September is when the archive becomes the product. During the season there is a show every few days from June through August and a recap due the next morning, and nothing in that schedule leaves room for sitting with four years of Finals results at once. So here is that piece: 2023, 2024, 2025 and 2026, score by score, at the two places on the sheet where a season is decided.

First place is a coin toss. Third place keeps falling away

Winning margins across the four Finals: Blue Devils by 1.237 in 2023, Bluecoats by 1.337 in 2024, Boston Crusaders by 0.175 in 2025, Bluecoats by 1.600 in 2026. There is no trend in that. The closest Finals of the four and the widest happened back to back.

The distance from first to third is a different number and it behaves differently. 1.312 in 2023, 1.675 in 2024, 1.725 in 2025, 2.062 in 2026. Four Finals, three increases, no year down.

2025 is the year that separates the two measurements. Boston Crusaders and Bluecoats finished 0.175 apart at the top, the tightest result in the set, and the spread back to third still grew, because Santa Clara Vanguard came in at 96.700, more than a point and a half behind second. A close win and a wide podium can happen on the same night, because they measure different things.

Twelfth place ends where it started

The score that holds the last spot in the top twelve: 86.375 in 2023 (Blue Knights), 87.050 in 2024 (Madison Scouts), 87.450 in 2025 (Blue Knights), 86.175 in 2026 (Blue Knights again). It drifts up for two years and comes back down. The whole range across four Finals is 1.275 points, and 2026 finishes lower than 2023 started.

So the top of the field is spreading out and the bottom of the top twelve is not moving with it. Those two facts are the piece. They are also the whole of what these numbers support.

What this does not establish

Four numbers pointing one direction are a pattern, not a cause. Nothing above identifies why the podium is spreading.

Panel composition changed across these four seasons, and caption weighting changed with it. Neither is controlled for in any figure here, and no arithmetic on final scores can control for them. A spread that widens while both the people assigning the numbers and the rules for assigning them are changing is a spread that widens for reasons this data cannot separate.

Programming risk, a top tier no longer bunching at the ceiling, judges more willing to spread a field: each is a plausible story and none of them is in these numbers. Nobody who judged these shows has been asked. Putting any of them forward as a reading of the scores would be inventing a finding, and blaming a corps for what might be a panel artifact would be worse.

What the archive supports is the shape. First to third has widened four years running, twelfth has not moved, and the two are not tracking each other. Why is a different piece, and it needs people rather than a spreadsheet.