Sports Analytics

• A. Video footage of a basketball game
• B. A spreadsheet recording each player's points, rebounds, and assists per game ✓
• C. Fan comments posted on social media after a match
• D. Audio commentary from a live radio broadcast

Why: Structured data is organized into predefined fields, such as rows and columns in a spreadsheet or database. A table of per-game player statistics fits this format, while video, audio, and free-text comments are unstructured.

• A. Data about the weather conditions at a stadium
• B. Ticket sales and attendance figures
• C. Granular records of discrete actions during a game, such as passes, shots, and tackles ✓
• D. Long-term financial data for a sports franchise

Why: Event data captures individual actions within a match or game, each typically tagged with a time, location, player, and outcome. It is the foundation for many advanced metrics like expected goals and pass maps.

• A. A player's average points per game ✓
• B. A model predicting next season's win probability
• C. A simulation of possible game outcomes
• D. A regression estimating a player's marginal impact on team performance

Why: Descriptive statistics summarize what has already happened, such as averages, totals, and percentages. Predictive models, simulations, and regression-based impact estimates go beyond description into forecasting or inference.

• A. Poor data only affects the visual appearance of charts
• B. Errors or missing values can lead to misleading conclusions and poor decisions ✓
• C. High-quality data is only needed for marketing purposes
• D. Data quality matters only when working with video footage

Why: Analysis is only as reliable as the data behind it. Missing records, mis-tagged events, or inconsistent definitions can distort metrics and lead coaches or managers to incorrect conclusions.

• A. A sample is always larger than a population
• B. A population includes every relevant observation, while a sample is a subset used to draw conclusions ✓
• C. A sample refers only to player data, while a population refers to team data
• D. There is no meaningful difference between the two terms

Why: The population is the full set of relevant observations, such as every shot taken in a league season. A sample is a subset of that population, and analysts use samples to estimate patterns when collecting full data is impractical.

• A. The total number of points a player has scored across the entire career
• B. The average number of points a player scores in each game they play ✓
• C. The percentage of shots a player makes from the field
• D. The number of points a player creates for teammates through assists

Why: Points Per Game is calculated by dividing a player's total points by the number of games played, giving a per-game average that allows fair comparison between players with different amounts of playing time.

• A. To measure how far a batter hits the ball on average
• B. To measure how often a batter gets a hit per at-bat ✓
• C. To measure the number of runs a batter drives in
• D. To measure how many pitches a batter sees per plate appearance

Why: Batting Average is calculated as hits divided by at-bats, showing the proportion of official at-bats that result in a hit. It excludes walks, hit-by-pitches, and sacrifices from the denominator.

• A. The total number of goals a team is expected to score over a season
• B. The probability that a given shot will result in a goal, based on historical data ✓
• C. The number of goals a player scores compared to their teammates
• D. The average distance from which a team takes its shots

Why: Expected Goals assigns each shot a value between 0 and 1 based on factors like shot location, angle, and type of assist, reflecting how likely a shot is to become a goal based on similar shots in historical data.

• A. A player's per-minute statistical contributions adjusted for pace and league averages ✓
• B. Only a player's defensive contributions such as steals and blocks
• C. The total salary a player earns relative to the team's payroll
• D. The number of minutes a player spends on the court per game

Why: PER condenses a player's positive and negative statistical contributions into a single per-minute number, adjusted for team pace and league average so that players can be compared fairly across different eras and teams.

• A. OBP only counts home runs, which are the most valuable hits
• B. OBP accounts for walks and hit-by-pitches, which also help a batter reach base ✓
• C. OBP is easier to calculate because it uses fewer statistics
• D. OBP measures how fast a batter runs between bases

Why: On-Base Percentage counts any way a batter reaches base, including walks and hit-by-pitches, giving a more complete picture of offensive value than Batting Average, which ignores these outcomes.

• A. Record the raw scores of every game played
• B. Estimate the relationship between one or more input variables and an outcome variable ✓
• C. Design the layout of a stadium
• D. Track player injuries in real time without any data

Why: Regression models quantify how changes in predictors (such as shots taken or minutes played) are associated with an outcome (such as goals scored or wins), which is central to statistical modeling in sports.

