The most expensive mistake in card investing is not paying too much — it is paying the right price for the wrong grade. Sellers anchor on the PSA 10 number, buyers anchor on the same number, and the PSA 9 that actually comes back sits at 30-50% less than what either party thought the deal was worth. Real comp-checking starts before you look at a single sale price.
Start With Condition, Not Price
Before you look at a single sale price, look hard at the actual card in front of you. Does it realistically have a shot at the grade you are pricing it for? A card with a visibly soft corner or centering that reads 60/40 is not a PSA 10 candidate no matter what a similar-looking listing sold for last week. If the listing photos do not let you evaluate all four corners, both edges, and the surface under direct light, ask for better pictures before you touch the comps. Pricing a PSA 9 candidate against PSA 10 data is how buyers get burned and sellers get flagged.
This is not a minor distinction. The PSA 10-to-9 price gap on a key rookie card regularly runs 2x to 5x. A 2020 Topps Chrome Tatis Jr. PSA 10 trades around $80-100; the PSA 9 trades around $20-25. If you comp the 10 and receive a 9, you have lost 75% of your projected value before fees. Honest condition evaluation is the highest-leverage step in the entire process. See also: how to negotiate to buy sports cards.
Pull Comps for the Realistic Outcome, Not the Best One
Once you know what grade range is actually plausible, pull recent sold prices for that range — not just the PSA 10 price, which is the number every seller and every excited buyer fixates on. For most raw cards purchased on eBay with average photo quality, a realistic submission targets PSA 8 or PSA 9, not 10. PSA grades raw submissions a 10 roughly 45% of the time for modern base cards in strong condition and considerably less for vintage.
Build a three-point price range: your realistic expected grade, one grade above as upside, and one grade below as the downside scenario. Look at:
- What the same card sold for at the grade you realistically expect
- What it sold for at one grade below, in case you are wrong
- Whether sales are clustered around a consistent price or scattered — scattered comps mean less certainty in your number either way
| Step | What to do | Common mistake to avoid |
|---|---|---|
| 1. Evaluate condition | Honest condition read before looking at any price | Assuming PSA 10 before evaluating corners or centering |
| 2. Pull comps for realistic grade | Sold listings for the grade you expect, not best case | Only checking PSA 10 price -- biased anchor |
| 3. Also pull one grade below | Price at PSA 9 or lower if you are wrong | Ignoring the downside scenario entirely |
| 4. Filter outliers | Check bid history on standout sales | Treating one inflated auction as the market rate |
| 5. Check spread | Tightly clustered = high confidence; scattered = low | Treating scattered comps as if price is known |
| 6. Run margin | Comps vs full cost stack (buy + ship + grade + fees) | Running margin against sticker price only |
Make Sure the Sales Are Real
This is the step almost everyone skips, and it is the one that protects your margin the most. Not every eBay sold listing represents a legitimate, arm's-length transaction. Two specific patterns inflate comp data regularly: shill-bid auctions and seller errors that get unwound via refund without delisting.
On auctions: check the bid pattern. A single sale that is 40% above the rest of the comps with a thin bidding history — two or three bidders, one making rapid late bids from a low-feedback account — is a probable shill scenario and should not anchor your price expectations. Treat it as an outlier until you see two or more subsequent sales at that level. On Best Offer listings: accepted offer prices reflect the seller floor, not the clearing market. Use them as a lower-bound data point, but do not set your asking price against them. A pattern of accepted offers across several comparable cards tells you something real; one accepted offer tells you almost nothing. See also: how to spot a shill-bid auction.
Check Whether the Price Point Holds Up Across Similar Cards
A single comp is a data point. Three comps for the same card and grade are a price. The test that separates a real price level from a noise spike is whether comparable cards — same set, similar tier player, same grade — land in a consistent range across multiple independent sales. If you can find four PSA 9 sales for the card you are evaluating within a $10 band, that band is the market. If your one comp is the only data you have, your confidence in that number should be low regardless of how attractive the price looks.
The cross-card check also catches set-specific liquidity problems. A low-population card that trades once every 90 days has unreliable comps by definition — the last sale may predate a player news event that moved the whole category. When population is under 50 in a grade, look at BGS equivalents and recent raw sales for additional signal rather than relying solely on the sparse PSA sold history. Off-season price softness of 10-25% is documented in baseball and basketball for non-superstar cards, so timing matters as much as the comp itself.
