SlateSignalsNBA stat milestones, current roles and matchup research in one daily workspace.
Loading NBA research…
Automatically archived TEST milestones, capped at one per team and two per game. Selection requires current source timestamps, at least three current-season team appearances with stable minutes, historical threshold support, and no selected questionable teammate. Expected lineups are not confirmed lineups. These are uncalibrated estimates—not confidence ratings.
Open a player for recent game logs, ranges and availability assumptions.
Compare thresholds across the slate. Each cell separates an uncalibrated model estimate from recent hit frequency. Open a player for role evidence, ranges and game logs.
Thresholds mean at least the listed number. Historical hit rates describe the last available games, including prior-team history. Milestones overlap and should not be treated as independent picks. No sportsbook odds feed is connected.
Open a matchup to compare team efficiency, scoring context, recent historical games and player projections. Add sportsbook lines to inspect model differences alongside the underlying assumptions.
Current roster identities, official injury status and publicly reported expected starters define a supported rotation of up to 12 players. Out/doubtful players and listed two-way assignments are excluded. Starter targets are anchored before bench redistribution. Every simulated team rotation totals 240 minutes with explicit role bounds.
Current-team per-minute production is blended with historical or preseason priors using 150 prior minutes. Small new-role samples cannot fully replace the prior. Players without regular-season history blend preseason data with a position prior and remain limited-evidence estimates. Dated, expiring overrides can capture verified minutes restrictions.
Questionable-player scenarios show alternate minutes if those players sit. Such teams are excluded from the test shortlist. Expected-lineup, stale-source and limited-sample notices remain visible. Model probabilities are uncalibrated; no STRONG rating is assigned.
Immutable pregame snapshots retain model version, exact player IDs, selected milestones and capture time. Final box scores grade milestones, with DNPs voided. Opening night’s V1 evidence is preserved separately from tonight’s V2 test.
The V1 rotation allocated 310 minutes to nonparticipants and underprojected starter minutes. V2 addresses that failure; tonight is a new forward test, not a claim of validated improvement.
Team efficiency sets a shared scoring budget. Current-season possession estimates are blended with 20 prior-game equivalents, and coverage is shown per team. A shared mean-one production scenario adds 5% game-level variability across players; this assumption is not fitted calibration. Player points, threes and assists adjust before simulations, and the game score equals summed player point projections. Raw player scoring and the size of each adjustment are visible in Game Research. This consistency does not establish accuracy or calibrated probabilities.
These diagnostics grade the published ladder, separately from highlighted TEST selections. Overlapping thresholds are not independent bets.
Public-source snapshots are refreshed locally, then published. No sportsbook prop feed is connected. Injury and lineup inputs expire after six hours for shortlist display; refresh close to tipoff.