• A. Categorical variable
• B. Ordinal variable
• C. Dependent (response) variable ✓
• D. Independent (predictor) variable

Why: Points per game is the outcome being predicted, so it is the dependent or response variable. Minutes played is the predictor, or independent variable.

• A. The model is 80% accurate on every single prediction
• B. 80% of the variation in team wins is explained by the model's predictors
• C. Teams win 80% of their games
• D. The predictors are 80% correlated with each other

• A. Possession time causes goals in every match
• B. There is a positive association between possession time and goals scored ✓
• C. Goals scored cause possession time to increase
• D. Possession time and goals are unrelated

Why: Observing that two variables move together shows an association, but it does not by itself prove causation. The relationship here is positive, meaning higher possession tends to accompany more goals.

• A. To make the dataset smaller so calculations run faster
• B. To evaluate how well the model generalizes to data it has not seen before ✓
• C. To ensure the model memorizes every past game exactly
• D. To remove all outliers from the dataset automatically

Why: Testing on held-out data reveals whether the model has learned real patterns or simply memorized the training data, which is essential for reliable predictions on future games.

• A. Pie chart
• B. Line chart ✓
• C. Scatter plot with no time axis
• D. Donut chart

Why: A line chart displays change over time, making it ideal for tracking a sequential metric like points per game across a season. Pie and donut charts show parts of a whole, not trends over time.

• A. To compare team payrolls
• B. To show the locations on the court where shots were taken and whether they were made ✓
• C. To display the total minutes played by each player
• D. To track referee decisions over a game

Why: Shot charts plot shot locations on a court diagram, often color-coded by make or miss, revealing shooting efficiency and tendencies by zone.

• A. Scatter plots automatically calculate player rankings
• B. Scatter plots show the relationship between two continuous variables, helping reveal correlation or outliers ✓
• C. Scatter plots are only used for categorical data
• D. Scatter plots display data as a single total value

Why: A scatter plot places each player as a point using two continuous variables, making it easy to see whether higher salaries tend to align with higher performance and to spot outliers.

• A. It cannot show more than three teams
• B. It can distort perception of values because depth and angle make bars harder to compare accurately ✓
• C. It only works with percentage data
• D. It requires specialized software to view

Why: 3D effects can distort bar heights and make accurate comparison difficult, so 2D charts are generally preferred for clear, honest visualization of sports data.

• A. Using red and green as the only distinguishing colors
• B. Using many similar shades of one color for different categories
• C. Using colorblind-friendly palettes and adding labels or patterns in addition to color ✓
• D. Avoiding labels so the colors carry all the meaning

Why: Colorblind-friendly palettes combined with labels or patterns ensure viewers with color vision deficiencies can still interpret the data correctly.

• A. To describe what happened in past games
• B. To estimate future outcomes or performance based on historical data ✓
• C. To record player statistics in real time
• D. To design stadium layouts for better fan experience

Why: Predictive analytics uses historical data and statistical models to forecast future outcomes, such as game results or player performance, rather than only describing past events.

• A. Calculating the average points scored by a team last season
• B. Creating a bar chart of player salaries
• C. Estimating the probability that a team will win its next home game ✓
• D. Listing all players who were injured during the year

Why: Estimating the probability of a future win is a forecast because it projects an uncertain future event. The other options describe or summarize past data.

• A. The data used to evaluate the final model on unseen cases
• B. The data used to fit or learn the model's parameters ✓
• C. The future games the model is trying to predict
• D. The list of variables excluded from the analysis

Why: The training set is the portion of historical data used to fit the model and learn its parameters. A separate test set is typically reserved for evaluating performance on unseen data.

• A. Test sets always contain more data than training sets
• B. Performance on training data can be overly optimistic due to overfitting ✓
• C. Training data cannot be used for any calculations
• D. Test sets are required by law in sports reporting

Why: A model can memorize patterns specific to its training data (overfitting), so evaluating on held-out test data gives a more realistic estimate of how it will perform on new games.