Look at the Top Grade Too — Just Not First
Once you have a realistic price anchored at your expected grade, check PSA 10 sales as well — but treat them as ceiling information, not baseline. You want to know the PSA 10 market is liquid before you submit, and you want a concrete upside number for your margin math. A PSA 10 market that is thin or trending down is a signal to be cautious even if the PSA 9 comps look fine.
The grade-premium ratio also tells you something about the card's market structure. A PSA 10 that trades at 3x the PSA 9 price is a card where gem-mint condition commands a real premium — common in vintage, star rookies, and low-pop parallels. A PSA 10 that trades at 1.2x the PSA 9 is a commodity card where grade barely moves the needle. Knowing which situation you are in shapes both your submission decision and your asking price if you are selling. See also: margin math before committing to a card.
Run the Margin Check Last
Comps tell you what the card is worth at various outcomes. They do not tell you whether the deal is good — that is a separate calculation. Once you trust your comps, stack them against your full cost to exit: purchase price, shipping to PSA, grading fee (currently $25 for regular service on cards valued under $499), return shipping, and marketplace fees on the sale (eBay is 12.9% on most cards). A card you buy for $40 and submit at regular service costs roughly $70 all-in before you sell anything. If the PSA 9 comp is $85, your margin is thin. If the PSA 9 comp is $150, the deal works even on a grade miss.
The margin check is also where timing decisions get made. If a card's PSA 10 comps have dropped 20% in the last 60 days and are still trending down, the right move is not to adjust the buy price — it is to question whether the trade makes sense at all right now. Comps are a snapshot, not a guarantee.
A Quick Gut-Check List
Before you act on a price:
- Have I actually evaluated condition honestly, not optimistically?
- Am I comping the grade I realistically expect, not the grade I want?
- Have I checked whether any outlier sale is explainable, or just convenient?
- Do I have more than one comparable sale supporting this price?
- Have I run this against my actual cost stack, not just the purchase price?
AgentGrail's comp engine runs this process automatically against real eBay sold data — surfacing the price range tied to the grade your card actually scans as likely, not the grade the listing photos are trying to sell you. When comps are thin or scattered, that uncertainty shows up as a lower confidence score and a REVIEW flag instead of a clean BUY, so you are not left guessing whether a single outlier sale is real. The market breakdown panel shows PSA 9 and PSA 10 sold comps side by side so you can see the grade-premium spread before you decide whether to submit.
Frequently Asked Questions
What counts as a real comp vs noise?
A real comp is a recent sold listing for the same card and same grade where the bidding history looks like normal buyer behavior — multiple bidders, no unusual late-bid spikes from low-feedback accounts. A noisy comp is an outlier, much higher or lower than similar sales, where something specific happened: a bidding war, a shill, a buyer error, or a moment of FOMO hype. Weight real comps heavily; treat outliers as information about range, not as the price.
How recent do comps need to be?
In a stable market, 90 days is defensible. In a market with active price movement — a player just signed a deal, a rookie just broke out, or a set just hit peak hype — focus on the last 30 days and treat anything older as context rather than current price. If you are seeing a clear trend week over week, weight the most recent three sales most heavily. Stale comps are worse than no comps because they give false confidence.
Why should I check one grade below my expected outcome?
Because grading is probabilistic, not certain. PSA grades a raw submission as a 10 roughly 45% of the time for modern base cards in strong condition, and materially less for vintage or cards with any visible flaw. A card that looks like a 10 to you comes back as a 9 more than half the time. If you have not priced the 9 scenario, you are running the margin math on a best-case assumption. A deal that only works at a 10 is a bet, not a margin play.
How many comps do I need to trust a price?
Three is the minimum for real confidence; five or more is better. One comp can be an anomaly. Two can be a coincidence. Three comparable sold listings in the same price range is a signal worth acting on. If you can only find one comp, treat your confidence in that price as low and factor in a wider margin of safety before committing.
What should I do if comps are scattered widely?
Do not average them and call it a price. Scattered comps mean the market is thin or uncertain — less liquid, less predictable. Apply a larger margin of safety before buying, or recognize you are entering a higher-uncertainty trade where your exit price could land anywhere in a wide range. Scattered comps are not a reason to pass, but they are a reason to not pay full ask and to size the position accordingly.
Do Best Offer accepted prices count as comps?
They count as lower-bound data. Accepted Best Offer prices reflect the seller's floor, not the clearing market price — buyers who use Best Offer are typically pushing below market, and the accepted price reflects that. Use them to understand how low prices go, not to set your asking price. If you are the seller, auction comps and fixed-price sold listings are more reliable benchmarks than accepted offer history.