Sell Price Optimizer Methodology

What the Tool Answers

For one card and one deadline, the tool answers a single practical question: what should I ask, and what are my odds? You choose an atom (a printing, grader, grade) and a horizon (how many days you can wait, from a few days to a few months). It returns the recommended ask price, the probability of selling by that deadline at that price, and the expected net after fees, the money actually in hand.

Underneath the headline answer is a full grid: the probability of selling at every modelled price and every horizon, so you can see the trade-off rather than just one number. A custom-price probe lets you type any price and read the odds back across each deadline. Everything keys off TCGBerg fair value as the reference point, so prices are expressed both in dollars and relative to fair value.

Probability of Sale

The core of the tool is a survival model: given a price, how likely is the card to sell within a given number of days. Two ingredients drive it. The first is how often the card trades at all, a sale rate estimated from its recent transaction history. The second is how price affects buyer demand, captured as a demand curve in relative price, where relative price is the ask divided by fair value.

Lower relative prices sell faster, but with diminishing returns. The demand curve is anchored so that pricing at the typical realized price gives middling odds, while cutting all the way toward free at most roughly doubles the clearing rate. There is no magic price that guarantees an instant sale, and the model says so: the probability of selling by your deadline rises smoothly as you wait longer or price lower, and is read straight off the fitted curves rather than guessed.

Borrowing Strength for Thin Cards

Most graded cards trade rarely, so the model pools information across a ladder of increasingly broad groups. From finest to coarsest: this exact atom; the same card across its printings at the same grade band; the set at that grade and price tier; the broader category at that grade and price tier; and finally the whole platform.

When an atom has plenty of its own sales, its own data dominates. When it has only a handful, the estimate leans on the coarser levels, borrowing the typical sale rate and price sensitivity of similar cards rather than overreacting to two or three sales. A card that has gone quiet still gets a sensible estimate rather than a collapse to zero. Crucially, the tool tells you which level it leaned on, so you can judge how card-specific the answer is.

Choosing the Recommended Price

The recommended price maximizes expected money in hand by your deadline, not the raw chance of a sale. For each candidate price the tool computes expected net as the after-fee proceeds multiplied by the probability of selling by the deadline, then scans across a dense range of prices and picks the best.

Because selling faster usually means accepting less, the optimum balances the two: priced too high, the odds fall; priced too low, you leave money on the table. Marketplace fees are tiered, so for high-value cards the best price can sit a little higher, where the marginal fee rate drops. When a card sells so readily that asking more would barely slow it down, the tool says the suggestion is capped at the top of the modelled range rather than pretending to find a precise peak.

Fees

Every figure labelled in hand is after marketplace fees. The tool models the marketplace's tiered final-value fee (the default is the eBay US trading-cards schedule) plus the per-order fee, so the recommended price and the expected net already account for what the marketplace takes. Money is computed exactly, with no floating-point drift, and fees are applied at the same rounding the marketplace uses.

Because the fee is tiered rather than flat, the effective rate on a large sale is lower than the headline percentage, which is part of why the optimal ask for an expensive card can land above fair value.

Confidence and Honesty

Every estimate ships with a 0 to 100 confidence score and is presented accordingly. High-confidence atoms show a single point estimate. As confidence falls, the tool shows a credible low-to-high band instead of a falsely precise number, and flags the thinnest cases outright.

An evidence panel makes the basis explicit: how many sales back the estimate (all-time and in the last year) and which pooling level the answer leaned on. The message is plain: based on this exact card and grade when the data is there, or estimated from similar cards when it is not. The tool would rather show an honest range than a confident-looking number it cannot stand behind.

Limitations

A few limits to keep in mind:
  - It is the market's clearing rate, not your personal odds. The probability reflects how fast this card sells across the whole market at a given price. Your own sale also depends on your photos, timing, and the competing listings you are up against.
  - Listing age is not yet modelled. The odds do not currently account for how long a specific copy has already sat unsold.
  - It needs a fair value. Pricing is relative to TCGBerg fair value, which is driven by graded-sale history; a card without a confident fair value cannot be priced here.
  - It is not financial advice. These are model estimates for information only, not a recommendation or a promise that your card will sell.