Tracking the Islanders: Shift-start Project

This season I will be tracking scoring chances in relation to the start of each center’s 5v5 shift for Islanders forwards in CLOSE* action. Over the summer I went back to last season and tracked this information for nine games, in an effort to begin to establish a baseline for each type of shift-start.

* CLOSE is defined as within-one-goal in first and second periods, and tied in the third period.

I categorized each shift-start for a NYI center into the following ten designations, five of which accounted for 332 of the 400 shift-starts:

DZ = defensive-zone faceoff: the center steps on the ice to begin his shift with a defensive-zone faceoff after neither team iced the puck.... 15% of shift-starts.

NZ = neutral-zone faceoff: the center steps on the ice to begin his shift with a faceoff in the neutral zone.... 18% of shift-starts.

AZ = attacking-zone faceoff: the center steps on the ice to begin his shift with a faceoff in the attacking (offensive) zone.... 12% of shift-starts.

Dump = dump-in: the opponent has clear possession of the puck, but does not have attacking-zone possession, when the NYI center enters the play from the bench.... 23% of shift-starts.

Ret = retrieval: NYI has clear possession of the puck, but has not established possession in the attacking zone, when the NYI center enters the play from the bench.... 13% of shift-starts.

DIS = disadvantage: The opponent has clear possession in its attacking zone as the NYI center enters play from the bench.... 8% of shift-starts.

ADV = advantage: NYI has clear possession in its attacking zone as the NYI center enters play from the bench.... 4% of shift-starts.

AZ+ = attacking-zone faceoff after an icing: The NYI center starts his shift while opponent cannot change players. They may or may not be tired.... 2% of shift-starts

DZ- = defensive-zone faceoff after an icing: NYI iced the puck before the entering center could be involved in the play.... Very rare: only 1 of 400 so far.

Other = anything that doesn’t fit into one of the other nine categories: Usually the puck is bouncing around with neither team able to gain clear possession as the NYI center enters from the bench.... 4% of shift-starts

After establishing the type of shift-start, I tracked which team earned the first scoring chance of that particular shift. The shift may extend beyond a whistle, if the NYI center remains on the ice for the ensuing faceoff.

There were 214 shifts that resulted in a scoring chance (108 for, 106 against) of the 400 tracked, whereas 186 resulted in no scoring chance for either team. After only nine games the sample sizes are very small for each type. The first half of this coming season, in particular, we will still be establishing what is expected for each type of shift-start. (And what is expected of NYI centers may be different that what is expected of TBL centers, for instance— due to the strength of the team at 5v5.)

Note that the NYI center is always fresh for a shift-start (and often his wingers as well), while the opposing center may be tired. Thus, we may be able to expect > 50% for NYI centers in this stat, when we combine Dump + Ret starts.

Note that a penalty-drawn counts as a scoring-chance for, and a penalty-taken goes down as a scoring-chance against.

How is a scoring chance defined?

I use the following visual to help define what counts as a scoring chance. The blue area is high-danger zone. Any shot-attempt is counted as a scoring chance in that zone, even if it is blocked right off the shooter’s stick. A shot in the green area typically counts as a scoring chance unless it is blocked by an opponent who is closer to the shooter than he is to his own goalie, when blocking the shot. The shot attempt must look at least somewhat threatening. (If it is such a casual save that a goalie will make 199 of 200 times, it does not go down as a scoring chance, for instance.... The velocity of the shot factors in, as well as whether a goalie needs to push side-to-side to make the save.)

The most difficult shots to evaluate are the ones from distance, particularly with a screen and/or a possible deflection. For those I consider several variables, including the location of the shot, the velocity of the shot, the reaction of the goalie (screen/tip) and even the commentary of announcers, since they may be able to see a deflection live which I cannot on video.... I do not consider who takes the shot from distance (whether Pulock or Hickey, for instance), but since velocity plays a part I imagine more Pulock long shots will go down as a scoring chance than shots by Hickey. (Sorry, Thomas. I would have used Calvin’s weak shot as an example if he was still with NYI.)

What may we learn from this data?

I am particularly interested in seeing how on-the-fly shift starts affect scoring chances. What is the expected scoring-chance % of a shift that begins with a puck-retrieval relative to a shift that begins with the opponent retrieving the puck? It will also be interesting to see how each line performs given “expected” scoring-chance % each game (calculated by what situation they start each shift). Natural Stat Trick has zone-start statistics, but they include self-inflicted defensive-zone starts and earned offensive-zone starts, while they ignore what situation a player starts on-the-fly (which team has possession, and where the puck is on the ice). The NST zone-start stats can be useful for context, but they aren’t very dependable for determining the advantage/disadvantage the skaters had each game.

Centers are used for “the start of each shift,” because tracking three different players at once would drive me crazy, when they enter, particularly since it is often off-camera. The centers are used for organizational purpose, but it is more of an on-ice stat for the entire line, since the wingers play a large part.... Defensemen also play a large part, as well as opponents. I am going to try to note when a NYI defenseman makes an obvious mistake for a scoring-chance against (or a great play for a scoring-chance for), but over time it should theoretically even out to some degree, as is the case with on-ice stats in general.

Most of my available time to track these stats will be Monday through Thursday, so I’ll be playing catch-up most of the season, each week. I will post from time-to-time with “yesterday’s” game, but I won’t often have all of the games in the most recent couple weeks tracked. Single-game stats will be telling some games and misleading other games, similar to how on-ice corsi stats are (but even smaller sample size).

If you have any questions, please ask below.