Valdar / Free tools / Sports Card Value Calculator
Sports card value calculator: what recent sold prices actually say
Paste the sold prices you found and this gives you the median, flags the sales that do not belong in the sample, and prints a confidence range that is honest about how few comps you have. It is the calculation a price guide does for you and never shows you.
10 comps used · 1 outlier removed · confidence: good
| Price | Date | z | Verdict |
|---|---|---|---|
| $38.00 | — | -0.82 | in sample |
| $41.00 | — | -0.55 | in sample |
| $42.00 | — | -0.45 | in sample |
| $44.00 | — | -0.27 | in sample |
| $45.00 | — | -0.18 | in sample |
| $47.00 | — | 0.00 | in sample |
| $49.00 | — | 0.18 | in sample |
| $52.00 | — | 0.45 | in sample |
| $55.00 | — | 0.73 | in sample |
| $61.00 | — | 1.27 | in sample |
| $940 | — | 81.25 | not this card |
Interval coverage 97.9%, from the order statistics of this sample.
What “worth” means when you say it out loud
A card does not have a value. It has a distribution of prices at which copies of it have recently changed hands, and the useful summary of that distribution is its middle plus an honest statement of how wide it is. Everything that goes wrong in this hobby’s pricing conversations goes wrong because somebody collapsed that into one number and then dropped the uncertainty.
So the output of this tool is three things, and all three matter. The median is the middle price: half the sales were above it and half below. The range is a confidence interval for that middle, computed from your own sample without assuming anything about the shape of the price distribution. And the spread — the width of the middle half, divided by the median — is the one-glance measure of whether this card has a price at all or merely a set of transactions.
The $940 problem
The tool opens with eleven real-shaped numbers: ten sales between $38 and $61, and one at $940. That eleventh sale is not a mistake in the data; it is the thing that always happens. It is the PSA 10 that matched your text search, or the auto parallel with the same player and year, or a lot of forty cards priced as one line.
Averaged in, it turns a $46 card into a $129 card — a number no copy has ever sold for, above every single observation except one. Remove it and the average is $47.40, which is close to the median of $46. The median of all eleven sales, outlier included, is $47 — one dollar away. That is what robust means, and that convergence is itself the signal: when the mean and the median agree, the sample is clean, and when they diverge by a factor, something in it does not belong.
The tool does not ask you to spot it. It computes each price’s distance from the median in units of the sample’s own median absolute deviation, and flags anything beyond 3.5 of them. That test has no opinion about card prices; it is the standard robust outlier rule, and it survives a sample where a third of the observations are contaminated.
| Contaminant | Plain average | Average moves by | Median of everything | Median after the test | Test verdict |
|---|---|---|---|---|---|
| none | $47.40 | — | $46.00 | $46.00 | nothing to flag |
| $120 | $54.00 | 1.1× | $47.00 | $46.00 | flagged, removed |
| $250 | $65.82 | 1.4× | $47.00 | $46.00 | flagged, removed |
| $500 | $88.55 | 1.9× | $47.00 | $46.00 | flagged, removed |
| $940 | $129 | 2.7× | $47.00 | $46.00 | flagged, removed |
| $2,400 | $261 | 5.5× | $47.00 | $46.00 | flagged, removed |
Each row is a separate run of the calculator on the ten clean sales plus one more. The average climbs to $261 and the median never moves off $46.00, which is the whole reason the headline on this page is a median.
Where to get the sold prices
The tool is only as good as what you feed it, and there is a hierarchy. Completed sold listings on a major marketplace are the best free source, because they are transactions rather than hopes. Auction house results are the best source for anything above four figures, because the buyer pool there is the market for that card. Asking prices are not data at all — an unsold listing at $400 tells you only that one person would like $400.
Filter hard before you copy. Same player, same year, same set, same parallel, same grade or all raw, and nothing that is part of a multi-card lot. It is worth being ruthless here: a sample of six clean, genuinely comparable sales beats a sample of thirty that includes three parallels and two lots, because the outlier test removes the obvious contaminants and cannot remove the plausible ones.
Then check the dates. Sold listings on most platforms only go back ninety days, which is a mercy — it means the sample is automatically recent. Auction archives go back years, and a 2021 sale of a modern card is a historical curiosity rather than a comp. If your sample spans more than a few months, put the dates in and set a half-life.
Reading the confidence range
Most pricing tools print a number to two decimal places and imply a precision the data cannot support. The range this one prints is computed from the order statistics of your own sample: for a sample of ten, the interval runs between the second-lowest and second-highest values and covers the true median about 98% of the time; for a sample of six it is the whole sample; and below six there is no interval worth printing, which is what the tool says.
