Here is a sentence every corps parent has said into a phone: “They were better tonight, and the score went down.” Usually both halves are true. The run improved and the number dropped, because the number doesn’t come out of a laser. It comes out of a chair, and the chair holds a different person almost every night.

A caption score has two parents: the corps and the room. The whole activity tracks the first one obsessively — trend lines, season highs, gain per show. Nobody puts a number on the second. So we did. No judge is named below, and that’s deliberate — this desk doesn’t print names it can’t verify against a recap header, and the argument doesn’t need them. The finding isn’t that any judge is wrong. It’s that the chair itself — any chair — is a bigger variable than the leaderboard admits.

The trick is the one this desk uses everywhere: trust the trend. Every corps’s caption scores climb through July on a nearly straight line, so we fit the line and look at what’s left over. A mark sitting above a corps’s own trajectory means that night’s chair read them warmer than the season’s consensus of every other chair; below means cooler. Statisticians call the leftover a residual. Fans call it “what just happened.”

Average one night’s residuals across everyone on the sheet and you get the room’s temperature in that caption. Do it for all 248 rooms this season — 1624 caption marks across 35 shows — and the typical room sits about 0.12 from neutral. Small number, until you remember corps separated by less than that are treated as different tiers by every message board in America. A hot room isn’t wrong, either. It’s just warm. That distinction carries the whole piece.

Look at the coldest row: Visual Proficiency at Midwestern Championship, where 7 corps landed an average of 0.65 under their own trend lines — on the same night, in the same caption, from the same chair. 7 simultaneous off nights in one caption is possible the way a family of six all getting food poisoning from different restaurants is possible. One cold read is the cheaper explanation. And the reverse holds for the warm rows: when a whole sheet runs two- or three-tenths high, nobody on it suddenly got better. The building did.

The widest split in the league: Seattle Cascades, Color Guard. At DCI West (Jul 5) the chair read them 1.39 under their own trend; at DCI New Mexico (Jul 14) a different chair had them 0.85 over it. Two rooms, 2.24 apart on the same book. The performers did not forget how to perform in between. Two sets of professional eyes weighed the same demand differently, which is what happens when humans grade art at 200 beats per minute.

Now the other end, because it’s the part everyone skips: Boston Crusaders’ GE Visual number carries a chair-to-chair deviation of 0.10 — every room they’ve walked into all summer landed within about a tenth of the same read, and these chairs never compare notes. That’s the system working. The contested rows aren’t the system failing; they’re the captions where taste legitimately has more room to operate.

The ranking is not random. Color Guard and Percussion sit at the top — captions where a chair is weighing vocabulary, risk, and responsibility against achievement, live, with no replay. The bottom of the table is the ensemble-analysis captions, where the rubric leaves the least room for a point of view. If you want to know which caption swing to shrug off on a Tuesday and which one to take seriously, this table is the shrug schedule.

There’s the season’s biggest storyline with its makeup off. Crown’s first win over Boston — DCI Southeastern Championship, +0.70, the result the entire Crown-is-coming argument leans on — came with a room swing of +1.74. Re-room that night onto a season-median panel and the margin reads -1.04 — a Boston win. The caption receipts say the warmth pooled where Crown live: Brass ran +0.46 off the expected gap, GE Visual another +0.41.

Be careful with the conclusion, because it cuts both ways. This does not say Crown’s win was fake — a warm room still watched Crown beat Boston on the sheet in front of it, and rooms have run warm for Boston too; read the whole swing column. It says the margin travels badly. Which is exactly what the model quietly concluded when it kept Boston ahead of Crown in the projections the morning after Atlanta, to this desk’s loud confusion at the time. The machine had already done the re-rooming. We’re just now showing the arithmetic.

This is not a bias audit, and if you quote it as one, quote it wrong somewhere else. DCI adjudicators are professionals working a detailed rubric at field speed, and the consensus table above is their character witness: independent chairs agreeing within a tenth across an entire summer is a level of calibration most industries would frame. The spread that remains is parallax, not agenda — different seats genuinely see different shows, and the activity is better for hiring judges with points of view instead of eight laser rangefinders.

What the data does say: panel composition belongs in every scoreboard argument, the same way this desk already treats split-field nights (different panels, different rulers) and one-chair caption swings like Crown’s Sunday dip. When a score jumps or sags between shows, check who the corps were standing in front of before you decide what it means — the Judges desk tracks every panel, show by show. And when rivals finally share a sheet, as Crown, Boston, and Vanguard will in Allentown, the room stops being a variable and starts being a referee. Both corps get the same eyes. That’s why shared sheets settle arguments and Tuesday-night scores start them.