Chapter 9 · §4.6

Human review is part of the design, not a patch

The final shortlist should not be formula-only.

A participant can have a good score and still be the wrong fit because the behavior is obviously unscalable, excessively noisy, too concentrated, dependent on simulator artifacts, or simply inconsistent with the kind of PM Finam would want to back.

That discretion should be explicit from the beginning.

STAGE 1 All participants ~ N entries STAGE 2 Formula-ranked candidates STAGE 3 Human review STAGE 4 Final shortlist A formula-only finalist may still be wrong because behavior is: · unscalable · excessively noisy · too concentrated · dependent on simulator artifacts · not a fit
← Prev: Alpha scoring All chapters Next: Weekly reports →