That behaviour is the useful part. A card with twenty tight comps produces a narrow range and you can price it with confidence. A card with five scattered comps produces a range as wide as the sample and the correct conclusion is that the card does not have a price yet — it has a negotiating position. Both answers are worth having, and only one of them is available from a price guide.
| Comps | What the tool reports | What you can say out loud |
|---|---|---|
| 1–2 | Median only, confidence “none” | “One copy sold for X.” Nothing about the card. |
| 3–5 | Median, quartiles, confidence “low” | “Recent sales were around X.” A working assumption. |
| 6–9 | A real interval appears | “The middle of the market is X, plus or minus.” |
| 10–19 | Interval narrows, outliers testable | “This card is worth about X.” Defensible in a negotiation. |
| 20+ | Spread becomes meaningful on its own | “X, and the market is tight/loose.” The most anyone can honestly claim. |
The confidence labels come from the sample size and from the width of the middle half relative to the median: above 60% the tool reports low confidence however many comps you have, because a card whose middle half spans that much does not have a settled price.
Underneath that table is a second quantity worth seeing, because it is the one that improves as you collect more comps. The interval always covers the true middle at least 95% of the time; what changes with the sample is how much of the sample it has to span to do it.
What the number is not
- It is not what you will get. The median is the middle of what copies sold for, before the platform’s cut and before postage. Put it through the fee calculator to find out what lands in your account, which is typically 12–15% lower and sometimes much more.
- It is not what a shop will pay. A dealer buying for stock pays a fraction of comp, because the discount is their margin, their capital and their risk. The route comparison prices that honestly against the alternatives.
- It is not a grade-agnostic figure. Run it separately for raw and for each grade. A single number across grades is the most common way a card gets mispriced by a factor.
- It is not a forecast. Recency weighting tells you where the market has been going; it does not tell you where it goes next, and nothing does.
Bottom line
To value a sports card, collect completed sold prices for the same card in the same condition, take the median rather than the average, remove the sales that are not of your card, and state a range rather than a point. Ten clean comps gives a usable answer; fewer than six gives a direction; one gives an anecdote. In the worked example on this page a single contaminated sale moves the average from $47.40 to $129 and leaves the median at $46 — which is why every number on this page is built around the middle of the sample and not its mean.
Questions this tool gets asked
How do I find what my sports card is worth?
Find completed, sold listings for the same card in the same condition — not asking prices, and not listings that are still live — then take the middle of that sample rather than the average. Ten sold prices is enough to be useful, six is the point below which no honest range exists, and one is an anecdote. Paste them into this tool and it will give you the median, flag the ones that do not belong, and tell you how wide the real uncertainty is.
Why the median and not the average?
Because card comps are a heavy-tailed sample with two predictable contaminants: a graded or autographed copy that matched the text of your search, and a lot listing where one price covers several cards. Either one drags an average a long way. In the worked example on this page, ten sales between $38 and $61 plus one $940 outlier average to $129 — a number no copy of that card has ever changed hands for — while the median moves by a dollar.
How many sold comps do I need?
Three is the minimum for any central value at all, six is where a distribution-free confidence interval becomes possible, and about ten is where the interval starts to narrow usefully. Below three the tool will still print a number and will label the confidence as none, which is the honest description of a sample that small.
What counts as an outlier here?
The tool uses the modified z-score: the distance of each price from the median, divided by the median absolute deviation of the sample and scaled by 0.6745, flagged when the result exceeds 3.5. That is the standard robust test and it assumes nothing about the shape of the price distribution, which is just as well because card prices are not bell-shaped. Two guards sit on top of it. When more than half the sample is the same price the median absolute deviation is zero, and the tool falls back to the interquartile fences rather than dividing by nothing. And no price within 25% of the median is ever flagged, however the robust test reads: an outlier here means a different card, and a different card is a multiple away, not a fifth away.
Should I include graded sales with raw ones?
No. A PSA 10 and a raw copy are different products with different buyers, and mixing them produces a number that describes neither. Run the tool once on the raw sales and once per grade, then compare the results with the grade premium calculator, which is the tool built for exactly that comparison.
What does the confidence range mean?
It is a distribution-free confidence interval for the median, read off the order statistics of your own sample. In plain terms: if you repeated the exercise with a different random handful of sales of the same card, the middle of that sample would fall inside this range about 95 times in 100. It makes no assumption about the shape of the price distribution, which is the point — card prices are not bell-shaped.
Why does recency weighting change the answer so much?
Because a card that moved 30% in six months is not described by a six-month median. If you supply dates, the tool weights each sale by an exponential decay with a half-life you choose, so a sale from one half-life ago counts half as much as one from today. On a stable card this barely moves the number; where it moves the number a lot, that is the tool telling you the market is trending and the plain median is stale.
Does my data leave the page?
No. The prices you paste are parsed and analysed by JavaScript that arrived with the page. There is no request to any server while you type, nothing is stored in the page or in your browser, and closing the tab disposes of it.
The card is in your hand. Point the camera at it.
These calculators work on numbers you already have. Valdar gets you the numbers: it identifies the card from one photo, pulls what that exact card has actually sold for, and estimates the four grading sub-dimensions before you pay a submission fee. Free to try on iPhone and Android.