Where it standsreading the corpus…
Where it disagrees with you
the model's number minus yours, across every prospect it can place
model is loweragreemodel is higher
Two models, and where they split
production reads what he did; pedigree reads everything production is blind to
The cohort calibration
the production model's raw scale, corrected against your board
The combined call
the two models weighted by how well each has actually tracked you
Your work list
where the two models disagree most — the only rows that carry information
What it has learned from you
every correction you make becomes a standing adjustment
What it actually knows
players per cohort, by league and age. Empty means the model stays silent.
Does the ladder hold up
a higher rung should produce better players, or the model is not measuring anything
The biggest argumentsworth settling first