How eBay Sold Comps Really Work in 2026

Published April 20, 2026 · Updated April 25, 2026 by HobbyCardIndex. Reports are updated as the market changes.

Why we are writing this

eBay is the single largest public comp source in trading cards. For most modern and post-war cards, it is the only venue deep enough to produce a real market price inside a 90-day window. Every pricing tool the hobby relies on, including HobbyCardIndex, pulls the bulk of its trade data from eBay sold comps.

And yet eBay sold comps are misread constantly. A listing without a price gets discounted, when the private-listing flag is the actual reason the number is hidden. A Goldin consignment sale on eBay gets averaged into a grid of private-seller comps, dragging the median up by hundreds of dollars on a mid-tier card. A Vault-to-Vault transaction gets compared to a ship-from-home listing as if they were the same instrument. A 90-day-ago sale is treated as current, when the last month of data actually shows a different trend.

The goal of this report is to explain how the eBay sold filter actually works, to name every category of sale that is hidden from public view, and to give a reader the specific rules we use at HCI to read a single comp correctly. Where we reference a price, it is a general reference band drawn from the public-tier market, not an HCI internal valuation. See our independence page for how we work with data. If you are reading eBay sold comps to decide whether to grade a card before listing it, the comp work upstream of that choice belongs in our should-I-grade-this-card decision framework, where the eBay-sold-grid math feeds the raw-vs-graded EV comparison directly.

What the eBay sold filter actually shows

The eBay sold filter is a search layer on top of completed listings. It shows listings where the sale status resolved to sold (auction cleared, Buy It Now purchased, or Best Offer accepted) within the trailing retention window. The public retention window is approximately 90 days. eBay does not publish the exact cutoff, and anecdotally it varies by category and by account status, but 90 days is the operational number the hobby has built its reading habits around.

Inside that window, a sold comp row carries four pieces of data: the title the seller used, the final sold price (sometimes), the end date, and the seller handle. It also carries the listing photos and the seller feedback count. Everything else, including how long the listing sat, how many watchers it had, how the sale closed (auction vs Buy It Now vs Best Offer), and whether the price included shipping, requires clicking into the listing to read.

The search filter itself is a title-keyword match. It does not understand card grade, set, or parallel unless those words appear in the title. This is the single largest source of bad comp reads: the tool looks for a string, not a card identity.

The 90-day window and how to work around it

For casual price checks, 90 days is enough. For anything that trades at a quarterly or slower cadence, 90 days is not enough. A 1986 Fleer Jordan PSA 8 might see five to ten public sales a month. A 1971 Topps Roberto Clemente PSA 7 might see three in six months. A 1990s parallel insert in PSA 10 with a pop of 18 might see one sale a year.

There are three practical ways to look at sales history older than 90 days. One is to use a third-party pricing tool that archives the window rolling-forward. HCI is one of these. CardLadder is another, with different methodology (see our HCI vs CardLadder comparison). The trade-off is that you are now reading a derived dataset, not raw sold comps. Every cleaning rule the tool applies is a decision you inherit.

Two is to look at auction-house archives (Goldin, Heritage, PWCC, Memory Lane), which keep their own records indefinitely and publish hammer prices publicly. Those archives are cleaner than eBay for flagship cards because the venue is curated and the provenance is clear, but they cover a tiny slice of the card universe, mostly pre-war and high-end modern.

Three is to keep your own scrapbook for any card you actively track. This is what serious collectors did before pricing tools existed. For a single flagship card you own, a notebook of sold comps with date and grade and seller is often more accurate than any aggregator, because you can apply your own filters.

Five categories of sales that are hidden from public view

The core reason eBay sold comps are misread is that several entire categories of sales either do not show a price or do not show up at all. Anyone pricing a card needs to know which ones.

1. Best Offer accepted

When a seller lists a card with Best Offer enabled and accepts an offer below the list price, the sold-status resolves but the final price is hidden from public view by default. The listing shows as sold with a struck-through list price and no accepted amount. Some sellers disable the hide-final-price setting, but most do not.

How much volume sits here depends on the category. On mid-tier modern cards listed at a round number ($250, $500, $1000), Best Offer accepted is commonly 40% to 60% of sold volume. On raw cards listed above $100, it is often the majority of sales. The final-price suppression means the public sold grid is systematically biased upward, because the visible sales are the full-price Buy It Nows and the accepted-list-price auctions, while the discounted Best Offer transactions are invisible.

If you are pricing a card that lists with Best Offer accepted and you see ten sold comps, assume 30% to 50% of the true sales volume is hidden and cleared lower than the visible list prices. The median you see is probably 10% to 20% high.

