Presented by WHKYHAC + Sportlogiq
Entries were judged by 4 industry experts, spanning all areas of sports including: coaching, media and analysis. Awards will be awarded in the following categories (ie - Judges Choice, Newcomer awards, Best in Code, Innovation) with visualisations being judged against the following criteria
The intention of this scoring play visualizer was to meet a small set of goals: find trends in scoring plays by player or team, observe play progression based on danger area, and track player contributions to scoring plays beyond goals and assists. Users can see that only three Sonnet players scored on plays starting in their defensive side of the ice, while a majority of other scorers had at least one long-distance build-up. Instead, nearly every Sonnet goalscorer found success when their team utilized the offensive zone perimeter. Over 60% of goals were scored from the manually-defined high-danger area (faceoff dots to nearest ends of trapezoid). Additionally, a text table displays the selected player’s contributions to all goals scored by their team. This includes passes, zone exits, and loose puck receptions. R was utilized for data preparation, which included designating three danger areas, grouping values by possession for ease of visualization, and filtering to scoring plays. Upon Tableau import, parameters were used to give dynamic danger colors to event locations, define shapes based on event type, and populate the text table after player selection. Additionally, actions were employed to maximize interactivity while preventing excess data points on the final product.
Rebound value modeling shows that shots from the inner slot are better in terms of xG and the value of the rebound generated. A modified form of xG using rebound value accounting can then be compared to regular xG to examine their difference the largest difference occurring in the inner slot region where more frequent and higher value rebounds occur. The result of this model comparison is also examined through comparing trends in PWHPA and other data.
Visualizing the location of passes leading directly to shot attempts shows that passes leading to high danger shots in terms of xG primarily originate from near the goal line, while those leading to low danger shots primarily originate from the point. Overall, these passes originate from regions outside the slot (where players are more likely to shoot).
Visualizing successful offensive zone entry locations show that they are concentrated near the point as opposed to center ice.
Forwards, reasonably, are ranked high in both high danger passes and successful offensive zone entries, reasonably, with only 2 defensemen in the top 20 for high danger passes, and no defensemen in the top 20 for successful entries.
Data cleaning and visualization were done using R.
With this project, we aimed to create an Expected Threat (xT) model for analyzing goal-scoring possessions in the PWHPA. Using a Long Short-Term Memory (LSTM model), the project sought to determine the probability of scoring based on various possession factors. Data points included puck recovery, breakout method, pass and shot locations, shot type, etc. While acknowledging the inherent randomness in hockey, the model provided insights into events and players contributing to goal-scoring opportunities. We also developed a visualization tool, a Plotly Dash Web Application, allowing users to explore possessions in detail and xT heatmaps for each player to explore where they are the most impactful in the offensive zone. Key findings included: confirming top players' high xT ratings, the importance of limiting opposing team possessions and shots on goal to reduce goals against, and potential applications for power play strategies.
Overall, the project aimed to understand goal-scoring sequences, identify impactful players, and explore the potential for defensive metrics based on the xT model. We can build on the model to create new metrics to evaluate individual players' impact on the ice. -
I've always been curious about how hockey teams can use data to aid in player development and usage optimization. Although I have very limited experience in coding (everything was self taught in the last half year), I wanted to see if I could answer some of those questions by creating a dashboard that explored favourable environments for a player to create offence. With the help of sportyR and tidyverse, I made 3 plots.
1. A scatter plot of the shot location with their xG. Extended research could be in how can the player receive pucks in favourable ice.
2. A bar graph outlining teammates who had a successful event within 3 seconds of the player's shot. Extended research could be which teammates and what events were most favourable to creating a shot for the player (teammate chemistry).
3. A matrix scatter plot of the player's offensive zone events comparing the success rate of the event and their total attempts. These are skills that a player could focus on improving or maintaining.
If I had more time, I would have liked to go even further and break down the events into their sub-types.
For my first ever Shiny Dashboard and data competition, I am quite happy with the result and can't wait to learn and create even more in the future.