The Spike Report · Corrections

Published: Monday, 5 October 2026. Covers: editions one through six. Source: JustTCG.

This is a standalone corrections notice rather than an edition. It exists so the record is in one place and can be linked to.

We spent the week before edition seven auditing our own build instead of adding anything to it. Six things came out. Two are missing filters, two are figures we described wrongly while computing them correctly, one is a superseded measurement, and one is a fault in the source data. Four of the six were in numbers we had already published more than once, and all six were found by checking the build against its own documentation rather than by anyone noticing from outside.

What is not affected: the board. Every Top 10 Movers list we have published had all four eligibility filters applied, because those filters live in the ranking code. Everything below is in the aggregate tables, the base-rate measurements, and the way we described two windows.

Editions one to six. Fixed 5 October.

The Spike Report defines a chase tier — ex, V, VMAX, VSTAR, Full Art, Illustration Rare, Special Illustration Rare, Hyper Rare, Secret Rare, Alt Art and their printed-rarity equivalents — and the board has always honoured it. The era, set, age and breadth tables were labelled chase tier and were not. They included every tracked card above $10.

On edition seven’s data that meant fourteen extra cards: nine Holo Rare, two Radiant Rare, one Amazing Rare, one Rare, one Uncommon. On the base-rate population, which applies the price floor differently, it was eighteen.

The effect on any single figure is small. It is not zero, and it lands hardest on the rows with the fewest cards.

Editions one to six. Fixed 4 October.

Same shape of mistake. The board excludes sets less than ninety days old, because a new set’s prices have no track record worth measuring. The tables did not.

We went looking for this because 30th Celebration entered our index on 16 September at eighteen days old. 30th Celebration turned out not to be the culprit. Pitch Black was — a 17 July set that has been under ninety days old for every edition since it was added, contributing five cards to the Mega Evolution row every single week without ever being eligible for it.

The one correction here that changes a headline number. Edition five’s Mega Evolution row was the centre of that edition. We published it as -4.34% median, 86% declining. On the corrected basis it is -2.88% median, 83% declining — a 1.46-point error on the number that edition was built around, and it made the era look materially worse than it was.

Every era row we have published in editions five and six, corrected:

Breadth figures: edition five was published as 312 cards and is 294. Edition six was published as 313 cards at a median of -0.195% and is 294 cards at -0.153%.

Editions one to four carried the same two gaps. We have not reprinted their era tables here, because the sets in the universe have changed since and a recomputation on today’s membership would be a different measurement rather than a corrected one. If you want a specific figure from one of those editions checked, ask in the comments and we will run it.

Both missing filters were one mistake made twice: we asked what the board excludes and never asked the tables the same question.

Editions four to six. Recomputed 5 October.

Because of correction 1, the base-rate measurements behind Early Signal were computed over a population that was not the board they were being compared to. The “rest of the board, same dates” column was comparing against a different board.

Recomputed on the chase-tier population, with cards whose prices failed to refresh removed and two corrupted series truncated (see correction 6):

The decline figures we printed were overstated, by between 0.42 and 1.10 points, worst at the first reading. The climb figures were understated, by between 1.57 and 2.05 points. Both errors came from the same cause, and they ran in opposite directions because the cards that should not have been there were cheaper and more thinly priced than the board they were standing in for.

The corrected series is flatter and less dramatic in both directions, which is what a real base rate usually looks like.

Editions four to six. Corrected in edition seven.

Every edition has described the base-rate window as “the trailing ninety days” and named a start date ninety days back. It was never ninety days. The harness uses every price point it holds, and our stored history begins on 26 May 2026. At edition seven’s cutoff that is 132 days, which at a 30-day forward horizon gives 102 usable signal dates per card rather than 60.

The figures were computed correctly over the full series. The sentence describing them was wrong, and it was wrong in a way that was visible in our own published numbers: edition six reported 42 distinct firing dates and edition seven reports 76, and neither of those can fit inside a window that only offers 60.

The practical consequence is that the six-month window we have promised for late November is a real extension rather than the relabelling of an existing one that it would otherwise have been.

Editions five and six. Corrected in edition seven.

The review list holds cards the artifact guard has flagged, which stay off the board until they record four distinct prices after their step. We have published a running total three times.

It is not a running total. The list is rebuilt every week by replaying the guard across the three most recent cutoffs, which is a design choice we described in edition four and then did not follow through on. The three-week window slides forward every week, so each rebuild silently drops anything that entered before it.

On the history available at edition seven’s cutoff the series reads 11, 16, 21, 22 — the list has grown every week since we started printing it. Edition six reported it as “twenty-four, down from thirty, the first time it has shrunk.” That shrinkage was the replay window moving, not cards clearing.

The weekly figures for cards added, cleared and suppressed are measured at each cutoff and are sound. The totals are not comparable across editions, and we will not print them that way again.

One related figure in edition six was also wrong on its own terms: we wrote that M Rayquaza EX was the longest-running entry on the list. Four entries had been on it longer.

Found 5 October. Flagged in the build from edition seven.

Altaria ex (253/182, Paradox Rift) and Gardevoir ex (233/091, Paldean Fates) each carry forty-two consecutive sub-$1 price points from 26 May to 6 July, and then jump on a single day:

Whatever happened upstream, the series before 7 July is not these cards’ price. Any window spanning that date is meaningless: Altaria ex’s 90-day change computes to +5,902.6%.

Altaria ex appears on edition seven’s board at number four. Its 7, 14, 30 and 60 day figures are all after the jump and are sound; nothing longer exists for it and nothing longer is printed. Gardevoir ex has not appeared on any board.

A $10 price floor was the only thing keeping these points out of the base-rate measurements, which is luck rather than design — a scale break above the floor would have gone straight through. From edition seven the build raises a named warning on any single-day move of 3x or worse, and we decide case by case which windows survive. We are not dropping real movers over a source fault; we are refusing to average across one.

The test suite covering the eligibility rules went from 34 cases to 40.

Four of these six were in figures we had already published more than once, and we found all six ourselves rather than being told. That is not a comfortable ratio, and the reason it happened is that the board and the tables were written at different times and never audited against each other.

If a number in any edition looks wrong to you, say so in the comments. Corrections get published, and they get published with the size of the error in them.