2. Private listings

A seller can set a listing to private, which hides both the bidder identifiers and, on Best Offer acceptances, the final price. Private listings are common on high-value cards for two reasons: the seller does not want the buyer identified (consignment relationships, dealer networks), and the seller does not want the price visible for tax or competitive reasons.

On cards above roughly $5,000, private listings are a meaningful share of total volume. A flagship 1952 Topps Mantle or a PSA 10 1986 Fleer Jordan will have more private sales than public. The public-only sold grid for those cards is a subset of the true trade, and it tends to over-represent the edge cases (the eye-catching high prices that get posted publicly for marketing) and under-represent the bread-and-butter dealer-to-dealer trades.

3. Pulled and canceled listings

A listing that sells and is then pulled or canceled (for fraud, payment failure, claim dispute, shipping issue, or seller change of mind) disappears from public sold view. If the cancellation happens within a few days of the sale, the listing often never shows up in the 90-day comp grid at all. Bigger than expected, because shipping disputes on high-value cards are common, and a nontrivial slice of auction sales get pulled for buyer-remorse reasons.

The practical effect is that the sold grid skews a bit toward uncontested transactions. If a card has a known problem (soft centering that grades inconsistently, fake autographs in circulation, a known counterfeit run), the pulled rate is higher and the public comp reads are thinner.

4. eBay Vault transactions

eBay launched the Vault in 2022 as a physical storage and authentication layer for high-value cards. A card in the Vault can be sold Vault-to-Vault, which means the card does not physically move during the transaction, no shipping is charged, and in many states sales tax is handled differently than a standard listing.

Vault-to-Vault sales are visible in the sold grid but are flagged with a Vault badge on the listing. They tend to close faster, clear at 2% to 8% higher than a shipped equivalent on flagship cards (buyers pay up for frictionless delivery and authentication), and carry different fee structures for sellers.

If you are reading Vault comps for a card that you personally would have to ship, the Vault comps are directionally correct but not strictly comparable. They are a premium tier. If you are reading comps for a card you would list in the Vault yourself, Vault comps are the correct reference set.

5. Auction-house consignment sales on eBay

Goldin, PWCC, Heritage, Memory Lane, and other auction houses list some of their inventory on eBay. Those listings clear through the auction-house consignment process, which means the seller (the original owner) paid a consignment fee (typically 10% to 20% depending on the house and the card value), the buyer paid a buyer premium (typically 15% to 20%), and the final eBay-visible price is the hammer price plus any eBay-specific fees.

The effect is that auction-house comps clear 15% to 30% above private-seller comps for the same card in the same window. This is a fee structure, not a price signal. A card that sold for $10,000 on Goldin on eBay is not the same as a card that sold for $10,000 from a private seller on eBay, because the Goldin buyer paid $11,500 to $12,000 all-in while the private-seller buyer paid $10,000 plus shipping and tax.

If you are pricing a card you would list yourself as a private seller, exclude auction-house comps or mentally adjust them down by 15% to 25%. If you are deciding whether to consign through an auction house, auction-house comps are the correct reference set.

Shipping, tax, and fees: three ways comps get misread

The price displayed in the sold grid is the listing price paid by the buyer. It does not include shipping in most cases, does not include sales tax, and does not reflect the seller's net after fees. All three distort the read.

Shipping on raw cards and low-value slabs is typically $4 to $8 via USPS First Class Package. On high-value slabs it is $15 to $50 for insured Priority or Registered Mail. Some sellers build shipping into the listing price (free shipping). Some charge it separately. The sold grid price is the price paid, so a $60 free-shipping listing and a $50 plus $10 shipping listing look different even though the buyer paid the same amount.

Sales tax varies by state. A buyer in Washington pays about 10% total tax on eBay sales. A buyer in Montana pays 0%. Tax is not shown in the sold grid. For a card purchased to resell or to be flipped, tax on the purchase and tax on the eventual sale compound into the break-even math. For a card purchased to keep, tax is a direct all-in cost.

Fees to the seller run around 13% to 15% on eBay for trading cards in 2026, plus payment processing. A private seller who lists a card at $1,000 and sells at $1,000 nets roughly $850 to $870 before shipping. The sold-grid price is gross, not net. If you are reading comps to decide what to list your own card at, the seller-net math matters. If you are reading comps to decide what to buy a card for, the seller-gross price is the actual market clearing price. See our selling cards on eBay guide for the full fee breakdown.

How auction format differs from Buy It Now

Auction and Buy It Now are structurally different price signals. They should not be averaged without thinking about what they represent.

