walrusquant/sports-analytic-skills

Review sports figures, tables, notebooks, and reports for chartjunk, fake certainty, cropped axes, baseline erasure, metric laundering, and weak reproducibility. Use when asked to clean up or audit analytical presentation.

Design and evaluate simple sports prediction baselines before accepting more complex models. Use for constant-rate, home-advantage, logistic, Elo-style, or market-reference comparisons.

Evaluate whether sports-model probabilities match observed frequencies. Use for Brier score, log loss, reliability bins, ECE, segment checks, and recalibration decisions.

Choose public sports data sources for a modeling question across NFL, NBA, MLB, NHL, college sports, soccer, and more. Use before acquisition code or whenever source coverage, grain, licensing, or historical depth is unclear.

Exploratory data analysis for user-provided sports data: grain, key integrity, coverage, missingness, entity balance, base rates, outliers, structural breaks, and leakage red flags. Use before feature engineering or model fitting.

Create and verify a portable Python environment for sports analysis. Use for machine setup, onboarding, dependency diagnosis, or reproducibility checks.

Record sports-modeling experiments with the hypothesis, data cut, validation charter, metrics, leakage status, decision, commands, and artifacts. Use for trials, model comparisons, and reproducible research history.

Define, review, and document point-in-time legal sports-model features. Use when creating rolling form, rest, matchup, rating, roster, injury, or contextual predictors.

Audit sports modeling tables and workflows for target, temporal, join, preprocessing, and split leakage. Use before trusting backtests or reported predictive performance.

Write a durable sports model card covering identity, intended use, target, decision time, data, features, baselines, validation, results, limits, maintenance, and kill conditions. Use when freezing or sharing a model.

Interpret sports models using held-out predictions, coefficients, error slices, largest misses, calibration context, and stability checks. Use after time-aware evaluation when explaining what drives a model and where it fails.

Load NFL schedules, play-by-play, rosters, and player or team statistics directly from nflverse with nflreadpy. Use for NFL acquisition, schema review, bounded snapshots, and preparing user-owned analysis artifacts.

Build and evaluate predictive sports models from user-provided modeling data. Use for binary outcome models, feature/model selection, chronological backtests, probability scoring, and comparison to baselines.

Load MLB Statcast, batting, pitching, standings, and player lookup data directly with pybaseball. Use for bounded pitch-level pulls, season tables, schema checks, and user-owned baseball data artifacts.

Build and evaluate sports strength models, including sequential Elo ratings, offense/defense splits, power ratings, and strict pre-event matchup features. Use for rankings, opponent adjustment, and strong prediction baselines.

Report sports analysis and modeling results with the question, data, methods, validation, baselines, metrics, interpretation, limits, figures, and reproduction pointers. Use for research notes, reports, and final answers.

Simulate game and season outcomes from user-supplied probabilities or ratings, summarize uncertainty, and test sensitivity to assumptions. Use for standings, win totals, matchup distributions, and scenario analysis.

Connect the optional sports_ds Python toolkit to the standalone sports analytics skills. Use when the user explicitly mentions sports_ds, wants its NFL/NBA/MLB public-data loaders or CLI, needs toolkit setup/troubleshooting, or wants to convert toolkit output into a skill's documented input artifact.

Define a sports analysis or prediction question, grain, decision time, baselines, primary metrics, validation, and acceptance criteria before choosing algorithms. Use at the start of any sports modeling project.

Create honest sports-analysis figures from user-owned data, including distributions, rates, rating trajectories, calibration plots, and walk-forward metric comparisons. Use for exploration and communication.

Load public multi-sport data directly with SportsDataverse Python. Use for NBA, MLB, NHL, college sports, soccer, and other supported league sources when a user needs schedules, box scores, rosters, or event data.

Guided statistical modeling for user-provided sports data: selecting models for binary, continuous, and count outcomes; assumption checks; effect sizes; time-aware inference; GLM diagnostics; and complete reporting.

Engineer and compare time-safe sports form features. Use for rolling windows, EWMA, rest, schedule gaps, early-season handling, and chronological evaluation.

Design chronological sports-model validation and a written evaluation charter. Use for walk-forward splits, grouped time folds, metric locking, tuning boundaries, and go/no-go rules.