Grade Flip.

Expected profit is the average net dollars per attempt, given the hit-rate profile at the top of the page (35% PSA 10, 45% PSA 9, 15% PSA 8, 5% below). The profile is a modern-near-mint prior — not measured, not per-card. Older cards and cards with known print defects hit lower.
Break-even PSA 10 is the P(PSA 10) you would need for the pick to exactly cover fees + grading + shipping + the raw buy — holding the other hit rates as assumed. The lower it is, the less the pick depends on the guess. Fees are 15% + $1.30 per sale (eBay), grading is the cheapest PSA tier that covers the expected slab value, shipping to the grader is $8. Time to grade is on the tile — a 20% expected return over 90 days is a real number; annualising is not.
Every pick's detail view also shows live snipes when available — matched eBay listings that identified as this card and scored PSA 10 or 9 on centering. The list refreshes every 12h via a scheduled job, so the buys under each pick reflect what's actually listed right now. The auto-snipe command is still there for on-demand runs.
Undervalued finds at the bottom of the page is a separate sweep — broad eBay queries where mistitled or generic listings ("pokemon lot", "pokemon holo") hide singles worth much more than the ask. Every hit is identified visually with a tighter gate than the per-pick snipe (pHash + ORB agreement + title cross-check + market ≥ 3× ask). A confidence bar rides on each row so you know how much trust to put in the ID.