Grading Assistant

Overview

The Grading Assistant helps you decide whether to send a card off for professional grading instead of selling it raw. For every card, RocketVault pulls live PSA 9 and PSA 10 comparable sales, subtracts the per-card grading fee and selling fees, and shows what you would actually net at each grade outcome — not a prediction, just the math.

The Grading entry point lives in the sidebar under AI Tools → Grading.

What it shows you

For each card, the Assistant returns:

• Raw price — what the card is selling for ungraded today.
• PSA 9 comp — current market price for the same card slabbed at PSA 9.
• PSA 10 comp — current market price at PSA 10.
• EV range — what you would net at PSA 9 (low end) and PSA 10 (high end), after grading fees, return shipping, and selling fees.

You also get the four pillar sub-scores from the photo (centering / corners / edges / surface) so you can form your own judgement on whether the card is likely to grade high.

Supported graders

The fee schedule covers four grading services and all common service tiers:

Fees come from the most-recently-verified rate card and fold into the EV calculation. The dashboard surfaces the Last verified date so you know whether a grader has just announced a price change.

How to run an analysis

The unified Grading page (/grading)

The Grading page is a single screen for filtering, picking, analyzing, and reading results.

1. Filter with the GradingFilterBar — set, year, player, status, and free-text search.
2. Select the cards you want to analyze with the row checkboxes.
3. Toggle "Include prices" if you want the SCP comp + EV computation in this run. Off by default for a faster, cheaper Gemini-only pass; flip it on when you actually want the dollar numbers.
4. Click Analyze. The backend enqueues a BackgroundTask and the BulkAnalysisRunner polls until completion. Up to 50 cards per run (mirrored from the backend's MAX_BULK_CARDS cap).

When the job finishes you'll see an EV range and pillar scores on every analyzed card, ranked by ROI.

Per-card detail (/grading/:cardId)

Click any card row to drill into its detail page. The page renders the same CenteringAnalysisPanel and the new PricesRangePanel — no PSA-band prediction anywhere.

Bulk candidate picker (/grading/bulk)

The /grading/bulk route is the dedicated Phase-2 surface for ROI-ranked candidate discovery: paginated, sorted by EV upside, designed for sellers who want to mass-evaluate inventory and queue a batch of submissions.

Reading the EV range

The two numbers the dashboard shows are calibrated to net dollars after every cost — that's what makes them comparable.

• EV @ PSA 9 = (PSA 9 comp × (1 − selling fee %)) − grading fee − return shipping
• EV @ PSA 10 = (PSA 10 comp × (1 − selling fee %)) − grading fee − return shipping

So if a card shows +$11 to +$140, you would clear an extra $11 over selling raw if it grades PSA 9, and an extra $140 if it grades PSA 10. If either endpoint is negative, grading would lose money even on the upside.

When a comp is missing (rarely-traded card, no recent PSA 9 sale, etc.) the dashboard shows "—" for that endpoint instead of fabricating a number.

Inspecting condition on a card

Click any card from the candidates table to drill into its per-card grading detail page. The page is built around one question: what's the condition of this card, at a glance? No scrolling, no decoding rulers — the verdict is the first thing you see.

Verdict-first hero

The top of the page is a single banner with everything you need to size up the card in under a second:

• Overall — the average of the four pillar scores (centering, corners, edges, surface). One color-coded number, sized large.
• Pillar chips — Centering · Corners · Edges · Surface. Each shows its score and is color-coded green / yellow / red so a weak pillar jumps out.
• Weakest area callout — when one pillar is dragging the grade down, an amber pill calls it out by name (e.g. "Weakest: Corners — worst corner BL"). Click it and the page scrolls and pulses the side viewer where that issue lives.

The hero stays sticky as you scroll on mobile so the verdict never leaves the screen while you inspect the photos.

Hover (or tap) any corner to zoom

The card photos render as the original images — no rulers, no labels obstructing the corners. To inspect a corner up close:

• Desktop: hover any of the four corners. A 4× zoomed popout appears next to the card, with the corner's score and a "Worst corner" tag if applicable.
• Mobile: tap a corner. A bottom sheet opens with TL / TR / BL / BR tabs. Pinch or double-tap inside the sheet for an even closer look. The worst corner pre-selects when you open it.

The corner the analysis flagged as worst wears a subtle pulsing red ring on the card image itself, so your eye is drawn there before you even start hovering.

Centering guide (optional)

A single button in the page header toggles a centering guide — four thin lines drawn on the card image at the detected card-art boundary. The pair on the side that's bottlenecking the centering score (left/right vs. top/bottom) renders thicker and color-keyed to severity. The guide is on by default and your preference is remembered across cards.

The guide is intentionally minimal — no numbers baked into the picture, no dimension bars over the card art. If you want the raw measurements, expand the Measurements & raw data accordion at the bottom of the page.

Measurements & raw data accordion

Collapsed by default. Open it to see the per-side breakdown: Left/Right and Top/Bottom ratios, worst-side percentages, sub-score chips, and any analyst notes. This is the same data the previous version showed below each card; it's just out of the way unless you want it.

Including centering images in eBay listings

The per-card panel also has the "Include centering images in eBay listing" toggle. When enabled, the next time the card is published to eBay, the annotated centering overlay (the version with rulers and dimension bars) is uploaded as an additional listing photo. Buyers see the centering measurements right on the listing — useful for higher-end raw cards where the buyer is going to ask anyway.

This is a separate output from what the grading detail page shows you. The page itself uses the clean image; the listing upload uses the annotated version.

Connecting the dots with the rest of the app

• Status pipeline: cards with status Ended (on the Cards page) are valid grading candidates. Run them through the Assistant before deciding whether to relist raw or send for grading.
• Republish path: if you skip grading and republish an ended card, RocketVault mints a brand-new eBay listing (new item ID) so analytics, watchers, and the URL all reset.

Common questions

Does the Assistant cover non-sports cards? Yes — anything with a PSA / SGC / BGS / CGC market. Pop reports drive the comp prices, not categories.

Why doesn't the EV use a weighted blend of PSA 9 / PSA 10? Because the blend depends on a grade prediction, and we don't predict. Showing the two endpoints separately lets you apply your own confidence to the call.

Can I override the per-card return shipping estimate? Default is $2.50/card based on a typical 5-15 card submission. Per-submission overrides are coming with the submission builder — track the changelog for that.

Are fees automatically updated when graders raise prices? The schedule is verified by hand and dated. We update it on each grader announcement; the Last verified date in the dashboard tells you whether you're on the current rate card.

What if my card has a borderless or no-detected-border layout? The analyzer treats null borders as borderless and computes pillar scores accordingly — earlier in 2026 a regression here would crash the page; that's been resolved.