PLAY’S · META · 26 Oct 2026

A deck that was on twenty-nine charts in a row is not on the last three

《“B-零朱”レイド》 was changed at the 22 October maintenance. On the three full days of play since, 赤白レイド is not a named slice on the daily ranked survey — it was one on all twenty-nine days before. 赤白轟轟轟(レイド無し), the build the chart names without that card, has gone 11.9, 11.6, 16.6 against a ceiling of 3.6 on the days before the change. What the chart records is which of the two builds gets a label, and that has changed hands. It cannot measure what the two colours hold together, the size of the move is inside what this survey does on quiet weeks, and it cannot say what caused any of it.

What stopped, and on which day

This site’s own readings, the seven days around the change. The red rule marks the change: everything below it was played under the new card text. The 22nd is about eight hours of play — the game was down from 23:15 the night before until 16:00 that afternoon — so its games are all under the new text and there are only forty-six of them. That chart also names nineteen slices and has no unnamed bucket at all, against eleven slices the day before: at forty-six games most decks are one or two, and the survey labelled all of them. The 4.3 in the Raid-less column that day is two games, and the day is after the change, so that build’s ceiling before the change is 3.6, which it reached on three days — the largest of them 17 October, six games out of 166. The “pair” column adds the two red/white columns and counts an unnamed build as nothing, so each figure is a floor and not a total.

赤白レイド was a named slice on twenty-nine consecutive days, 24 September to 22 October. Its readings in that run go from 6.5 on the part-day of the 22nd to 24.4 on 24 September, which is a partial six-column read. It was the chart’s leading named deck on thirteen of the thirty-two days and tied for the lead on one more — more often than any other deck in the series; ジョーカーズ is next with seven and two. Counting only the twenty-five days read in full, it led eleven and tied one against ジョーカーズ’s six and two.

Then nothing, three days running, across samples of 227, 146 and 290.

Of the eight decks with a long run, seven are still on the chart

Twenty-five of the thirty-two days are full reads, where every slice was recorded. Eight decks are named on twenty or more of those twenty-five. Six of them — ジョーカーズ, バラギアラループ, ゼニス天門, 赤単バルガ, HADESドラゲリオン and コットンゼニス — have never missed a full-read day. 青魔導具 has missed exactly one, 18 October, and is on all three charts since the change.

赤白レイド is the eighth, and the only one of them absent from the last three charts. All three of its absences are those three days.

“Not a named slice” is not zero, and the ceiling here is the chart’s own

A deck under the survey’s labelling cutoff falls into the unnamed bucket rather than leaving the format. That bucket reads 13.4, 8.2 and 11.9 on the three days — about 30, 12 and 35 games. So the honest statement is that the deck is below the chart’s smallest named slice, not that it is gone:

• 23 October — smallest named slice 2.6, which is about six games in 227. So at most five of the day’s games.
• 24 October — smallest named slice 2.1, about three games in 146. At most two.
• 25 October — smallest named slice 2.1, about six games in 290. At most five.

And that ceiling is inferred, not published. This site does not know the survey’s labelling rule. It is not a fixed number of slices — the three days name ten, twelve and thirteen decks plus the bucket, and the 22nd names nineteen with no bucket. It does not look like a flat percentage either: 2.1 is named on two of the three days, while the bucket on the 25th holds about thirty-five games. The figures above are read off each chart’s own smallest label and are worth exactly that much.

What the two colours hold together cannot be read off this chart

Add the two red/white columns and the series looks different from the headline — but only up to a point, and the point is the whole of it. The sum is a floor, not a total: the chart names a build only above its labelling cutoff, and everything under that sits in the unnamed bucket. Across the twenty-one full days before the change that floor runs 10.7 to 27.7. Since the change it reads 11.9, 11.6 and 16.6.

Two things stop that being a finding, and both were missed in the first version of this page.

The 10.7 is in the previous ranked season. Restrict the comparison to the ladder the three days are actually on — the six full days from 16 to 21 October — and the floor ran 11.9 to 27.7, averaging about 20. Against that, the 11.6 of the 24th is below anything the pair had held in the comparable window, and the part-day of the 22nd at 10.8 is lower still. So the pair is not sitting in the lower half of a band it already occupied; on one of the three days it is under the band.

