Eleven good players are not a good team. A structure is.
Team KR consumes the players and never re-grades them, but it is not their average. It weights each player by how much his position moves winning (the positional-value layer the grade deliberately leaves out), by his real cross-package snap share, and by how central he is to the scheme, then rolls three phase KRs into one on the correct engine. It knows a pass-protection unit goes as its worst blocker and a pass rush goes as its best, that a missing load-bearing piece collapses everything downstream of it, and that a quarterback with no backup is the most fragile roster in the sport.
Grade the player at his position. Weight the position by winning.
A composite roster on the Pro engine. Team KR consumes the finalized player numbers, but it is not their average. It weights each player by how much his position moves winning, by his real snap share, and by his scheme demand, then rolls three phases into one.
Talent KR is the fit-blind aggregate: OVERALL x snap share x positional value, no fit and no fragility. The gap down to Team KR is the fit, the talent the scheme is not getting out.
A star back at 0.75 moves the team less than a good quarterback at 2.60, and the College engine forks the table where the run and option game lifts the back, the interior, and the linebacker. Each player pulls on one phase plus his special-teams term. The grade is position-relative on purpose, so the team layer can carry which positions win.
Illustrative on the real Team KR aggregation (the two-engine WAR layer, the 50/30/20 weight blend, the three-phase split, the Talent-versus-Team gap). Composite roster, demonstration figures.
Some units go as their worst. Some go as their best.
A roster is a set of units, and units do not all fail the same way. Some go as their worst member. Some go as their best. The engine adjusts the unit's contribution accordingly, and it never touches a player's OVERALL.
All else equal a team prefers an elite pass rusher to elite coverage depth, and the handling encodes it. The adjustments are bounded, disclosed, and applied to the unit, never a player. Fragility flags compound, and three or more is structurally fragile regardless of the number, with the quarterback cliff weighted heaviest. A roster is a set of units, and a unit is only as sound as the way it fails.
Illustrative on the real weak-link and strong-link handling (the bounded adjustments, the football-native fragility flags led by the quarterback cliff). Composite units, demonstration figures.
A missing piece does not cost one position. It cascades.
When a scheme is missing a load-bearing piece, the loss is not one position. It is every position that leaned on the structure. The cascade takes the whole phase down and shows its work.
Before the cascade runs, the Pythagorean reality check maps the team's actual scoring to a legend tier and flags a Team KR that floats more than twelve points above the results, and the fix is upstream, never a silent adjustment. The output is a set of layers, not one figure, and the engine names the structural gap and stops. Diagnose before optimize, at the roster level.
Illustrative on the real scheme-fit cascade (the compound phase penalty, the structural overrides, the disclosure), the Pythagorean check, and the multi-layer schema. Composite roster, demonstration figures.
A team is not the average of its players, and treating it as one is how a roster of good names underachieves and nobody can say why. The grade tells you how good each player is at his position, on purpose leaving out how much the position matters, so the team layer can carry the thing the grade must not: the positional value that says the quarterback moves winning two to three times more than anyone else, weighted before the player and forked by engine because the two games value positions differently. On top of that sit the interactions, because a roster is a structure and structures fail in structured ways, a pass-protection unit goes as its worst blocker and a pass rush as its best, and the engine credits depth where the unit is weak-link and stars where it is strong-link. And when a scheme is missing a load-bearing piece the loss is not one position but every position that leaned on it, so the cascade takes the whole phase down and shows its work. What comes out is not one number but a set: how good the roster is, how good it can become, how fragile it is, and whether it survives January. Build the list and you get a spreadsheet. Build the structure and you get a team.
Grade the players. Then build the structure.
Team Intelligence weights each player by the winning his position moves, reads the units by how they actually fail, cascades the compound breakdowns, and returns Team KR as a set of layers, never a flat average.