Recent Posts

The Penguins Have Improved Consistency In Their Shot Outcomes

For the past few seasons, the NHL has transitioned from a heavy checking game to a speed game. With new elite players like Connor McDavid and Auston Matthews, the NHL has seen an influx of high-speed and high-accuracy players. The Penguins were one of the first teams to lead this change. With a Stanley Cup victory in 2016, the Penguins showed that a high-speed high -accuracy team could prosper in the NHL.

A trend that has piqued my interest in the modern NHL is shots per game as well as shot accuracy (shooting percentage). Let’s take a look at how the past few Stanley Cup champions fared in shots per game and shooting accuracy.

Team Shots per Game Shooting Percentage
Chicago Blackhawks (2014-2015) 1st (33.87) 26th (7.9%)
Pittsburgh Penguins (2015-2016) 1st (33.12) 16th (8.9%)
Pittsburgh Penguins (2016-2017) 1st (33.48) 4th (10.1%)
Washington Capitals (2017-2018) 31st (29.01) 2nd (10.8%)

As we can see, the Capitals were an anomaly by being last in the league in Shots per Game compared to the previous three champions. With the data, we can state that the more shots per game, the more likely a team will win. Current research shows that more shots per game will lead to more goals, and thus more victories. The reasoning behind this is that each shot has a probability of scoring. In order to maximize the probability of scoring, teams are shooting more and therefore giving themselves more opportunities to score.

When looking at shooting percentage we see a strictly increasing trend. In 2016-2017, shooting percentage dominated how teams were attacking on offense. The Penguins mastered how to shoot accurately as well as shooting with volume. The next year (2017-2018), the Capitals continued with shooting accurately but did not shoot with volume.

A key component in shot volume, as well as shot accuracy, is the type of shot. There are a total of seven shots types of shots that are tracked by the NHL: backhand, deflection, slap shot, snap shot, tip-in, wraparound, and wrist shot.

To see how certain players and teams excel with types of shots, I created a neural network model and a nearest neighbors model on shot data from 2015-2016 to 2017-2018. With three seasons of data, the models will have enough data to be at least 95% accurate. The output of both models is a probability that a shot will go in. The models take in the x coordinate and y coordinate of the shot, the type of shot, the distance of the shot, and the angle of the shot. Here is what happens when we run the 2018-2019 Penguins’ shot selection through the neural network models:

As we can see, the team excels at taking wrist shots in large quantity while the highest probability shot belongs to a tip-in. We can also look at how their shot probabilities fared versus shot distance:

As we can see, the yellow highlight around the curve is the confidence interval of the neural network. It follows that as the distance increases, the model gets slightly less certain given that there are very few shots from over 50 feet on net.

Throughout this entire season, the Penguins have had a lack of scoring depth. However, there has been an explosion of scoring from the Penguins back end. Let’s take a look at a few specific players and how they have performed this year when passing their data through the neural network.

Jake Guentzel has had a breakout year. His production and consistency have skyrocketed ever since he established himself as a top line option for the Penguins. Flanking Sidney Crosby, Guentzel has not shied away from the spotlight. Here’s how his shot selection has fared this year:

We can see that Guentzel has become adept at using his wrist shot for high scoring opportunities. Additionally, Guentzel is extremely consistent with his tip-in attempts.

With Guentzel’s rise, we have seen the exact opposite from Bryan Rust. Though Rust has been injured for a good portion of the season, he has not been producing to his expectation. Let’s see how he fared statistically:

Looking at Rust’s results, there seems to be a disconnect from the statistics and what seems to be happening on ice. His wrist shot has an average probability around 12%, which speaks to the quality of the shots he takes. However, we see his lack of production due to his lack of opportunities.

Someone who has played at his elite potential is Kris Letang. If he stayed healthy, he would have been a dark horse for the Norris trophy. Kris Letang has been on fire this entire season. When healthy, he’s been a shining star for the Penguins defensive core. Let’s look at how he fared in the model:

Like Rust, Letang has had his fair share of injuries. However, even though he has missed some games, he continues to play at an exceptional level. With his snapshot probability being a little over 11%, Letang has helped the Penguins score from outside the crease and slot areas.

With the Penguins’ season ending, it has become quite apparent that they have had a lot of production from their top 6 forwards. However, during the playoffs, it will become vital that the Penguins’ top 6 forwards have support from the third line as well as the back end. With shot volume increasing each year, I expect the Penguins to employ a shot-first mentality this playoff run.

Sasank is a student at Carnegie Mellon University majoring in computer science and minoring in business. He is a huge hockey and baseball fan, but also enjoys following the Steelers and Spurs on the side. His passion lies in applying machine learning to sports analytics. As a mentor in the Carnegie Mellon Tartan Sports Analytics Club, Sasank does sports analytics research on hockey in order to bring more clarity to skater and goaltender performance. He can be reached on Twitter at @_svish.