MLB consensus rankings

How the rating is built. Every team gets a rating in runs better than an average MLB team on a neutral field, from each of three pillars, and the composite is their weighted sum. Results is what actually happened, in two parts: a run-differential rating solved across the whole schedule, so beating good teams counts for more than beating bad ones and margins are capped at 5 runs, blended 65 / 35 with a win rating so a win counts beyond its margin. Betting markets is what money says. Baseball has no spread worth reading, since the runline is fixed at a run and a half with the price doing the work, so the rating is fit to ten weeks of closing moneylines, the most recent weighted heaviest, and converted to runs at the rate the season itself sets. The futures board is shown at 0% and scores nothing: by September it had separated all thirty teams into fifteen distinct prices, and what it is really pricing is the playoff path rather than the team. Every club has a moneyline in all six of its games a week, so the lines already carry what the futures would, and carry it better. Analytics models is what the models say: every live model, placed on the same points scale by its position against the market. Full-season weights are results 15%, markets 50%, models 35%; results phase in over the first 8 weeks because a September record predicts almost nothing, and a pillar with nothing to say this week hands its share to the others. Results carry their full 15%. There are no media rankings and no human polls in the score. Voters anchor on where a team started the season, reward reputation, and see a fraction of the games they rank; the three pillars here are each accountable to something real, a score, a price, or a measurement. Two columns are shown and not scored. vs MKT is wins above what the closing moneyline expected, and Luck is runs scored above or below the BaseRuns estimate from the team’s own hits, walks and extra bases: the part of the record that was sequencing rather than production. Both were tested and neither made the ranking better, which in a 162-game season is the point, so they are reported rather than scored. Résumé vs market is the results rank minus the market rank: a large positive number is a team whose record the market does not yet believe, a large negative one is a favourite that keeps losing. A source whose data is more than 30 days old is excluded from scoring and shown struck through.