NHL Analytics Glossary

Corsi and Fenwick

Why Does it Tell Us and Why Does it Matter?

Both of these metrics are tracks of shot attempts for and against a team or player. Corsi tracks all shot attempts, including blocked shots and shots that miss the net. Fenwick removes blocked shots from the calculation. Corsi and Fenwick are both based on attempts at 5v5.

The advantage of Corsi has been that it is seen as a better predictor of future success than just evaluating a team’s goals for and against – a team that’s been outscored so far in the season but is generating more shot attempts is more likely to be successful moving forward than a team in the opposite position. Corsi is also more likely to carry over from year to year than measuring goals alone.

Fenwick has been shown to be less predictive than Corsi but does have one benefit – it can help to analyze a team whose defensive strategy is focused on blocking shots. If a team is able to make an advantage in that area a major focus (or conversely, if a team struggles with shot blocking), Fenwick can aide in analyzing how much of a team under or over performing their Corsi is due to blocked shots.

What are the Shortcomings?

The biggest shortcomings of both metrics are straight forward – all attempts are treated the same, even though clearly all shots are not equal. No one would argue that a point blank shot from the front of the net is equal to a shot from the blue line with no traffic in front of the net. Because of this, while Corsi is a better predictor of future success than goal based metrics, it has been shown to be less effective than metrics based upon the quality of the chances that are created.

Scoring Chances, High-Danger Scoring Chances, and Expected Goals

Why Does it Tell Us and Why Does it Matter?

These metrics show how effective a team is at generating chances to score. The first component of these metrics is the area of the ice that a shot comes from, with these zones given a rating from 1-3. A shot on a rebound or on the rush has one added to this, and a blocked shot has one deducted. For the most used versions, a score of 2 or better is a scoring chance and 3 or more is a high-danger chance.

Expected goals uses the same approach, but measures each chance by the likelihood of a goal resulting from it – adding these up results in how many goals should have been scored.

The ability to generate these chances has much better predictive properties of results than measuring all shots equally. Because of this, it can be better used to measure a team’s luck. It also has the ability to provide stronger analysis of goalies by measuring their performance against these opportunities.

What are the Shortcomings?

At least in the publicly available version of these metrics, there are still a number of other items missing that help to determine the quality of a chance. For example, traffic in front of the net and whether the shot is coming off of a pass would both give a clearer idea of the likelihood of a goal coming from a shot.

Additionally, the publicly available versions of the metric have not been shown to be more predictive of future success than Corsi. Some versions of the metric do claim to be better at this, but those typically do not have the information available for this to be measured independently.

Finally, in terms of being a predictive measure these statistics do not include the quality of the goaltending a team is facing, or the quality of the player taking the shot. Each opportunity is evaluated as if an average shooter is taking the chance against an average goalie.


Why Does it Tell Us and Why Does it Matter?

PDO combines a team’s shooting percentage with their save percentage, or does the same for when a player is on the ice – by default, this must always average out to 100% across the league.

PDO has widely been accepted as a form of evaluating luck – if a team’s PDO is higher than 100% they can be seen as having outperformed expectations so far, while a team below 100% can be expected to play better moving forward.

What are the Shortcomings?

The key shortcoming of PDO, and one I am interested in exploring further, is that most people use it by anticipating that every team and player should regress towards 100%. Typical to my last point on expected goals, this anticipates that all goalies are equal and all players have the same shooting percentage.

Instead of this, I believe teams and players should be evaluated based on their performance against an “expected PDO”, which would be based on individual past shooting and save percentages. However, the viability of this method would require insight into how much past save and shooting percentages can be predictive of future success.

Why Should These Matter to GMs and Coaches?

As with most analytics, the biggest advantage for GMs is too find assets who are being undervalued, or to avoid overpaying players who are outperforming their underlying metrics. This is especially true for metrics that do tend to stick from season to season – a player can easily become undervalued if they aren’t producing the points that their performance dictates, and that then creates an opportunity for a general manager to find a bargain.

For coaches, these metrics provide more data points to evaluate the best way to put together lines and combinations. Ultimately they will want to find the combination of lines that has put together the best metrics that are predictive of future success, like the ones above, instead of just measuring based on the goals for and against their current combinations.

4 views0 comments

Recent Posts

See All