Estimated Market Value (EMV) is our estimate of what the next copy of a card, at that grade and printing, would sell for at auction today.
We start from individual completed sales, not averages: graded sales on eBay, Fanatics Collect, Goldin and Alt, matched to the exact card, printing, grader and grade, with duplicates merged and best-offer sales counted at the price actually paid. Lots, reprints, raw cards, autographed or altered slabs, and slabs with PSA qualifiers such as off-center are set aside and listed with the reason, as are sales that another sale contradicts until they are checked. Several copies sold in the same auction count as one group.
Recent sales count more than older ones, and the estimate changes only when a sale arrives. When the latest sales break away from the old level and nothing in between contradicts them, the estimate resets to the new level.
When a grade has few recent sales, we estimate it from its older sales adjusted for how the market has moved since, from the card's other grades and printings, and from comparable cards, and we mark it Modeled.
The fast read is the middle of the last three sales, so you can see where the latest sales sit against the estimate. The range shows where the next sale is likely to land, based on how accurate our estimates have been on similar cards.
Every number shows how many sales it rests on, and every sale links to its listing.
No one can predict a graded card's value exactly. Sales of the same card in the same grade usually differ by about 7.5 percent. The causes are things no one sees: who was bidding that night, how each copy looked, how buyers acted. Our math is imperfect too. We measure how close we get and show you.
Inputs. Graded sales from CompsAPI, Scrydex and PkmnPrices, plus sales verified by hand.
Validity. Copies of one sale count once. Lots, reprints, raw cards, autographed, altered or PSA-qualified slabs, and titles whose card, printing or grade disagree are excluded with the reason. A cert sold again within 14 days voids an earlier eBay auction. Outliers against other venues or a well-supported estimate, and 1st Edition sales below 0.8 times Unlimited, are held for review unless a different copy within 21 days and 25 percent corroborates them.
Estimate (version 1.2). Prices move to eBay-auction terms by venue gaps per era. The estimate is a weighted median: weights halve every 30 days and taper to zero between days 60 and 120, counted from the latest sale; copies sold together at one venue count as one group; sales beyond 2.5 dispersion units (at least 0.08) count less until confirmed. It is published with five or more valid sales in 90 days. When the two latest sales sit beyond 2.5 units on the same side, the level resets to them and later sales; it becomes permanent after 30 days and moves at most a factor of two until three sales support it.
Confidence. From the evidence (recent sales, weighted by age and plausibility) and dispersion: A with 4.5 or more and under 0.15; B with 2.8 or more and under 0.30; C otherwise.
Ranges. 80 percent ranges from past errors by era and confidence; at confidence C and during resets, a studentized conformal interval whose level is adjusted daily towards 80 percent coverage.
Modeled estimates. For grades with fewer than five recent sales: the grade's older sales moved to today by a set or era sales index (120-day half-life), the nearest grades and other printings scaled by comparable-card ratios, and same-set peers, weighted by how much data stands behind each, kept within 0.9 to 1.1 times two or more recent sales, published where they beat the index-adjusted last sale.
Fast read. The median of the last three valid sales.
Momentum. A sequential change-point test on individual sales; it describes the sales and never changes the estimate.
Safeguards. For the 50 highest-value cards, a nightly move of more than 50 percent is held until it is confirmed.
Market signals. Breadth, Buy It Now share and pace are descriptive only.
Validation. Walk-forward backtests: every estimate uses only earlier sales.
Automated review of suspicious sales. Sales that look out of place (far from the grade's earlier prices, a name or number that could belong to another card in the same set, a grade with very few sales, or a price that disagrees with other venues) are read by an automated reviewer. It sees the listing title and the seller's details, the card and every similar card in its set, and the grade's earlier sales, never anything dated after the sale. It decides whether the sale is this card, this grade and a real single-card sale. When it is at least 90 percent sure, the sale is moved to the right card or grade, or left out; otherwise a person checks it. Every decision is logged with its reasons and the date it could first have been made, so past estimates can be rebuilt exactly as they would have been.
Population counts come from Scrydex and are checked every night; monthly spot checks against PSA's published counts start with the first month of entries.
The catalog lists every English set and its subsets (Shiny Vault, Trainer Gallery, Galarian Gallery, Classic Collection and the like) from TCGdex, checked against the official card totals reported by Scrydex and pokemontcg.io; cards missing from a set are added from those sources. Card images come from TCGdex, Scrydex or pokemontcg.io, and every image is checked to load; a card without a working image shows “No image yet”.