An auction clears at whatever two bidders were willing to pay in the final minutes. On a liquid card (high view count, multiple serious bidders) that price is a market consensus. On a thin card (low view count, one serious bidder) the price is whatever the winning bidder thought was fair minus a few increments. Thin-card auctions produce the widest distribution of outcomes in the hobby. The same card can clear 30% below median one week and 20% above median two weeks later.

Buy It Now is a seller-set price. If it sold at the listed price, two things are true: the seller thought the price was fair, and at least one buyer thought the price was fair. A Buy It Now that actually sold is a cleaner comp than an auction on a thin card, because both sides of the transaction reached the same price. But Buy It Now comes with survivorship bias: the listings that sat for weeks without selling are not in the sold grid, so the visible prices are skewed toward the ones the market validated.

Practical rule: on liquid cards with 20+ sales in a 90-day window, auction and Buy It Now comps converge and can be averaged. On thin cards with 1 to 5 sales, read each comp individually, note the format, and weight accordingly. Our how to value a card guide walks through the weighting process with worked examples.

How listing titles ruin comp searches

The eBay sold filter is a keyword search, not a card-identity search. This is the single largest operational problem with raw sold-comp reads.

Consider a 2020 Topps Chrome Luis Robert Chrome Refractor base RC in PSA 10. A clean seller might title the listing "2020 Topps Chrome Luis Robert #200 Refractor RC PSA 10". A lazy seller might title the listing "Luis Robert Refractor Gem Mint 10". A confused seller might title the listing "Luis Robert Chrome RC 2020 HOT PSA10 LOW POP". A wrong seller might title the listing "2019 Topps Chrome Luis Robert Refractor PSA 10", which is the wrong year.

A keyword search for "2020 Topps Chrome Luis Robert Refractor PSA 10" will match the first title and miss the other three. Miss-rates on title-only searches commonly run 15% to 40% on modern cards. On vintage cards with looser naming conventions, miss rates can be higher.

Three cleaning rules the hobby uses to tighten title searches:

• Search by card number, not just player name. A card number is a hard identifier. If the title includes "#200" or "/200" or "card 200", the match is exact. If it does not, the listing is a candidate for manual review.
• Search for the grade as a separate pass. Run one search for "PSA 10", one for "PSA10", one for "Gem Mint 10", and deduplicate. Lazy sellers use the shortened forms. The grade word in the title is sometimes absent even when the card is slabbed, because the photo shows the slab and the seller assumes the buyer will read the photo.
• Exclude the wrong parallel with negative keywords. If the card is a base Refractor, exclude "Gold", "Red", "Orange", "Atomic", "Prism", "X-Fractor", "Purple", "Speckle", "Aqua", "Silver", and other parallel names. Missing this step is how a $75 base Refractor gets averaged with a $450 Gold /50 Refractor.

For more on parallel-specific reading, see our guides on what is a parallel and what is a refractor.

Outlier trimming: three rules

Every aggregator trims outliers. There is no industry standard, which means every tool produces a slightly different median. The rules HCI uses are simple and documented, and most professional readers apply some version of the same three.

Rule one: trim the top and bottom 10% before computing a median. On 20 sales, this drops the top 2 and the bottom 2. The remaining 16 are where the middle of the market sits. The 2 that got trimmed are not noise in every case, but treating them as noise is a more honest default than treating them as signal.

Rule two: exclude sales more than 30 days old from a current-price read. The full 90-day window is useful for depth. For a real current comp, the last 30 days is the honest read, because cards move on a faster cycle than 90 days during trend changes. Many cards compressed 20% to 40% inside 90-day windows during 2022.

Rule three: flag the sale seller and separate out auction-house and Vault sales. Median the private-seller comps on their own. Note the auction-house comps as a premium reference. Note the Vault comps as a Vault-to-Vault reference. Do not blend them.

Applied together, these rules turn a messy sold-comp grid into a clean reference price. They are not sophisticated. They are the baseline.

Reading a single comp vs a distribution

A single eBay sold comp is a data point. A distribution of 20 comps is a market. The difference matters because a single comp can be anything.

Consider two people who sell the same PSA 9 modern rookie one week apart. Seller A lists at $275 with two good photos and a clean title and sells to a local collector after two days. Seller B lists at $500 with Best Offer enabled, accepts $210 three weeks later. Both sales are real. The sold grid will show Seller A at $275 visible, and Seller B as struck-through $500 with no accepted amount shown. If you read only Seller A's comp, you will think the card is a $275 card. If you know Best Offer hides the true clearing price at around $210, you will think the card is more like a $240 card at the median.

