Korean and Japanese baseball, priced off team run rates from the official league sites and shrunk toward the market consensus across every book quoting the game.
NPB publishes its starters; KBO does not. Japan files announced starters (予告先発) the day before, so where both are known this model prices the actual arms and says Starters known on the card. Korea only files after the previous game ends, so KBO stays team-level — it knows how good each roster has been over 120+ games, not who is on the mound. Guessing a rotation would look confident and be wrong often enough to be worse than admitting it.
Weighting follows that honestly: 50% model where the starters are known, 35% where they are not, with the disagreement capped either way. It is built to catch a lazy line, never to claim it knows something it structurally cannot.
Still being calibrated. Early slates show the model projecting more runs than the market on most games. Until that settles, treat the totals as research and lean on the caps — which is exactly why they exist.
Not on the tracked record. This is a research board while the model proves itself. Nothing here counts toward the graded W–L on the site.