• A. The model is definitely accurate
• B. The model is definitely inaccurate
• C. Nothing conclusive, since one outcome cannot confirm or refute a probabilistic forecast ✓
• D. The model's probability must have been wrong

Why: Probabilistic forecasts must be judged over many predictions. A single outcome is consistent with a 70% probability, so one result alone cannot validate or invalidate the model.

• A. The total number of shots a team takes in a match
• B. The probability that a given shot will result in a goal, based on factors such as shot location and type ✓
• C. The number of goals a player is expected to score over a season
• D. The average distance a team covers during a match

Why: Expected goals assigns each shot a probability of scoring based on historical data about similar shots, such as distance, angle, and body part used. It helps evaluate chance quality rather than just the final score.

• A. Points per game
• B. Offensive rating ✓
• C. Total rebounds
• D. Minutes played

Why: Offensive rating estimates points scored per 100 possessions, which adjusts for pace and allows fair comparison of efficiency across teams that play at different speeds.

• A. A rule that limits how many players can press at once
• B. A specific event or cue, such as a bad touch or backward pass, that prompts a team to start pressing the opponent ✓
• C. The moment a team switches from defense to attack
• D. A substitution made to increase defensive intensity

Why: A pressing trigger is a situational cue that signals the right moment to press aggressively, such as a poor first touch by the opponent or a pass to a player with their back to goal.

• A. To show the total distance each player runs
• B. To visualize how players connect with each other through passes, revealing key players and structural patterns ✓
• C. To track the speed of each pass
• D. To count the number of goals scored from passing sequences

Why: Passing network diagrams represent players as nodes and passes as links, helping analysts identify central players, preferred combinations, and how the team's structure shifts during a match.

• A. Possession percentage is always inaccurate
• B. Possession value considers where and how possession is used, since possession in dangerous areas matters more than possession far from goal ✓
• C. Possession value is only used in tennis
• D. Possession percentage cannot be measured in team sports

Why: Holding the ball deep in one's own half adds little threat, while possession near the opponent's goal creates scoring chances. Possession value weights touches by their positional threat, giving a more meaningful picture of control.

• A. Using too many charts in team presentations
• B. Collecting and using athletes' personal data, such as biometric or health information, without clear consent ✓
• C. Charging high subscription fees for statistics websites
• D. Publishing player statistics before games are finished

Why: Sports analytics often relies on biometric, health, and performance data about individual athletes. Ethical use requires informed consent, clear rules about who owns the data, and protection against misuse such as using it to discriminate or to pressure players.

• A. Historical data is always perfectly accurate
• B. Past performance may not account for changes in age, injury, coaching, or team context ✓
• C. Historical data cannot be stored digitally
• D. Predictions based on the past are always more reliable than current observations

Why: Models trained on past data assume the future resembles the past. Players age, get injured, change roles, or move to new teams, so relying solely on historical patterns can produce misleading forecasts.

• A. Because models cannot process numbers
• B. Because the model may be overfitted to noise in the training data and fail to generalize to new situations ✓
• C. Because sports data is always incomplete
• D. Because analytics models are never tested

Why: Overfitting occurs when a model captures random fluctuations in the training data rather than genuine underlying patterns. Such a model looks impressive in-sample but performs poorly on new seasons or players, which is why validation on held-out data matters.

• A. Analysts only work during evening games
• B. Analysts focus on metrics that are easy to measure rather than those that matter most ✓
• C. Analytics is only used by teams with bright lighting in their arenas
• D. Data is only collected in well-lit stadiums

Why: The streetlight effect refers to searching where it is easiest to look rather than where the answer is likely to be. In sports, teams may overweight easily counted statistics like points or goals while neglecting harder-to-measure contributions such as defensive positioning or off-ball movement.

• A. Keeping all model outputs secret from the players involved
• B. Using analytics only as one input alongside human judgment, and being transparent about how metrics are used ✓
• C. Allowing the algorithm to make all final roster decisions automatically
• D. Ignoring player feedback because data is objective

Why: Analytics should inform rather than replace human judgment. Being transparent about how metrics are used, and giving players a fair opportunity to understand and respond to evaluations, reduces the risk of unfair or opaque decisions that affect livelihoods.