Anaheim DucksArizona CoyotesBoston BruinsBuffalo SabresCalgary FlamesCarolina HurricanesChicago BlackhawksColorado AvalancheColumbus Blue JacketsDallas StarsDetroit Red WingsEdmonton OilersFlorida PanthersLos Angeles KingsMinnesota WildMontréal CanadiensNashville PredatorsNew Jersey DevilsNew York IslandersNew York RangersOttawa SenatorsPhiladelphia FlyersPittsburgh PenguinsSt Louis BluesSan Jose SharksSeattle KrakenTampa Bay LightningToronto Maple LeafsVancouver CanucksVegas Golden KnightsWashington CapitalsWinnipeg Jets

SCORING CHANCES SPREADSHEET & APPLICATIONS

In the first of my three-part series on scoring chances, I argued that the time has come for the ‘advanced hockey stats’ community to track more scoring chance related events.  Last week a grading system was established and presented because scoring chances vary greatly in terms of their goal-scoring expectancy and this needs to be recognized within the tracking system.

Scoring Chance Tracker (Ana-Pho20131123)

Now we get down to the nitty-gritty.  I present my scoring chances tracker spreadsheets, complete with game summaries for team and individual statistics.  I finally chose the Anaheim at Phoenix match-up from last Saturday (Nov. 23rd) as the subject-game for this presentation.  This was a big match-up of Western Conference powerhouses and sure to be filled with playoff-type intensity and speed...  And it lived up to expectations.

Before we get to game analysis using the downloadable spreadsheets I’ve provided, let’s set up the “Scoring Chances Tracker” legend and explain its contents.

For every scoring chance tracked, we will start by recording the team credited as well as the period and time of occurrence.  Then as the NHL does for goals, the man strength situation (Str) is recorded along with all the on-ice players’ numbers for both teams.  The legend for the rest of the category columns can be found on the file’s last sheet.

NB: Let me start by clearly re-stating that this whole exercise is simply an individual effort to contribute a new tool that helps record scoring chance events as objectively as possible; a tool that would also provide more contextual detail than is currently available (i.e. shot attempts and shot distances).  I am very much aware of the constraints my methodology presents: it took me several hours of work to prepare the spreadsheets for just this one game.  Gathering enough data using this tracking method to make it a worthwhile statistical endeavor would be a painstaking task and I certainly don’t qualify as a good candidate to carry it out at this moment.

ANAHEIM AT PHOENIX GAME ANALYSIS

The first thing I noticed in this particular game is the obvious discrepancy between what the traditional shot attempts stats tell us and the scoring chances final tally.  Here is what Corsi and Fenwick proponents would use to analyse the general outlook of this game:

Anaheim total shot attempts: 51 (incl. 10 blocked), Phoenix: 83 (incl. 24 blocked).

But the Ducks won the scoring chances battle 18-16.  This was in reality a very free-flowing, but tightly contested game featuring great team speed on both sides and allot of offensive involvement from the back end.  There were evident shifts in puck possession momentum and the first sheet of the file (“Chance Tracker”) illustrates that very well.  Notice how the Ducks came off an early penalty kill to completely take over the second half of the first period (even though shots were 15-8 PHO in the first).  The Coyotes had their customary push-back (they’ve been having difficulties at the start of games all season) after Anaheim scored their fourth goal late in the second.  The momentum carried on to much of the third period as Anaheim seemed content to play prevent defense, but the three-goal deficit was simply too much to overcome.

The other sheets of the file summarize the scoring chance related stats from the game.  The team stats are shown on the second sheet, detailing chances by period, by grade and by man strength situation.  Goals are also shown by chance grade (NC denoting goals scored on non-scoring chance situations, none in this particular game).

We can easily decipher the dominant skaters in the game with a quick glance at the individual numbers (third sheet): Radim Vrbata, Keith Yandle, Shane Doan and Mike Ribeiro lead the way for Phoenix while Dustin Penner was a clear standout for Anaheim.  An interesting number to focus on is the SC+/-, which works the same way the traditional +/- stat does, but for EV chances instead of goals.  Since there are more scoring chance events than goals, this provides us with a much clearer picture of a player’s overall efficiency, especially over multiple games.  Offensively, Penner and Vrbata lead the way with 4 SC each, although Penner was able to finish twice on Grade “B” chances, whereas Vrbata ended up with only 1 assist despite authoring one of the 7 Grade “A” chances in the game.  Ribeiro was the top playmaker with 3 SCA, even though he finished the game with a goal and no assists.

