depth, unless you're Steve Tambellini or Garth Snow...in which case you'd need about seven years through which you'll cycle through too many coaches to get a good read on zone starts anyways.

So let's try an experiment, just to shake up the analysis a little; let's look at all the coaches who have undergone a coaching switch, and see who seems to stand pat on their deployment.

Like pinning Jello to a wall here. The drastic differences (the triangle denotes "delta", or "change") suggest that, when player personnel shifts drastically as it would in a team change, so do the zone start deployments. Only one coach changed less than the average change among the 1st, 2nd, and 3rd lines (4%); while the shifts of Tortorella and Laviolette average out to low change, you'll note that's because the two larger shifts between 2nd and 3rd lines average out. I think we can pretty safely say that depth and perceived talent (be it offensive or defensive), along with a fair amount of randomness, guide these deployments more than coach preference. That only becomes more apparent when you compare coaches' 2nd line deployment to the next years' (controlled to 1, as above; r = 0.24) and when you do the same for the 3rd line (r = 0.12).

I guess this shouldn't be extraordinarily surprising, since there's still some resistance to the amount of control advanced stats should have over coaches' decisions. With perhaps Vigneault being the exception, there are not many incentives for straying from the averages among the forward player population, where the better offensive players tends to get more minutes, hence top 6 duty, and they tend to get a higher Zone Start %. For these reasons, I'll be sticking with the averages of the 1st line (52.5%), 2nd line (50%), 3rd line (48.5%).

Benjamin Wendorf was co-manager of the SB Nation Winnipeg Jets blog Arctic Ice Hockey (formerly Behind the Net); he is currently co-editor of Hockey-Graphs.com. He has been writing about NHL analytics for six years. He can be reached via email at wendorf DOT benjamin AT gmail DOT com, and tweets from @BenjaminWendorf.