Scoring Pokémon card lots against eBay sold comps

TL;DR

• Active eBay listings are asking prices, not realized prices—they lie. Use SOLD comps (~90-day history) to anchor your math.
• Card identity is everything: name + set + edition + variant + grade. “Charizard” is not a price; “Base Set Unlimited Charizard, PSA 9” is.
• Sell-through velocity (sales per week) predicts how long your capital sits locked up and is the variable most buyers ignore.
• Expected value = (median sold price × (1 − fee rate)) − shipping − grading. Apply a 30% margin of safety before you bid.
• Confidence gates protect you: skip lots where you can’t find ≥3 grade-matched comps or where the price spread exceeds ~6× the median.

Buying Cards Rewards Discipline

I’ve overpaid for Pokémon lots I “knew” were steals. The problem was never the cards—it was that I was bidding on nostalgia and gut feeling, not math.

A lot listing forces the question: “What’s the highest I can pay and still come out ahead?” Most buyers don’t ask it clearly. They see a binder of holos, think “that’s gotta be worth something,” and bid until their pulse climbs. That’s the winner’s curse—if you win, you were probably the most optimistic person in the room.

The antidote is a defensible max bid: a number anchored to real market data, with a safety buffer baked in. This post walks through building one for cards.

Why Active Listings Lie—Use SOLD Comps

Here’s the trap: I search eBay for a card, see five listings asking $400–$500, and think “Great, my estimate is $450.” Then I overpay because those listings never actually sold.

Active listings are ceilings, not realized prices. Sellers list high hoping for a negotiation or a desperate completist. The realized price—what someone actually paid—lives in SOLD comps only.

eBay’s Marketplace Insights API gives ~90 days of sold data for a given search. If you don’t have API access, third-party comp tools (there are several built specifically for cards) scrape and republish this. Last resort: take active prices and apply a 0.65× haircut as a conservative estimate.

The gap between asking and sold can be brutal. A card listed at $500 might average $350 sold—a 30% spread. Ignore it and your margin evaporates before the card even lands in your mailbox.

Pulling Sold Comps

The mechanics depend on your data source:

• Direct API (eBay Marketplace Insights): query a precise search string, filter by condition, and get a histogram of sold prices over the lookback window.
• Third-party tools: card-specific comp sites are often cheaper than raw API calls for bulk work; query their endpoints and parse JSON.
• Manual (fallback): use eBay’s “Sold items” filter, eyeball the prices, and compute the stats by hand. Tedious but free.

For bulk lot work, I write a small Python script to:

1. Build the distinctive search token from the card (e.g., "Charizard Base Set Unlimited PSA 9").
2. Query comps with that token.
3. Filter by grade or raw condition.
4. Trim outliers (more on that below).
5. Compute median, mean, and velocity.
6. Spit out a JSON blob.

The script takes ~30 seconds per card. Run it before you commit to a bid.

Reading Comps Smartly

Not all comps are equal. Three gotchas matter most for cards:

Singles vs. Bulk Lots

A “lot of 100 holos” has comps for individual cards—but those are single-card sales. Multiply a single-card comp by 100 and you’re lying to yourself; most of a bulk lot is near-worthless commons and bulk rares.

Value the chase cards individually, then bulk the rest. If three cards in the lot carry the value and the other 97 are penny bulk, price the three from real comps and assign the rest a per-card bulk rate. A handful of grade-worthy cards usually carries the whole lot.

Identity & Grade Matching

This is where card valuation gets unforgiving. A search for “Charizard” returns Base Set, Base Set 2, reprints, promos, and a dozen modern sets—wildly different prices. Worse: the same card in different grades is a different asset. A raw Base Set Charizard, a PSA 8, a PSA 9, and a PSA 10 can span an order of magnitude.

Filter comps to the exact identity: name, set, edition (1st Edition / Shadowless / Unlimited), and grade. If your card is a PSA 9 but your comps include raw copies and PSA 10s, your median is fiction. This 30-second discipline saves more money than any other step.

Condition & Authenticity

A listing that says “PSA 10 Gem Mint” is not comped to a raw “NM” copy, and neither is comped to a trimmed or restored card masquerading as clean. Read the photos, not the titles—check centering, edges, and surface. And for anything expensive, comp slabbed copies only; raw high-value cards carry authentication risk that the market prices in.

Outlier Trimming

You’ll find comps at $1 (typos, lot accidents, mis-listed grades) and at 4× median (a gem-mint pop-report rarity or a shill-looking sale). These drag your median in opposite directions.

Trim outliers to ~0.2–4× the median. It’s a heuristic, but it kills the obviously broken data without overthinking.

Sell-Through Velocity—The Hidden Signal

Here’s the insight that changed how I buy: how fast does this card actually sell?

eBay reports sold dates. Pull them. A card with 10 sales per week is liquid; one with 1 sale per month means your capital is locked up 4× longer.

Velocity matters because:

• Working capital drag: if you pay $300 for a card and hold it 60 days, that’s $300 sitting idle—opportunity cost if you could’ve flipped it into three faster cards.
• Risk: longer holds mean more exposure to a set reprint, a market dip, or a hype cycle cooling off.
• Predictability: fast movers are easier to forecast and price.

