[1910.11292] Predicting In-game Actions From the Language of NBA Players
It would be interesting to investigate and model these connections, and perhaps develop a theoretical bound for the predictive power that language could have in such situations.

Abstract: Sports competitions are widely researched in computer and social science,
with the goal of understanding how players act under uncertainty. While there
is an abundance of computational work on player metrics prediction based on
past performance, very few attempts to incorporate out-of-game signals have
been made. Specifically, it was previously unclear whether linguistic signals
gathered from players' interviews can add information which does not appear in
performance metrics. To bridge that gap, we define text classification tasks of
predicting deviations from mean in NBA players' in-game actions, which are
associated with strategic choices, player behavior and risk, using their choice
of language prior to the game. We collected a dataset of transcripts from key
NBA players' pre-game interviews and their in-game performance metrics,
totaling in 5,226 interview-metric pairs. We design neural models for players'
action prediction based on increasingly more complex aspects of the language
signals in their open-ended interviews. Our models can make their predictions
based on the textual signal alone, or on a combination with signals from
past-performance metrics. Our text-based models outperform strong baselines
trained on performance metrics only, demonstrating the importance of language
usage for action prediction. Moreover, the models that employ both textual
input and past-performance metrics produced the best results. Finally, as
neural networks are notoriously difficult to interpret, we propose a method for
gaining further insight into what our models have learned. Particularly, we
present an LDA-based analysis, where we interpret model predictions in terms of
correlated topics. We find that our best performing textual model is most
associated with topics that are intuitively related to each prediction task and
that better models yield higher correlation with more informative topics.

‹Figure 1: Shot location of all attempted shots for all the players in our dataset. A darker color represents more shots attempted at that location. Black lines represent the structure of one of the two symmetric halves of an NBA basketball court. (In-game Play-by-Play)Figure 2: The LSTM-TM Model. hn = hforward n ⊕ hbackward n , |hn| = |hforward n | + |hbackward n |. ⊕ denotes the vector concatenation operator. (The BiLSTM Models)

$$$$Figure 3: The BERT-L-T Model. n denotes the number of Q-A pairs in a given interview. Each Q-A pair is fed into BERT to produce a Q-A vector, and the resulting vectors are then fed in sequence to the BiLSTM. hn is generated in the same way as in the LSTM-TM model (see Figure ??). (The BERT Models)Figure 4: The BERT-A-T Model. Attention is applied over a sequence of Q-A vectors, which are produced by feeding the interview’s Q-A pairs into BERT. hcontext is randomly initialized and jointly learned with the attention weights during the training process. (The BERT Models)Figure 5: The BERT-A-TM Model. We use the same notation as presented in Figure ??. (The BERT Models)Figure 6: BERT-A-T prediction accuracy per player, relative to its accuracy for all players, for each prediction task. Each point in the graph is defined by Equation ??. (Results)Figure 7: BERT-A-T’s averaged positive class prediction confidence (p̂(ym i = 1)) as a function of the positive class topic probability (θi zm+ ) assigned to each interview. The computation of f(θi zm+ ; j), is described by Equation ??. (Associating Topics with Classifier Predictions)Figure 8 Figure 9 Figure 10: Correlation heat-maps of LDA topic probability (θi zm ) and the DNN positive prediction confidence (p̂(ym i = 1)) (top: (a) LSTM-T; bottom: (b) BERT-A-T), for each prediction task. (Topics Correlation with LSTM-T and BERT-A-T)