Image use: card images are shown only beside a card's data, at display size. They are never offered for download or resale, never taken from fan sites, and are removed on request from the rights holder (write to scheduledgeoff@gmail.com).
October 16, 2026: version 1.1 is live. When the two most recent valid sales both sit more than 2.5 dispersion units from the current level, on the same side and with no sale in between, the level resets: EMV becomes the recency-weighted median of those sales and every sale since, and sales from before them are left out of the estimate for up to 30 days. The reset ends early if a sale lands back near the old level, after which the usual 90-day window resumes. While a reset is active, the confidence grade is C unless five or more sales have arrived since, the range is the lowest to highest of those sales (since October 17, the conformal 80 percent interval), and the card page says when the level was reset and after how many sales. Version 1 (frozen October 15, 2026) keeps running alongside for comparison.
October 17, 2026: version 1.2 is live. Weights taper from day 60 to 120 and are counted from the latest sale; copies sold together count as one group; confidence comes from weighted evidence; a reset that holds 30 days becomes permanent; outliers are corroborated by any different copy. Tested on 41,128 sales from 2025 and 2026 against version 1.1: median error 12.4 against 12.2 percent, error after a shift 12.6 against 12.5 percent on the same sales (those within 30 days after a shift found by either version), fewer reversals (13.3 against 15.0 percent). These comparisons applied today's sale checks to the past; see Accuracy for the figures as the site behaved at the time. Daily history before October 17 is version 1.1.
October 17, 2026: new 80 percent ranges. Every estimate's range, live or modeled, now comes from how far past sales landed from the estimate, measured separately for grades never sold, grades with no sale in 90 days, grades with 1 to 4, 5 to 8 and 9 or more sales in 90 days, and grades in a reset, with the low and high ends set separately and refreshed weekly. Ranges widen a little for grades that have not sold in a while. Tested on 74,127 sales from July 2025, each against a range built only from earlier sales: 79 to 82 percent landed inside it in every group (the earlier ranges: from 37 percent during resets to 85 percent).
October 17, 2026 (morning). Slabs with PSA qualifiers (such as OC for off-center) and autographed or altered slabs are set aside with the reason (71 and 5 sales); tested on 38,577 sales from 2025 with no change in accuracy.
October 17, 2026 (night). Modeled estimates, version 2: set sales index, nearest grades, a 120-day half-life chosen on pre-2025 sales, weights by data richness and publication by era and recent-sale count (2025 test: 17.4 percent median error against 18.0 for the index-adjusted last sale). A 1st Edition or Shadowless sale priced below 0.8 times a well-supported Unlimited estimate is held for review. For the 50 highest-value cards, a nightly move over 50 percent is held until confirmed. Ranges at confidence C and during resets are conformal 80 percent intervals instead of the lowest to highest sale (2025 coverage 80.9 percent at C and 58 percent during resets, against 83.8 and 35 percent).
October 17, 2026. The reset cap (a reset cannot move EMV more than a factor of 2 from the last settled level until 3 sales support it), corroboration in the review queue, modeled estimates for grades with fewer than 5 recent sales, the fast read as the middle of the last three sales, and nightly population checks.
Daily EMV history before October 16, 2026 was computed afterwards with the current version from the stored sales, each day using only the sales before it.
Momentum describes what a grade's own sales have done. It never changes an estimate. The recent level is the adjusted median of the grade's last K sales (K grows with how often the grade sells, from 3 to 15); the earlier level is the median of its sales from 30 to 365 days before, moved to the present by the era or set index. A move starts when the recent level is far enough from the earlier level for this grade's usual spread (at least 13 percent, more for grades with scattered prices or few sales); it is confirmed after several separate days of later sales that hold the new level (2 to 8, more for busier grades), and it ends when sales move back toward the earlier level, when the estimate catches up with the sales (Settled), or when no sale holds the new level for a while (7 to 45 days, longer for grades that sell rarely). Settled and Reversed notices show until the grade sells again after 7 days, and for at most 45 days, so a grade that sells rarely keeps the notice while the move is still its latest news. The recent-sales summary splits the last 90 days into periods of at least 5 sales and 10 days; the latest period can be as short as 3 sales when it jumped at least as far as a move needs to start, and then shows its lowest and highest sale.
