Why the same card sells for $40 and $90

You find five sold listings for what looks like the same card and they read $41, $52, $55, $58 and $89. The instinct is to treat four of them as real and one as a mistake, or to average them and move on. Both are wrong, and in opposite directions: the spread is the most useful thing on the page.

What actually differs between two “identical” sales

Before blaming randomness, most of a spread is explainable, and each explanation changes what the number means for you:

• Auction versus Buy It Now. An auction discovers what two determined bidders will pay; a fixed price discovers what one buyer would accept. Different measurements of different things.
• Delivered versus item price. A $52 card with $8 shipping cost more than a $55 card shipped free. Every figure here is delivered for that reason.
• Condition and grade. Raw cards vary; slabs of different grades are simply different cards, and a PSA 10 comp does not price your PSA 9.
• Timing. A sale during a hot week for that player is real evidence about that week, and weaker evidence about today.
• Who was selling. A seller with no feedback and bad photos gets less for the same card, and that discount is about them, not about the card.

Which number to trust: the median, not the average

On the five sales above the average is $59 and the median is $55. The average is dragged by the $89 — one buyer having an uncharacteristic day moves it by four dollars, and a single fat-fingered $300 sale would move it by fifty. A median barely notices, which is exactly the property you want from a number that has to survive strangers.

Ours is also recency-weighted: a sale's weight halves every 45 days, so a sale from six weeks ago counts half as much as one from this week, and a sale from three months ago counts a quarter. Nothing is discarded for being old — it just stops outvoting the present.

An outlier is flagged, never deleted

A sale more than 3.5 median-absolute-deviations from the middle is marked as an outlier and kept in the pool. That is a deliberate choice, and the reason is that the alternative is unfalsifiable: a system that quietly drops the sales it dislikes can produce any fair value you want and always looks tidy.

It also matters because outliers are sometimes the story. Three $89 sales in a row are not three mistakes — they are a market moving, and a filter that deleted the first one would have hidden the first evidence of it.

How many sales before a number means anything

This is where most price guides stay quiet and this one does not. The confidence label attached to every fair value here comes from a stated rule, not a feeling:

Two recent sales of the exact card earn MEDIUM because two independent witnessed transactions are genuine evidence. One sale never does, however convenient its number is — and the difference between those two rows is the difference between a market and a coincidence.

The range is the answer; the single number is a summary of it

Every fair value here ships with a range rather than a point, because the point alone overstates what five sales can tell you. The range is the middle half of the evidence — the 25th to 75th percentile — and the floor and ceiling quoted alongside it are the 10th and 90th. A card whose middle half spans $41–$58 is a different proposition from one whose middle half spans $52–$55, even when both medians say $55.

Practically: buy against the low end, sell against the high end, and treat a wide range as a warning that the market has not made up its mind. What you keep after fees is a separate calculation again — the net proceeds calculator does that one.

Your own purchases are the comp to trust least

Cards you bought yourself are real trades, so they count — but a valuation built mostly from them measures your behaviour rather than the market, and it is circular. Overpay once and that price becomes the fair value that makes the next identical card look like a bargain, which is how one bad habit compounds into five. So a pool that is more than half your own buys cannot reach HIGH confidence here, and the page says how many of the sales were yours.

Run it on your own card

None of this is theory you have to take on trust. The sold comp fair value calculator runs the exact function described above — the same one that prices every card on the board — over sales you paste in yourself. It prints the range, the midpoint, the outlier count, and the weight each individual sale carried, so the claim that a five-month-old sale barely moves the answer is something you can watch happen rather than something this page asserts.