The optimizer answers a simple question for every card: should you send it to a grader, and if so, which one?
The math is straightforward, but the inputs aren’t:
- Per-grader, per-grade pop counts — how many PSA 10s, CGC 10s, BGS 9.5s, SGC 10s, TAG 10s exist?
- Per-grade sold comps — what’s the median PSA 10 sale price in the last 90 days?
- Pull-rate-style probabilities — given a card’s current condition, what fraction lands at each grade?
- Fees + shipping — PSA Express, CGC Premium, BGS Standard each have different cost structures.
The expected value calculation: for each grader, multiply the probability of each grade by that grade’s median sale price, sum, and net against fees. The verdict is GRADE if EV − fees > current raw price by a meaningful margin, SKIP if not, INSUFFICIENT_DATA if pop or comp data is too thin.
The v6 design (per mockups/output/13_optimizer_tool_v6.png) shows all five graders side-by-side, with a probability bar for each grade and the EV contribution surfaced. Users see the math, can disagree with a probability if they think their specific card is unusually clean, and trust the verdict because they can audit it.
Activates the moment commercial pop data integrates. See docs/decisions/2026-05-03-pop-data-sourcing.md for the GemRate plan.