Design Basketball Shot Outcome Prediction
Problem Statement
Design a machine learning system to predict the outcome of a basketball shot (made or missed). The prediction must rely only on information available at the exact moment the player releases the ball.
Discussion Areas
Candidates should discuss the following aspects of the system design:
- Problem formulation
- Data collection strategies
- Feature design and engineering
- Modeling approach selection
- Choice of loss function
- Evaluation metrics
- Feature selection strategy
Example Input Signals
Consider the following types of input signals:
- Shot-level physics: Release speed, release angle, distance to the rim.
- Player-specific information: Player identity, historical performance, potentially represented via learned embeddings.
- Contextual data: Arena conditions, current game state, player fatigue, minutes played, and other pre-shot context.
Specific Question
If a univariate linear regression of the target against a single feature shows that the feature is not statistically significant, is it always safe to discard that feature? Justify your answer.
AI interview prep
Core dilemma
Balancing the complexity and real-time acquisition of high-fidelity physical and contextual data with the need for a robust, explainable, and deployable ML model that can make accurate predictions at the precise moment of shot release.
3 tradeoffs and 2 stress tests locked
Interview prep
Unlock the interviewer X-Ray rubric, tradeoff matrix, stress tests, clarifying questions, and key concepts.
Design Basketball Shot Outcome Prediction
Problem Statement
Design a machine learning system to predict the outcome of a basketball shot (made or missed). The prediction must rely only on information available at the exact moment the player releases the ball.
Discussion Areas
Candidates should discuss the following aspects of the system design:
- Problem formulation
- Data collection strategies
- Feature design and engineering
- Modeling approach selection
- Choice of loss function
- Evaluation metrics
- Feature selection strategy
Example Input Signals
Consider the following types of input signals:
- Shot-level physics: Release speed, release angle, distance to the rim.
- Player-specific information: Player identity, historical performance, potentially represented via learned embeddings.
- Contextual data: Arena conditions, current game state, player fatigue, minutes played, and other pre-shot context.
Specific Question
If a univariate linear regression of the target against a single feature shows that the feature is not statistically significant, is it always safe to discard that feature? Justify your answer.