With every direction label, three numbers: the last sales against the sales before them (counts and months shown), the last sales against the estimate (the estimate, its sales and date shown), and, during a move or notice, how far the estimate has changed since the move began (from and to shown). Unconfirmed labels also show how often past signals in the same liquidity group were confirmed or settled within 30 days (grades with 9 or more sales in 90 days), 45 days (5 to 8) or 90 days (fewer), rounded to the nearest 5 percent and shown only with 30 or more cases. Tiles show Rising, Falling and Settled for every grade, and Unconfirmed only for grades with 5 or more sales in 90 days that are not on a modeled estimate.
Up states are blue and down states orange, unconfirmed ones in a lighter tint, notices grey; every label also carries its arrow and its word.
How many cards are selling higher. Each card counts once, whatever its price, and is compared with its own earlier sales. A grade is matched when it has at least 1 valid sale in the last 30 days and at least 2 from 30 to 150 days ago; the earlier sales are moved to the present by the change in their era or set index since the sale. A grade's difference is its current adjusted median minus its earlier adjusted median, in log terms; a card's difference is the median of its matched grades' differences (the mean of the middle pair when even). About the same means within 5 percent either way.
Labels, in this order: Rising broadly when 70 percent or more of matched cards sold higher; else Mostly rising when more than 50 percent sold higher; else Falling broadly when 70 percent or more sold lower; else Mostly falling when more than 50 percent sold lower; else Steady when at least 50 percent were about the same; otherwise Mixed. A group shows a label only with 8 or more matched cards. Percentages within 2 points of a threshold are shown to one decimal place, and the label is computed from the unrounded value.
Counts completed sales only; it does not predict prices and never changes an estimate.
Accuracy line. From the same table as the accuracy explorer: each estimate as it stood the day before a sale, against that sale, grouped by how many sales the grade had in the 90 days before (none, 1 to 4, 5 to 8, 9 or more) and by era, from January 2025. “Half landed within X percent” is the median symmetric error in that group; the line links to the group in the explorer. It describes past estimates, not a guarantee for one card.
Evidence grade. Strong: at least 4.5 effective recent sales (older and clustered sales count for less) and a spread under 0.15. Moderate: at least 2.8 and a spread under 0.30. Limited: anything else with 5 or more valid sales in 90 days. Modeled: fewer than 5 sales in 90 days. Each grade shows the median error of past estimates with that grade.
Next-sale interval. An estimated range for the next sale, built to hold it about 8 times in 10; the page shows how often it did in past checks of the same group. It is not the range of past sales, which is shown separately as “Sales in the last 90 days ranged from…”.
Recent level. The venue-adjusted, plausibility-weighted median of the last sales, in eBay-auction terms; raw prices are in the sales list.
Exceptional sale. When one accepted sale from an auction house that publishes its results, or one accepted in review, is 5 or more times the grade's earlier level on a grade with fewer than 9 sales in 90 days, the page shows that sale, what supports it and what does not, and a working range from past jumps like it, until a later sale lands inside the range or 180 days pass.
As the site would have behaved at the time, measured October 18, 2026. Each sale we counted is compared with the estimate as it stood the day before: for sales before October 18, 2026, rebuilt from only the sales and sale checks available then (the Stage 0 causal replay on the frozen evaluation set from January 2025); from then on, the estimate the site published. 120,444 sales from January 1, 2025 to October 16, 2026: median error 16.3 percent across the catalog (grades weighted by how often each kind sells); 12.9 percent on grades with 5 or more sales in 90 days; 11.9 percent on grades with 9 or more sales in 90 days; 16.6 percent on grades with 5 to 8; 20.8 percent on grades with 1 to 4; 31.2 percent on grades with no sale in the previous 90 days. Unweighted, half of all sales landed within 15.0 percent of the estimate and 80 percent within 44.2 percent; the estimate was above the sale 37 percent of the time. Explore by era, grader, grade, venue and price.
The floor: two copies of the same card and grade sold within 7 days of each other typically differ by about 7.5 percent, so no estimate can be expected to land closer than that to a single sale. An earlier figure on this page (12.2 percent) applied today's sale checks to the past, which made the estimates look more accurate than they were.
Range coverage (how often the next sale landed inside the 80 percent range, on sales from 2025 each judged before it happened): 79.9 percent.
Momentum: possible moves that became the new price within six weeks, over two years: 47 percent; last 90 days: 58 percent.
Refreshed every night.