One corpus, two surfaces
The browsable database and the ML predictor read from the same source of truth. No divergence between what you see and what the model learned.




























The browsable database and the ML predictor read from the same source of truth. No divergence between what you see and what the model learned.
Free-text posts are parsed into structured fields (stats, activities, outcomes, round) so filters actually work. Post IDs strip any personal detail.
The stratified pool spreads across focus and outcome buckets so heavy-STEM / heavy-t20 dominance doesn't bury the rest.