The scoreboard case first, because it is simple: Crown sit at 97.038, fourth — 2.1 behind Bluecoats and a tenth behind the two corps tied above them. Between DCI World Championship Prelims and DCI World Championship Finals — the 3 most recent nights when all four top-tier corps stood in front of the same judges — Crown gained +0.56, the most of the four. No cross-panel asterisk applies. Same sheets, same rooms, fastest climb.

Now the model’s answer: 0.0% to win gold, against 100.0% for Bluecoats. The trajectory fit labels Crown steady — not accelerating — and projects a Finals median of 97.65 to Bloo’s 98.96. The machine has watched the same season you have and concluded the climb is already priced in. The question this piece exists to answer: is that discipline, or is that a blind spot?

Read the gain column and the ordering is unambiguous: Crown first, then the corps the model actually likes. The gap series tells the same story from another angle — Bluecoats’ lead over Crown on a shared sheet has gone 1.82, then 1.76, then 2.06. That is a lead being ground down, not defended. And the grinding is happening at the exact point in the calendar when scores compress and gains get expensive.

The build behind it is brass-first: Crown’s 19.45 at DCI World Championship Finals was the best brass number on the sheet — better than Bluecoats’ — the signature caption doing signature-caption work. That matters for the projection argument, because brass and GE are the captions that historically keep climbing through August. The caption that pays Crown is the caption that ages well.

There is a forward-looking number that cuts the same way. The stats account @dcistats published a “cleaning potential” read on the tier — an estimate of the points a corps stands to recover simply by tightening execution as the book gets cleaner through August — and put Crown at +9.6, the most headroom of the top four. Read against the projection argument, that is the tell: the corps already climbing fastest on shared panels is also the one carrying the most uncollected points into the stretch. A model that treats the last week of the calendar as flat is, in effect, betting Crown leaves all 9.6 of those points on the field.

The model is not being stubborn for sport; it has three defensible reasons. First, the fit is damped by design — it has watched enough Julys to know that every slope in the tier flattens in August, so it refuses to run Crown’s last week out to Indianapolis at face value. Second, the ledger: Crown are 0–10 against Bluecoats this season. A model that hands big gold odds to a corps that has never once beaten the leader is not a model, it is a mood. Third, percussion: Crown’s 18.45 at DCI World Championship Finals was the lowest battery number in the tier, 1.4 off the best on the sheet. Closing on the leader while conceding a full caption is a hard trick to sustain.

But here is the crack in the model’s armor, and it is visible in its own output: Crown’s 95th-percentile Finals projection is 97.9, and Bluecoats’ median is 98.96. The model’s own upper tail puts Crown within reach of the leader’s expected number — it just considers that tail unlikely, because the damped fit treats the last week as noise. If the last week is signal — a design change landing, a corps finding its August gear early — the damping is exactly backwards, and 0.0% is the artifact of a prior, not a probability.

Here is the desk’s position, stated so it can be wrong: the model is underrating Crown, and the next two shared panels will show it. Specifically — if Crown cut the gap to Bluecoats under 1.5 on a shared sheet by the next shared panel, the climb is real, the damping is too heavy, and gold odds belong in the high teens, not single digits. If instead the gap holds at 2.0 or wider through those meetings, the model read the compression correctly, the desk was seduced by a good week, and we will say so in this space — in those words.

One number decides it, and it goes up on a scoreboard soon. That is the whole appeal of writing about a sport that keeps score.