Trends transition analysis doesn't seem consistent with examples?

I’ve just started exploring the trends database, and thought it looked really cool. However, I quickly stumbled across a number of examples where the network diagram analysis shown on the left hand side seems to be directly contradicted by the examples on the right, and now I’m very confused and not sure what to make of the analysis. Perhaps I am missing something?

I’ve pasted a screenshot below that illustrates what I’m talking about (as a new user, I am apparently prohibited from pasting more than one). I stumbled across this example by accident, and then took only a few minutes to find quite a few more. While I can’t paste screen shots, I describe a couple of these examples below too. Given how easy it was to find these, I’m now a bit concerned about whether I can really trust the analysis in general…

Example 1: Rel Minor network map suggests 100% of transitions from i → iiø7 go to i6, but many of the examples appear to contradict this (one of these, which goes to V, is displayed on the right, but it’s not the only counterexample).

Example 2: Rel Major network map shows no further nodes after a I → II4/2 transition, but e.g. The Globalist by Muse transitions I → II4/2 → V6.

Example 3: C Major network map shows the only transitions from C->B are to Bb, C, or F, but e.g. Baby Bride Rag by Roar (sorry there’s no link to this one; as a new user I’m also apparently limited to 3 links in a post) has multiple instances of C->B->G7.

As mentioned above, perhaps I am missing something obvious here. If so I would love to be enlightened!

1 Like

@conchis thanks for the report. My guess is that it has something to do with our thresholding algorithm… right now, we don’t report a progression in trends unless a minimum number of songs use it (I think right now it’s set to something like 3). This is because if a chord sequence only occurs once, we feel like it’s not really a “trend”. However, these songs can still be navigated in the tool, which is why their statistics aren’t showing up correctly.

1 Like

Thanks Ryan, that makes sense, and appreciate the response.

(Although it seems like the threshold may be set higher than 3 as I’m pretty sure in the first example I can see 3 or more transitions for each of i->iio7->V and i->iio7->V7 - as well as some with other variations on V.)

Obviously this won’t matter much unless you’re already in the realm of relatively rare progressions.