Monday, January 12, 2015

Advanced Metrics Explained, or, Fancystats 101

Lately throughout the NHL people have been using “advanced statistics”, or as I like to call them, fancystats, to better quantify a player’s quality, or a team’s quality.  Some “old school” blowhards like to disregard them because they feel people who swear by them “don’t watch the game” (seriously you’ll see a variant of that phrase in countless message boards and said countless times by former players and the like).  I personally feel that they’re usually a pretty good predictor of a team’s chances to win, and they only enhance my viewing of a game. Unfortunately, some of the phrases can be pretty daunting, so here I made a glossary of various advanced statistics that you might see come up in a discussion, with simpler explanations than I usually see for them. Enjoy.

Editor’s note: some of these are taken from an email conversation I had with friends, which inspired this article, I edited the explanations so they don’t reference the article they shared as much, but it’s worth noting.  Also this won’t include basic stats, such as goals for/against, shots, and time on ice, as those are very simple and usually found on the NHL’s website.  It’s for brevity purposes. Another editor’s note: all numbers are hypothetical, unless explicitly stated otherwise, usually to make calculations easy.

Corsi %: The percent of shot attempts (including shots on/in goal, blocked shots, and missed shots) that the player's team takes as a percent of all shots taken while he's on the ice. For instance, if the Blackhawks attempt 50 shots while Patrick Kane is on the ice, and Detroit attempts 10, because they suck, Kane's Corsi % would be 50/60 (the 60 being the 50 from Chicago and the 10 from Detroit), making it an 83.3%.  That's unrealistically high, but it should explain it well enough.  Sometimes this can be called the “Corsi For %”, but they mean the same thing, shot attempts divided by total shots while a player (or team) is on the ice.

Fenwick %: Fenwick is a similar metric to Corsi, in that it measures the percentage of shots taken.  Fenwick, however, does not take shots blocked into account, so in the example above, if 5 of the 50 shots the Blackhawks took while Kane was on the ice were blocked, and 2 of Detroit’s shots were blocked, Kane’s Fenwick % would be 45/53 (the 53 is the 45 Chicago shots on goal (SOG)/missed shots plus 8 Detroit SOG/missed shots), which is 84.9%. Again, those numbers are wholly hypothetical.

PDO: PDO (which, unlike a lot of stats with three letters, like ERA, RBI, SOG, and TOI, stands for nothing at all, and is merely the online handle of its creator) is the closest that the hockey community has yet quantified luck.  The algorithm for it is simply shooting percentage + save percentage, all at even strength (power plays and penalty kills don’t count).  The thinking behind it is that the number will always flow back to 1, aka the mean.  A team or player (save percentage is literally taken as the team’s save percentage while the player is on the ice for individual player measurement, same with shooting percentage) that is above 1 will regress back down, and a team/player below will rise back towards 1.  It makes a lot of assumptions, but is generally regarded as the best way to quantify luck that the community has yet come to.  Taking the Blackhawks as an example again (what? I’m a fan). This season the Blackhawks have a 8.54 shooting percentage, noted for math purposes as .0854.  The team’s save percentage is 92.6%, written here as .926.  That puts the team’s PDO at 1.011.  Being very close (within 1.1%) of the mean number means that the Blackhawks are performing about as well as would be expected. (shooting/save percent numbers from

Offensive Zone Start % This is determined by when a player starts his shift on a faceoff (so during play line changes don't count), and usually disregards neutral zone starts. Joakim Nordstrom, this season, starts 12% of his shifts (of the ones counted) in the offensive zone, meaning he starts 88% in the defensive zone (again, that's only for shifts that start with a faceoff).  Obviously Q isn't expecting him to light the lamp as much as (the currently injured) Kris Versteeg, who starts 70% of his shifts in the other team's end. The inverse of this is Defensive Zone Start %, which calculates how often a player is on the ice for a faceoff in the defensive zone as opposed to the offensive zone. (these numbers come from

TOI% of Competition: This one is relatively simple.  Basically it measures the quality of opposition by averaging the season long ice time of the players they play against. I'm not sure how SCH uses it, but it's typically split between defensemen and forwards (from this article from a couple years ago you see that, while Toews plays against both forwards and defensemen with high TOI, Bolland only plays against forwards with high TOI, so Bolland usually plays against the top six forwards, but the second or third defense pair, but Toews plays against the top lines and top defense pair).

Also here’s a few relatively simple ones quickly described:

Plus/Minus (+/-): the net number of goals scored on even strength and shorthanded. Usually used only on players.

Goals/Points per 60 minutes: a (usually season or career long) average of the number of goals/points a player scores in 60 minutes of ice time. For instance, a player who plays 600 minutes in a season and scores 10 goals and 50 points has 1 goal/60 minutes, and 5 points/60 minutes.

Penalty +/-: This is simply a calculation of how many penalties a player draws divided by how many he took (usually the box score will read “player x (penalty)- 2/5 min against player y” in that case player y draws the penalty, and player x takes the penalty).

Scoring chance: While there is no official designation, the Elias Sports Bureau (a well respected cornucopia of stats for all sports) defines a chance as a shot taken within an area shaped like a baseball home plate, starting from the goal crease, out to the faceoff dots, and up to the tops of the circles.  Here is a handy illustration:

It also includes shots taken on an odd man rush, a 3 on 2, a 2 on 1, etc.

So those are a few of the more well known advanced stats that will typically come up in a discussion.  There are others, but often they are only used by one site and have a competing metric on another site.  Corsi, Fenwick, and PDO are the “big ones”, and are generally accepted to some extent.  If you try and calculate these yourself, remember that the NHL’s site doesn’t list shot attempts as a stat, just shots on goal, which will skew the numbers dramatically (they have missed shots as a metric, and shots blocked by, but not shots blocked, and they’re both on a completely different page from the shots category).

There is one other metric that I would like to see, and that is zone time.  Basically calculate the amount of time the puck spends in each zone for players and teams.  I believe it can be a great metric, as it would include time spent passing and creating the best scoring opportunity, and controlling the play.

No comments:

Post a Comment