Team Intelligence grades the roster. This changes it.
A rating tells you how good the team is. It does not tell you what to do next. Roster construction is the other half: it takes the same Team KR pipeline and re-runs it with a player added, cut, or moved, then reports how the number changes and why. The delta is the whole answer, and it reuses the one number you already trust. Same engine, different question, not how good are we, but who to keep, add, and move to get better.
Rank moves by what they do to the team, not by the player alone.
The engine does not invent a new score for a move. It re-runs the whole Team KR with the player in, and the change is the answer. That is why the best available player is rarely the best move. Here is the roster board for the composite Duke read, candidates ranked by Team KR delta, not by their own rating.
The 90 adds almost nothing; the team already has elite shooting, so a great shooter duplicates a strength. The 84 adds the most, because he fills the one thing the roster cannot do. The board is sorted by delta, and the delta does not follow the rating.
Real Duke Team KR read re-run with illustrative composite candidates. Demonstration deltas.
A duplicate star adds little. A gap-filler adds a lot.
The Team Intelligence read already found the one thing this roster cannot do. Roster construction points every move at it. Two candidates, one a bigger name, and the gap decides which is worth more to this team.
Six points of individual rating separate them, and the lower one is worth almost eight times more to this team. Sign the 90 and you have a better highlight reel; sign the 84 and you have a better team. Only a move scored against the gap can tell them apart.
Illustrative engine read on the real gap-fit delta structure, scored against Duke's real coverage soft spot. Composite candidates, demonstration figures.
You cannot just add. Each move trades against the cap, the budget, and the whole team.
The roster limit is full and the budget is finite, so a real add is a trade: someone comes off, the money moves, and the whole team re-prices. The engine runs the full chain and previews it live, nothing commits until you say so.
The add is worth +3.1 alone, but the cut costs a little back, so the true team gain is +2.5, and it comes in under the roster limit and inside budget while closing a fragility flag. That is the honest number: not what the player is worth, but what the whole move does, after everything it costs. Change any piece and the panel re-prices before anything commits.
Illustrative engine read on the real add-cut-reprice structure (roster limit, budget, Team KR delta, fragility). Demonstration figures.
A roster is not a pile of ratings; it is a system, and a move is only as good as what it does to the system. The engine refuses to score a player in a vacuum. It re-runs the team you already have with the move made, reads the change in the one number you trust, and ranks every option by that, the duplicate star that adds nothing, the gap-filler that adds everything, the add that pays for itself only after the cut. Do not build the best collection of players. Build the team with the highest delta.
Do not collect players. Build the team.
College Roster re-runs the team you trust with every move and ranks by the delta.