The single-comp read produces confident wrong answers. The distribution read produces less confident right answers. On a card with 20+ sales in 90 days, the distribution is always the correct read. On a card with 3 sales in 90 days, the right answer is "the market is thin, my price has a $50 uncertainty band, and I should price inside that band, not at a single point."

How to look up eBay sold listings the right way in 2026

The single phrase most readers type into Google is "ebay sold", usually meant as "show me what this card actually sold for, not what someone is asking." The eBay sold filter answers that question, but only if you know how to query it. Three operational habits separate a clean eBay sold lookup from a noisy one.

First, always start the search inside the Trading Cards category, not from the homepage search bar. A category-scoped query restricts results to the trading-card taxonomy and silently excludes dolls, action figures, and the long tail of mistyped listings that share a player name with a card. Path: eBay home, Sports Mem, Cards & Fan Shop, Sports Trading Cards, then "Sold Items" filter on the left rail (or "Completed Items" filter for the broader view including unsold listings). On mobile, the toggle sits inside the filter sheet under "Show only".

Second, append "Sold listings" or "Sold Items" with the explicit URL parameter when you bookmark a comp search. The URL parameter is LH_Sold=1&LH_Complete=1. A bookmarked sold-comp URL with those parameters baked in pulls the same filtered view every time, which matters if you track a card across weeks. The eBay UI sometimes resets the sold-items toggle on tab reload; the URL parameter is sticky.

Third, use eBay sold filters together, not one at a time. Combining "Sold Items" with the price-range slider, the seller-location filter, and the condition filter (Graded vs Ungraded) collapses a 200-listing comp grid into 20 rows that actually match the card identity you are pricing. Adding a numeric grade to the keyword string ("PSA 10" with quotes) further narrows. Most readers stop at the sold-items toggle; the cleaner reads come from stacking three or four filters together.

Two patterns to avoid: using the eBay app's "Sold" sort without the category constraint (the app's default scope is too broad), and trusting the eBay-sold preview that appears in the Google rich-result snippet (Google's snippet is built from eBay's structured data and frequently lags the actual sold grid by hours to days). Click through to the live sold filter every time.

Eight rules for reading a 2026 eBay sold comp

Everything above collapses into a practical checklist.

1. Date-stamp every comp you cite. A comp from 87 days ago and a comp from 2 days ago are not the same comp. Prices drift even on flagship cards.
2. Check the slab generation. A PSA slab from 2018 and a PSA slab from 2024 can trade differently for the same card and same grade. Pre-2021 slabs sometimes carry a small premium on flagships. Post-2021 slabs with new labels are the default for modern cards.
3. Identify the sale format. Auction, Buy It Now, or Best Offer. Note Best Offer acceptances separately, because the visible price is the list price, not the clearing price.
4. Identify the seller class. Private seller, auction house (Goldin, PWCC, Heritage, Memory Lane), dealer, or Vault. Auction-house comps carry fee-inclusive prices. Vault comps carry frictionless-delivery premiums.
5. Match the parallel exactly. Base Refractor is not Gold Refractor. Silver Prizm is not Green Prizm. One numbered parallel can trade 5x to 20x another parallel of the same card.
6. Match the grade exactly. PSA 9 is not PSA 10. BGS 9.5 is not BGS 10. SGC 9 is not SGC 10. Ignoring this collapses the grade-ladder premium and produces a garbage median. Our what is a PSA 10 guide walks through grade math.
7. Trim outliers before computing a median. Top and bottom 10%. If the distribution is tight, the trim does not move the number much. If the trim moves the median by 20%, the market is thin and the median is weak.
8. Note the volume. Twenty sales in 90 days is a market. Three sales in 90 days is an estimate. One sale in 90 days is a guess. A comp with volume is always worth more than a comp without.

Two worked examples

Example one: a liquid modern card

Card: 2020 Topps Chrome Luis Robert base Refractor RC #200, PSA 10. A liquid card, 2020 rookie year at the pandemic peak, continues to trade regularly in 2026.

What the sold grid shows in a 90-day window: approximately 35 to 50 public sales depending on the week. Median visible price somewhere in the $60 to $80 band in mid-2026, down materially from the 2021 peak of $200 to $250. Dated comps from the trailing 30 days cluster tighter, around $65 to $75.

Cleaning: exclude Gold Refractor (/50) and Orange Refractor (/25) with negative keywords. Exclude 2019 Topps Chrome with the "2020" title requirement. Flag Best Offer accepted listings and assume those cleared 10% to 15% below the visible list price. Trim top and bottom 10%. Separate Vault comps (typically 3% to 6% higher).

Resulting read: a current-market comp in the $60 to $75 band with roughly $5 to $10 of uncertainty based on slab condition and venue. For a card with this volume, the median is reliable and the comp set is a real market.

