Peer-review panels (modelled here on the NIH 1–9 impact-score scale, where 1 is exceptional and 9 is poor) score five criteria, then average them into an overall impact score that maps to a percentile and, against a funding cutoff ("payline"), a probability of an award:
Direct costs = Budget / (1 + F&A_rate)
Indirect costs = Budget − Direct costs
Approach score = 1 + penalty(personnel% outside [55%,75%])
+ penalty(F&A_rate above the 58% negotiated cap)
+ |aims_focus| / 40
Innovation score = 1 + |aims_focus − 60| / 10
Overall = 0.20·Significance + 0.20·Investigator + 0.15·Innovation
+ 0.30·Approach + 0.15·Environment
(Significance, Investigator, Environment held at a fixed
baseline of 2.0 — this simulator isolates the two criteria
a grant writer actually controls: Approach and Innovation)
Percentile ≈ (Overall − 1) / 8 × 100
P(funded) = 1 / (1 + e^((Percentile − Payline)/6)), Payline = 15th pctile
- Personnel share — reviewers expect roughly 55–75% of direct costs on salaries/wages for most research grants; drifting outside that band reads as either an under-resourced team or padded personnel lines, and hurts the Approach score.
- F&A / indirect rate — most US universities negotiate facilities-and-administrative rates in the 50–60% range; requesting above the effective 58% cap used here without justification is flagged by program officers and directly penalises Approach.
- Aims focus — too broad (near −100) reads as unfocused and hurts Approach; too narrow/incremental (near +100) reads as safe but unoriginal and hurts Innovation. The sweet spot modelled here is a moderately bold, still-feasible aim (~60).
- Funding probability — a logistic curve centred on a 15th-percentile payline, the rough historical cutoff for many federal research-grant mechanisms: even a strong score only becomes a near-certain award once comfortably past the payline.