And the ceiling is wide open. The bucket holds 13.4, 8.2 and 11.9 on the three days, so the pair’s true share could be as high as 25.3, 19.8 and 28.5. On 21 October it could have been as high as 32.9. Those ranges overlap, so this chart cannot say whether the two colours hold more or less of the format than they did — not in either direction, and a page that said the combined share had not grown was reading a floor as a total.

What the chart can say is narrower: which of the two builds carries a label, and that has changed hands.

The size of the move is ordinary for this chart

赤白轟轟轟(レイド無し) went 4.3 on the part-day to 16.6 on the 25th, 12.3 points across four charts. Measured against this survey’s own volatility that is not remarkable. The two widest four-chart swings in the series are both larger, and both happened with nothing altered in the game:

• 赤白レイド read 24.1, 15.0, 8.3, 18.2 across 17 to 20 October — 15.8 points.
• ジョーカーズ read 10.8, 11.4, 26.5, 14.3 across 16 to 19 October — 15.7 points, though that window opens on the first day of the new ladder.

Over the three full days the Raid-less build moves only 5.0 points — 11.9, 11.6, 16.6 — against a widest three-chart swing since the ladder reset of 15.8, which is Raid’s own, 17 to 19 October. Either way round, the rise on its own needs no explanation: this chart produces moves that size on quiet weeks.

The run ending is the harder thing to put down to the ordinary. It is still not unique

Raid’s twenty-nine days is the longest run that ends inside this record, and the only one that ends on the first full day under a changed card. But it is not the only run that ends, and the nearest precedent is already on this page:

• 青魔導具 was a named slice for twenty-four days running, missing on 18 October, and named again the next day — with nothing altered in the game. This page counts that gap two sections above; the first version of it never came back to what the gap means.
• その他, the unnamed bucket itself, ran twenty-eight days and stopped on the 22nd, because that day’s chart named all nineteen of its slices.

Twenty-nine against twenty-four, inside a window only thirty-two days long, is a margin and not a category. A run ending is not by itself evidence of anything, and the 18 October gap is the reason to say so. What is left of the finding is the date: the streak stops on the first full day under the new card text, and that is a coincidence of dates until something else confirms it.

The chart’s other build named for omitting the card did not move

赤単轟轟轟(レイド無し) — mono-red rather than red/white, and named the same way for leaving the card out — reads 2.1, 3.6, 2.6, 2.3, 2.5, 3.8 and 2.4 on the seven full days it was named before the change, 4.3 on the part-day, then 2.6, 2.1, 2.8. It is flat across the change.

That is a fact about the chart and not an explanation of one. At two to eleven games a day — eleven of them on 15 October, the largest sample in the series — the column is inside its own noise either way, and nothing in this data says why a deck did or did not move.

What the players said is on another page

Twenty-one posts from 19 to 25 October were collected and written up separately, as talk and not as fact: five describe the deck as gone, three name the Raid-less build as what replaced it, two are from people still playing it, and the largest community post in the set argues the complaints are overblown. That page deliberately stopped short of the result this one reports, and said so at the time.

The two should not be stacked into one piece of evidence. A search that returns between seven and eleven posts is not a sample of players, and a survey of ranked games is not a survey of opinions. They agree here; that is worth noting once and not leaning on.

What this page is not

It is not a verdict on the change. The notice of 20 October states the change as balance adjustment and names no deck, no tournament and no figure, so there is no stated intent to measure the chart against. A page saying the change worked would be inventing the goal first and then scoring it.

It is not a claim that the change caused this. Eighteen things changed on the same afternoon, and the survey has one column for the outcome and none for the cause.

The survey counts decks met in ranked games, not players. One build taking another’s place on a pie chart is not evidence that the same people rebuilt the same deck; the chart has no way to follow anybody from one day to the next.

And it is three days. The standing example on this site is ジョーカーズ reading 11.4, then 26.5, then 14.3, then 22.7, then 20.8 across five days with the game untouched. A week is the shortest window this site has been willing to call a move. Week five of these readings covers 22 to 28 October and is not written yet; week four is the most recent one that is.