Example two: a thin vintage card

Card: 1968 Topps Nolan Ryan rookie #177 (shared with Jerry Koosman), PSA 7. A flagship rookie with ongoing demand, but PSA 7 is a middle grade with moderate supply.

What the sold grid shows in a 90-day window: approximately 6 to 12 public sales, with a spread from $1,800 to $3,200. Some sales are through major auction houses on eBay (Goldin, PWCC) and carry buyer premium baked into the visible price. Some are private-seller Buy It Nows that sat on the market for weeks. One or two are Best Offer accepted with no visible final price.

Cleaning: separate Goldin and PWCC comps from private-seller comps. The house sales cluster $2,800 to $3,200 and include buyer premium. The private-seller sales cluster $1,900 to $2,400. The Best Offer listings probably cleared in the $1,800 to $2,100 band, invisible. Trim outliers at the top and bottom. Weight recent sales heavier.

Resulting read: for a card you would list as a private seller, the honest comp is the $1,900 to $2,400 band, with the 30-day median around $2,100. For a card you would consign through Goldin or PWCC, the gross comp is $2,800 to $3,200 but the net to you after consignment fees (typically 10% to 15%) is similar to a private-seller net. Volume is too thin for a single-point answer. See our BGS grading guide, SGC grading guide, and PSA grading guide for cross-grader context on vintage slabs.

How HCI builds comp sets

HobbyCardIndex archives eBay sold comps as they roll through the 90-day window and stores them in a structured catalog keyed to card identity (player, set, year, card number, parallel, grader, grade). We apply the three cleaning rules above, plus slab-generation awareness, plus venue separation (Vault, auction house, private seller). We do not publish the full raw daily time series on public pages, because that is a paid data surface (see our independence page and the paywall policy section on llms-full.txt).

What we do publish on public card pages and hub pages is the reference band for the card: a recent-window median, a grade-ladder snapshot (PSA 9 vs PSA 10 where data supports it), and pop-aware notes for thin cards. That public view is built from public-tier data only and is meant to be a starting point, not a replacement for a reader's own comp reading. For the full story on where HCI's public surfaces end and the paid dashboards begin, see the HCI vs CardLadder and HCI vs TCDB comparisons.

Why this matters for collectors

A collector who reads comps well has a real edge. On thin cards, the difference between a competent comp read and a lazy one is 10% to 30% of the price. On liquid cards, it is 3% to 8%. Compounded across a collection, that gap is the difference between a year of market-parity results and a year where you systematically bought high and sold low because the comp grid misled you.

The hobby has more pricing tools in 2026 than it has ever had. CardLadder, Cardbase, Mavin, 130point, PriceCharting, and HCI all exist, and they all read the same underlying eBay data with different cleaning rules. Read more than one tool, read the raw sold grid yourself on flagship trades, and build your own scrapbook for any card you actively hold. For a side-by-side on the tools, see the HCI vs 130point, HCI vs Mavin, HCI vs Cardbase, and HCI vs PriceCharting comparisons.

Five action items

1. Pick one card in your collection worth more than $500. Pull the last 90 days of eBay sold comps yourself. Count Best Offer accepted listings (no visible price). Count auction-house listings. Count Vault listings. Compute the private-seller median after trimming top and bottom 10%. Compare to what a pricing tool shows. Note the gap.
2. For any card you are about to list, look at the 30-day median of private-seller comps, not the 90-day average. The 30-day median is the current market. The 90-day average is a lagging indicator.
3. For any card you are about to buy, subtract 8% to 12% from the visible median to approximate the private-seller clearing price after accounting for hidden Best Offer acceptances.
4. Keep a one-row-per-sale scrapbook for any card you hold longer than six months. Date, grade, price, seller class, sale format. The scrapbook becomes more accurate than any tool over time because you control the cleaning rules.
5. Read comps with the assumption that every hidden-price listing cleared 10% to 15% below the list price. The visible sold grid is systematically biased upward. Adjust for it.

What we track

HCI tracks eBay sold comps using public-tier metadata (card identity, sale date, sale price where visible, grade, parallel, venue, seller class) and publishes derived reference bands on public card pages. We do not publish raw daily time series, per-user data, alert thresholds, or any paid-tier analytic output on public pages. See our independence page for the full data posture and the state of PSA 10 premiums report and grading cost comparison for the kinds of analysis we do run on top of clean comp data. If you want the full map of what HCI is versus what it is not, see our baseball card hub, basketball card hub, football card hub, hockey card hub, and Pokemon card hub for sport-by-sport context, and our 2020s hub and 1990s hub for era-level framing.