ND-USC: A Rivalry Recalibrating

The calendar has moved again, and the thing I have been doing all summer — reading a single ambient signal because there was nothing else to read — is not the thing this chapter can do. Chapter 9 left us on August 31, watching the USC line spike past Notre Dame's on August 29 as Week 0 finally arrived and actual football took over from curiosity. What I want to do here is note, briefly and honestly, that the evidence pool I have in front of me for this chapter is thin in a specific way, and then say what can be said with the thin pool rather than reach past it.

The reason to be up front about this is that the whole premise of the thread, going back to Chapter 6, has been that I would not invent specificity to fill space. The Wikipedia pageview lines we walked through in Chapters 7, 8, and 9 were useful precisely because they were the same measurement, day after day, from the same source — the daily counts published by the Wikimedia Foundation — and I could put them next to each other and let the shape do the talking. When the pool narrows, the honest move is to narrow the chapter with it.

“I grew up in Indiana 60 miles east of Notre Dame. And I was born with cerebral palsy, so I could not play football or sports. As a Notre Dame football fan for nearly sixty years, I have never been so excited for Irish…”

— a viewer comment on Notre Dame Fighting Irish, YouTube, Sep 21

Here is what the pool for this chapter actually gives me. It gives me GDELT article-count observations for two team queries, labeled team:298 and team:338, across a run of September dates. GDELT is a different instrument than Wikipedia pageviews. It measures how often a team surfaces in the news index on a given day, which is a rougher and jumpier signal than the ambient curiosity we have been tracking. But it measures the same thing every day, which is what I need.

For team:298 — the query that has been carrying the Notre Dame side of the ledger — the counts sit at 1 on September 8, 1 on September 9, 1 on September 10, then jump to 6 on September 11, back to 1 on September 12, 1 on September 13, then a gap, 2 on September 15, 1 on September 16, 1 on September 17, and 1 on September 19. For team:338 — the USC-side query — the pool gives me one observation in the same window: 1 article on September 18.

That is what I have. And even with a blunt instrument, the shape is worth sitting with.

The Notre Dame line is almost entirely a floor of 1, with one spike. That single September 11 count of 6 is the entire signal of the month, in the data in front of me. It is a six-fold jump against every other day in the window, and then the line settles back to its floor immediately, with a mild half-echo of 2 on September 15. Everything else is baseline noise.

This is the honest sentence in the middle third: when your daily signal is a flat line of ones and a single spike, the spike is the story, and the shape of it — sharp up, sharp down, no sustained plateau — tells you something happened on one day that did not become an ongoing conversation.

I do not know from this evidence pool what that something was. The pool does not tell me. It tells me that on September 11, the volume of articles indexed against the Notre Dame team query jumped, and it tells me the jump did not carry. The Wikipedia line in Chapter 9 behaved differently on August 29 — it jumped and then held elevated for two more days before easing, which is what you would expect from a signal that had a live game inside it. The GDELT spike I am looking at here is shaped more like a news item than a game weekend.

The USC side of the September pool is even thinner. One observation in the window, on September 18, at a count of 1. I cannot draw a curve from a single point. I can only say that the pool does not, in this window, give me anything like the August 29 spike we walked through in Chapter 9, when USC's Wikipedia line tripled Notre Dame's in a single day. Whatever moved USC's ambient signal that Saturday is not visible in the September article-count observations I have.

Which brings me to the honest limitation of this chapter, and I want to name it rather than paper over it. The ambient-signal argument we have been building across Chapters 6 through 9 was strongest when I could put two lines next to each other, day after day, and let the reader watch the gap widen or close. The GDELT pool here does not let me do that side-by-side comparison the way the Wikipedia pool did. It gives me a longer Notre Dame series and a single USC point, and it gives me a shape — spike on September 11, flat around it — that I can describe but cannot fully explain from inside the evidence.

What I will say, and what I think is defensible from what is on the table: the September signal for Notre Dame, as measured by the article-count series in this pool, is not a ramp. It is a floor with one bump. That is a different shape than the August pageview lines we walked through in Chapter 7, which held above 1,000 for ten straight days on the Notre Dame side, and it is a different shape than the sustained widening we tracked in Chapter 8 from August 19 through August 24. Whatever the season is doing to ambient interest in this rivalry, the piece of it visible to me right now looks episodic rather than accumulative.

The next chapter will need a better pool to do more than that. If the daily counts fill in on both sides — if I can get the USC series back to something I can lay next to the Notre Dame series the way we did with Wikipedia — the argument I have been building can pick back up where Chapter 9 left it. If not, the honest thing to keep doing is what this chapter has done: describe the shape of what is actually in front of me, note where the September 11 spike sits inside that shape, and refuse to fill the space around it with specifics I cannot source.

The ambient window Chapter 9 closed has not reopened in this evidence. It has narrowed to a floor and a bump, and I would rather leave the reader with that honestly than dress it up.