Should I start Andrew Beck or Nicholas Singleton in Week 18 (2026, Standard)?

FF uses position-specific provider weights, rescored CBS and FFToday stats, and the existing bounded form, defense and metric adjustments. Saved Boosted Trees and Workload have bounded model influence when available. Monte Carlo supplies the distribution; it does not get a duplicate mean vote.

Every player is ranked by the FF expected score (mean), whether or not a simulation distribution is available. Equal displayed scores use the higher FF-minus-ESPN margin when both ESPN forecasts are known, then unrounded FF, a comparable FantasyPros/Sleeper estimate, and stable identity order if still level. This tiebreaker is not a measured probability of beating ESPN. The median marker is additional information and never changes the order. Missing or availability-conditional ranges do not switch the ranking statistic. The window shows the experimental middle 80% (P10–P90), rather than the difference between two model means. Its observed coverage under this new version is still collecting. Availability-conditional distributions are labeled and are not shifted to a different center.

Missing sources and incompatible scoring are excluded. Bye and confirmed unavailable players remain excluded. ESPN, Sleeper and FantasyPros are the benchmark markers. The common scale covers the displayed outcome windows and available benchmarks.