### Fit-for-Purpose Verdict **What this dataset supports well:** This dataset enables **single-bout analysis** across 387,427 professional boxing matches. You can profile fighter attributes—age, height, reach, weight, stance, and record—at the column level, compare distributions across weight classes, and examine outcome patterns (wins, knockouts, decisions) within individual rows. Basic descriptive statistics, frequency counts, and univariate exploration of how attributes correlate with results are all reliable. The 100% referential integrity score confirms that every record is internally consistent, and the single-table structure means no broken foreign keys will derail queries. **What it cannot support and why:** **Cross-bout aggregation and fighter-level career analysis are unreliable.** With no validated joins and no separate fighter dimension table, you cannot safely roll up a boxer's career win rate, track performance over time, or compare the same athlete across multiple bouts without risking double-counting or misattribution. The 64% completeness score is driven by pervasive nulls in critical columns: **reach_B is 90% null** (349,554 of 387,427 rows), **judge1_B is 87% null** (335,367 rows), and **judge1_A is 87% null** (335,179 rows). Any analysis that depends on reach measurements, judge scorecards, or complete attribute sets for both fighters will exclude the majority of your data or produce biased results. Predictive models trained on these columns will either discard most observations or impute values that may not reflect reality. Mismatch detection that relies on reach differentials or judging patterns is similarly compromised. **Top remediation steps:** 1. **Investigate and backfill boxing_matches.reach_B** (90% null, 349,554 of 387,427 rows)—reach is a core predictor of fighting style and mismatch; without it, physical-attribute models are incomplete. 2. **Investigate and backfill boxing_matches.judge1_B** (87% null, 335,367 of 387,427 rows)—judge scorecards are essential for analyzing decision outcomes and competitive balance; this gap blocks scoring-pattern analysis. 3. **Investigate and backfill boxing_matches.judge1_A** (87% null, 335,179 of 387,427 rows)—parallel to judge1_B, this null rate prevents side-by-side judging comparisons and limits decision-outcome insights. Address these three columns to lift completeness above 80% and unlock the fighter-attribute and outcome analyses your goal requires. Until then, restrict conclusions to the subset of bouts with complete data and flag sample-size limitations in every finding.