Active listings carry zero velocity signal—you can’t tell if a card will move in a week or rot for a year. Sold dates are a proxy for category health. A modern chase card with five sales a week has demand; a niche promo with one sale a quarter does not, no matter how “rare” the seller swears it is.

Fees & Costs

Every dollar that leaves your hand doesn’t come back whole.

eBay fees are the big line item. The standard final value fee is ~13.6% of the sale price (look up the current rate—it moves). Subtract it upfront.

Shipping is small for raw singles—a penny sleeve, a toploader, and a plain envelope—but it climbs fast for graded slabs that need bubble mailers, rigid boxes, and signature/insurance on anything pricey. Budget it honestly.

Grading is the cards-specific cost that wrecks naive math. If your plan is buy-raw-grade-and-flip, you’re paying a submission fee plus eating weeks or months of turnaround where your money is locked inside a grading queue. And you’re betting on the grade: a card that comes back a 9 instead of the 10 you hoped for can erase the entire spread. Price the realistic grade, not the dream grade.

Other costs: your time to research, photograph, list, and handle returns; payment processing; the occasional refund.

Expected Value + Margin of Safety

Here’s the formula I use:

EV = (median_sold_price × (1 − fee_rate)) − shipping − grading

That’s your realistic expected payout if you win the card and sell it under typical conditions.

Now, margins matter. I only bid if I’m keeping at least 30% gross margin (profit at 30% of the sale price). Why 30%? It’s a buffer for:

• Comps being off (the market moved, or the grade mix was wrong).
• A grade coming back lower than hoped.
• A buyer return, an INAD claim, or a chargeback.
• Cards that don’t sell and need relisting.

I use three verdict bands:

These are my personal thresholds; yours might differ based on risk tolerance and how much capital you’ve got on the sidelines.

The Max-Bid Formula

If you’re buying on eBay directly, there’s no buyer’s premium—your bid is your price. But the moment you bid through a dedicated card auction house (the big consignment platforms), a buyer’s premium gets added on top of your hammer price—often ~20%. Forget to reverse it out and you’re 20% over your margin plan.

Max bid formula:

max_bid = target_value ÷ (1 + premium_rate)

If your target value (the most you can pay after fees, shipping, and grading and still keep your margin) is $1000, and the house premium is 20%:

max_bid = 1000 ÷ 1.20 ≈ $833

Bid $833, pay $1000 all-in including the premium, and you’re still at your margin target.

Confidence Gates—When NOT to Bid

Math is only as good as its inputs.

I lose money when I bid on thin data: only two comps, or a 10× spread between the highest and lowest sale. The data’s too noisy and the market’s too fragmented to trust.

Confidence gates I use:

1. Minimum comps: ≥3 real, grade-matched SOLD listings. If I can’t find three, I skip or downgrade from BUY to WATCH.
2. Price spread: max ÷ min should be under ~6×. If the cheapest comp is $50 and the priciest is $500, something’s off—mismatched grades, condition variance, or outliers. Retrim or skip.
3. Velocity minimum: if there are only 1–2 sales in 90 days, demand is thin. I might still bid, but I lower my margin target to compensate.

These gates saved me from a “can’t-miss” lot of vintage promos that turned out to have zero buyers below my cost.

What the Data Taught Me

Over a year of buying, patterns emerged:

• Hype cards bleed value faster than I expected. The asking-to-sold gap on freshly-hyped modern cards is brutal, and velocity drops the second the hype cools.
• Bulk lots are mostly a trap. One or two chase cards carry the lot; the rest is bulk you’ll be sleeving for hours. Comps assume you can sell the headliners, not the 95 commons.
• Graded vintage has legs. Slabbed, era-defining cards hold value better than raw modern hits—fewer comps, but tighter spreads and steadier demand.
• Timing matters. Cards that move in 1–2 weeks are margin-positive; the ones that sit for months quietly turn into dead inventory.
• Grading is a bet, not a guarantee. I’ve paid PSA-10 prices on a raw card that came back a 9. Now I price the conservative grade and treat the bump as upside, not the plan.

The meta-lesson: action beats perfection. I started with a messy spreadsheet, upgraded to a Python script, then folded it into a proper data tool. Each version was ~20% better than the last. I didn’t wait for the perfect system—I improved it off real losses.

Wrapping Up

Card lots are won by people with better information, not just deeper pockets. A clear data model—SOLD comps, grade matching, velocity, fees, margin of safety—shifts the odds in your favor.

The tool doesn’t have to be fancy. A Python script, a spreadsheet, or disciplined manual research all work. What matters is asking the hard question before you bid: “Can I make money at this price, and by how much?”

Once you have a number you trust, bid with confidence. The discipline pays off.

Before you trust a comp, trust your eyes: a decent loupe catches off-center cuts, print lines, and trimmed edges that photos hide—and a stack of sleeves and toploaders keeps the cards you win from losing a grade in transit.