AHL Fancystats

Recently I talked to , who’s doing really interesting work estimating fancystats for the AHL.  As we all know, the AHL doesn’t publish play-by-play data that the fancystats wizards at , war-on-ice , and Stats.HockeyAnalysis can use.

Josh was so kind as to share the data he has generated with me.  Because of the way the AHL publishes its data, we only have all-situation statistics to work with (as opposed to the farm-raised, cage-free score-adjusted 5v5 stuff we get from the NHL).  So, keep in mind that we can’t draw the same narrative-busting conclusions like we do with NHL data. These are just general trends, but it’s better than nothing.

(click the images to get the full interactive page)

The takeaway I get from this is that John Hynes runs a real tight defensive ship in WBS. This confirms a lot of what we hear about that club from the media.  While I would guess that WBS usually doesn’t appear at the top of the Shots For category every year, they have been hit hard by call-ups to the NHL this year ( @nafsnep at chirpsfromcenterice.com is my go-to WBS Penguins/I-80 traffic blogger).

Like the possession chart, the PDO data is clouded by the lack of pure 5v5 information.  However, it’s probably reasonable to assume that Norfolk is on the unlucky side of the die, and Oklahoma is getting all the bounces.

Another way to look at the Penguins 2014-15 so far is to look at the shot attempts displayed on a line graph.

Here we have Corsi For and Against (shot attempts for and against) plotted out across the 41 games played so far (displayed as 10 game moving averages to smooth out the data).

The Penguins have done a pretty good job of attempting more shots than the opponent through this half of the season. You can see there are some instances where the opponent outshot the Penguins…let’s explore that more.

The picture below is the Corsi Differential (or plus/minus for you traditional statistic folks). Corsi Differential subtracts the shot attempts against from the shot attempts for (Corsi For-Corsi Against).  If the differential is +3, the Penguins out-attempted the opponent by 3.  If the differential is -3, the opponent out-attempted the Penguins by 3.  This statistic illustrates the same data points as in the graph above, just in a different way.

As indicated by the dotted line, the Penguins out-attempted the opponent by 2 attempts, on average.  However, there are two parts of the graph that dip below the zero line.  These are the same points in the previous graph in which the opponent outshot the Penguins.

I also introduced another variable into this graph: PDO.

The Corsi Differential chart above shows how the Penguins did when it came to shot attempts.  When we factor in PDO, which shows how the Penguins came out on the goalscoring and luck side of the picture, we get a more complete view.  The line is colored red/blue to indicate PDO at that particular point in the season.  Blue is lucky, red is unlucky.

PDO (the acronym doesn’t stand for anything) is the sum of Shooting % and Save %.  There is lots of research that indicates that the components of PDO regresses heavily to their respective means (SV% regresses heavily to the league average, which is around 92%.  SH% regresses heavily to 8%). This means that PDO also regresses to the league average of 100.  PDO is commonly used as a way to measure how lucky a team is at any point in a season.  Anything below 100 is bad luck, anything above 100 is good luck. Either way, you should expect it to regress back to 100.

Notice how the color shifts back and forth with no discernible pattern.  The hockey gods are a finicky bunch.

We can break down how exactly PDO fluctuates by breaking it out into its component parts: Save Percentage and Shooting Percentage.

Notice how both shooting and save percentages vary wildly, with no discernible pattern.  This is why the NHL is crazy and unpredictable.  See also: fun and maddening.

Since the Penguins recently played their 41st game, I thought a good first post would be to recap the first half of the season. The Tableau viz below is a scatter plot of the teams the Pens have played so far. The X axis (horizontal) is “Corsi For Per 60”, or the amount of shot attempts (shots, missed shots, shots blocked, goals) the Penguins took against the opponent , normalized against 60 minutes of TOI (Time on Ice). The Y axis (vertical) is “Corsi Against Per 60”, or the amount of shot attempts the opponent took against the Penguins , again normalized against 60 minutes of TOI.

(click on the picture to get the actual interactive visualization)

The easiest way to read this graph is to break it down into 4 sections (quadrants).

  • Top left: The opponent attempted many more shots than the Penguins did in the game
  • Top right: Both the opponent and the Penguins attempted a large amount of shots
  • Bottom right: The Penguins attempted many more shot attempts than the opponent
  • Bottom left: Both teams had few shot attempts

The team logos on the graph are also sized according to how many shot attempts (by both teams) were taken in that game. So, a small logo indicates that there were few shot attempts in that game, while a large logo indicates a large amount of total shot attempts.

The label below the logos is the Corsi +/- (the differential between the Penguins shot attempts and the opponent’s shot attempts). To take two extreme examples: on December 20th the Panthers smoked the Penguins, attempting 27 more shots than the Penguins did. On November 14th the Penguins demolished Toronto (as most NHL teams are wont to do), attempting 31 more shots than the Leafs.

There are also some live filters I enabled on the right side of the viz. You can select only the Penguins’ home games, only opponents in the Penguins’ division, or filter by Conference.  Feel free to fool around with those, you won’t break anything.

If you hover over a logo on the graph, the tooltip for that particular game will appear. This popup contains key information about that game, i.e. date and Corsi data.

This is the first post at Null_Hypothesis Hockey. I hope to regularly put up posts about hockey from a particular perspective. The goal of this site is to look at hockey from an objective and rational point of view, using statistics. I’ll be posting a lot about the Penguins because they are the team I know the most, but I will probably write about other players, teams, and the NHL at large as well.

A lot of the data I will use in my posts will contain data taken from War on Ice and Stats.HockeyAnalysis . These are two of the main websites I use to research advanced stats (aka fancystats), though of course there are many more that I will mention in the future.

Anyways, thanks for reading.