The “Summary (Indiv.)” sheet also goes on to detail each player’s participation in Grade A and B chances by man strength situation (Grade “C” chances are excluded here as they present a much lower goal scoring expectancy).  It also shows all the chances a player was on the ice for and against, and for each man strength situation and chance grade.

The final sheet shows goalie stats, with saves and shots faced on scoring chances.  It also distributes the numbers according to chance grade and man strength situation.  Of course, the value of goalie’s scoring chance related stats would be better served with a much bigger sample size than what we get in one game.

APPLICATIONS

Obviously it is impossible to cover all the possible applications stemming from an eventual materialization of this project, because they are seemingly endless.  But I thought I would at least discuss some of the most intriguing ones in my view.  Who knows, the potential possibilities might hit a nerve with the right kind of people out there, the kind that can give this project outline a real lift-off.

Have you ever wondered which NHL player actually creates the most offense (not just goals and assists)?  Divide (SC+SCA) by TOI and multiply by 20 to get SCC/20 (scoring chances created by 20 minutes of ice-time).  Last season while talking about the Caps’ new acquisition Mike Ribeiro, head coach and former playmaking genius Adam Oates coined a phrase that resonated with me.  He said Ribeiro was very good at creating the “scoring pass”.  That got me thinking: how can we define a scoring pass?  My suggestion: a Grade “A” SCA (highlighted in light green on the spreadsheet).  The best playmakers should be at the top of the ‘A’SCA/20 list every year.  Why have rebounds never been tracked by the NHL?  The NBA recognizes their importance.  Ever wondered which goalie is the best at rebound control?  You can now consult the Freeze% (non-rebound producing saves/SC saves) or DReb% (percentage of total rebounds recovered by the defensive team) leaderboards. 

Want to push it a little further?  Weigh all the stats according to scoring chance grade.  And this is the crown jewel of this charting system: we instinctively know that the chances classified as Grade “A” have a much better chance of ending up in the net than the others, but exactly how much better of a chance?  Charting a few complete seasons of games would give us that answer. 

Matrix experts could then cross-reference as many of the scoring chance events categories as they want to provide hockey fans with new insights on the game. For example, which passing play has a better goal scoring expectancy: a clean one-timer on a hard 10-to-7 zone pass or an own chance shot from zone 4?  The original grading chart I suggested could even see some shifting of chance types once enough empirical evidence becomes available.  What Sabermetrics did for baseball (Run Expectancy for each possible situation), this kind of study could eventually do for hockey.

VERY EARLY RETURNS

I did perform a partial charting (team chances only) of last spring’s Stanley Cup Finals (all six games) using my scoring chance grading chart and here are a few numbers and interesting observations it spawned:

-          Total chances (A-B-C): BOS – 117 (23-76-18), CHI – 108 (30-65-13)

-          Total chances/G (excluding OT): 31,0 (BOS: 15,0 – CHI: 16,0)

-          Total “A” chances/G (excluding OT): 7,7 (BOS: 3,2 – CHI: 4,5)

-          Goal distribution by chance grade (32 total goals, excluding one empty-netter): A-21 (65,6%),

B- 8 (25,0%), C- 0, NC- 3 (9,4%)

I have tracked many more games in the past and can confirm that most NHL games range from 25 to 35 total scoring chances.  We often hear NHL coaches say that limiting the opposition to less than 15 chances is a good indicator of sound defensive play. 

There were more that 2,5 times more goals scored on “A” chances than “B” chances during the six games despite premier chances representing only 23,6% of all series chances.  I would argue that the Hawks’ ability to create more “A” chances than the Bruins gave them the edge in an otherwise evenly contested Finals.  Not to toot my own horn, but I did mention Chicago’s edge in elite skills being key on this site last spring as I successfully called the six-game Hawks’ victory.

I concede that six games is a very limited sampling size, but this leads me to believe I’m on the right track with my grading chart and it emphasizes my point that we need to not only track scoring chances, but grade and contextualize them with empirical evidence.  Surely the NHL’s brass can see the value in putting a few more statisticians to work on this.  Anybody out there has Gary Bettman’s ear?