Delta

posted Aug 6, 2014, 12:21 PM by Robert Vollman
Courtesy of Tom Awad, creator of GVT, here are the 2013-14 individual totals for the Delta statistic, in spreadsheet and CSV format.

What is Delta, and when is it useful in place of the more familiar Corsi (attempted shots) or Fenwick (unblocked attempted shots) statistics? 

Expressed in its simplest form, Delta is simply Fenwick adjusted for Shot Quality, leading to an “Expected Goal” value. Tom first introduced Delta way back in 2010, before the concept of Expected Goals had become more common (courtesy of Michael Parkatti). Delta is a way of expressing Fenwick numbers while keeping them in units of goals, which is the fundamental currency that hockey is expressed in. 

While Shot Quality is not a significant contributor to team success or failure in the NHL, it does exist, and it can be expressed at the individual level as well as the team level: certain players have a play style that leads to them having fewer, high-quality chances, while others prefer outside shots, leading to high shot differential. Both styles have their benefits and may be suited to different types of players, but Fenwick cannot differentiate between them.

Corsi and Fenwick also have other limitations, they are highly influenced by zone starts as well as the score of the game. When the score isn't tied in the third period, the trailing team will typically have a huge advantage in shots that doesn't translate into goals, they are taking lower-quality shots and allowing high-quality ones. As a result, “Fenwick Close”, defined as Fenwick when the score is tied or within 1 goal in the first two periods, has become the default Fenwick measure. This works well in that it avoids bias, but ends up throwing away almost 35% of the data in the NHL, which is unfortunate. Delta can naturally correct for that by weighing each shot by the game score, achieving the same thing as score-adjusted Fenwick. 

Like Fenwick and Corsi, Delta can also be adjusted for Quality of Competition and Quality of Teammates. However, Delta is also easier to compare to other measures. When a player has an Adjusted Delta of +5.0 over the entire season, you know that it means that he contributed +5 goals through his possession play. It's harder to perform that equivalency when we say a player has a Fenwick of +3.0 per 60 minutes. 

Lastly, comparing Delta with Fenwick acts as a proxy for Shot Quality. This can be useful when trying to understand why a player or team has the results that they do; for example, Phil  Kessel last season had a Corsi For of 1230 and an Corsi Against of 1557, giving him very bad possession numbers (a Corsi % of only 44.4%). However, when weighted by Shot Quality, we can see that he was on the ice for an equivalent of 59 Expected Goals For and 64 Expected Goals against, a more reasonable 48%. This goes part of the way in explaining why, despite the terrible Corsi numbers, his team still managed to outscore its opponents by 8 goals at 5-on-5 while he was on the ice.
ċ
13_14_delta.txt
(141k)
Robert Vollman,
Aug 6, 2014, 12:21 PM
Ĉ
Robert Vollman,
Aug 6, 2014, 12:21 PM
Comments