Alongside the Research Paper competition at the MIT Sloan Sports Analytics Conference are the research poster displays. Research posters have always been a part of the SSAC (at least since I started attending in 2010), and even the research paper finalists had accompanying posters. In recent years more attention has been paid to the research posters and prizes have been awarded to the best ones as determined by conference attendees.
The format of the Research Poster competition is similar to the Research Papers, with abstract submissions used to determine which authors are invited to submit full papers. The full papers are then evaluated by the SSAC jury who select two groups: the eight Research Paper finalists who will present on stage and the twelve Research Poster finalists who will display their work throughout the conference.
To put it in soccer terms, it’s a bit like the paths into the UEFA Champions League and Europa League. To be sure, the Research Paper competition carries a lot of prestige, but all of the work presented on the Research Posters is of high quality.
You can find the Research Poster abstracts and their full papers here (scroll past the Research Paper finalists).
One difference between the Research Posters and the Research Papers is that there are no dedicated tracks in the Research Posters, so there is a greater diversity of sports represented. As expected, there are six basketball-related posters, followed by three soccer posters, and one each for tennis, ice hockey, American football, and Australian rules football. (One poster is more general but uses data from basketball and ice hockey.) The fact that a quarter of the posters are soccer-related is indicative of the continued high volume and quality of soccer analytics research at the moment.
Here are the finalists:
- Automatic classification and player tracking of NFL game video [VERY interested in this, lots of cross-over appeal]
- A critique of win probability models in basketball [Again, a lot of cross-over appeal]
- Studying NBA double-team strategies using deep learning
- Reassessing NBA possessions by modeling them as Markov chains [Very intriguing as always from Luke Bornn’s group]
- Quantifying and classifying transitions in soccer [Work from Patrick Lucey’s group at STATS]
- Evaluating soccer set-pieces using expected goals and deep learning [More from the Lucey group]
- Learning a basketball ghosting model using wearables and deep learning [Yet another from Lucey group. Seems like a companion piece to the Research Paper finalist]
- Predicting emotional states in tennis players from in-game broadcast images [An application of image classification ]
- Drafting errors and decision-making theory in the NBA
- Classifying and identifying teams from deep learning of individual trajectories [From the Bornn group. Lots of implications for sports data companies]
- Extending expected goals by evaluating off-ball scoring opportunities [Very interesting paper]
- Injury risk modeling using neural network models
The posters show a good range of state-of-the-art research in sports analytics, and the overarching themes of this year’s research are deep learning, player tracking data, deep learning, wearables, and deep learning.
Take a look at the posters in the halls when you get the chance, and have a nice chat with the presenters as well. They worked just as hard as the Research Paper authors to be able to present at SSAC.