FLB Analysis — Lebanese Division A (2025–2026)

A full basketball analytics pipeline for the Lebanese Basketball Federation's Division A, built on data scraped directly from RealGM: standings, team power rankings, player efficiency, playing-style profiling, win-correlation analysis, and a statistical player-team fit engine.

View the live report

Sample outputs: player writeups

Two narrative writeups generated from the analysis engine's own numbers — fact-checked against the underlying data before publishing:

• Player to Watch: Christophe Khalil — an under-the-radar forward on a non-competing team, surfaced by the fit engine's rebounding and efficiency signals.
• The Hidden Gem: Desi Washington — the league's assist leader, buried on a losing team, ranked the #1 statistical fit for three different contenders.

What's in the report

• League standings, team power rankings, and team strength scores
• What actually correlates with winning (evidence-based, not assumption-based — see caveat in report)
• Team identity/style profiles (pace, shot profile, ball movement, defensive identity) and each team's statistical needs
• Team profile vs. roster radar charts for the league's top teams
• Full player rankings and efficiency rankings, including a Lebanese-players-only breakdown
• Player-team fit engine: best external targets per team, "hidden gem" recommendations, and underused-player signals from current rosters
• Home/away splits and head-to-head matchup records

Pipeline

scrape_realgm_*.py          → data/raw/          (Playwright scrapers)
process_realgm_data.py      → data/processed/    (clean season stats)
process_boxscores.py        → data/processed/    (clean per-game box scores)
build_standings.py          → standings_2025_2026.csv
build_team_profiles.py      → team_profiles_2025_2026.csv
rank_teams.py                → team_power_rankings_2025_2026.csv
build_team_strengths.py     → team_strengths_2025_2026.csv
build_win_correlation_analysis.py → win_correlation_analysis_2025_2026.csv
build_team_style_profiles.py → team_style_profiles_2025_2026.csv
build_player_archetypes.py  → player_archetypes_2025_2026.csv
build_player_team_fit.py    → player_team_fit_scores_2025_2026.csv
rank_players.py / rank_player_efficiency.py → player rankings
build_home_away_analysis.py → home_away_analysis_2025_2026.csv
build_matchup_summary.py    → matchup_summary_2025_2026.csv
build_charts.py             → chart PNGs (embedded in report)
generate_html_report.py     → index.html (this report)

Validation scripts (validate_*.py) check for scoring/winner consistency, duplicate games, and missing team mappings.

Notes on methodology

• All normalized "score" and "tier" fields are relative to this season's 12-team field only — not an absolute or cross-season benchmark.
• The win-correlation analysis is based on n=12 teams (one season) — treated as a hypothesis to re-check next season, not proven fact.
• "Lebanese Players" tables use a hand-reviewed name heuristic (RealGM has no nationality field) — see the note in that section of the report.

Tech

Python, pandas, matplotlib, scipy (Pearson/Spearman correlation), Playwright (scraping). Report is a single self-contained HTML file with charts embedded as base64 PNGs — no external dependencies except Google Fonts.

Run it yourself

pip install pandas matplotlib scipy playwright
python scrape_realgm_lebanon.py
python scrape_realgm_games.py
python scrape_realgm_boxscores.py
# ... run remaining build_*.py / rank_*.py scripts in the order above ...
python generate_html_report.py