What about other recipes which remain close to the average in my data ? Pet food for dust relics is more likely to be in high-frequency pool, due to symmetry reasons with other Pet foods costing relics. Similarly, 2 Runeshards recipe is more likely to be in the lower pool. For 3KP, as 1KP is the only other recipe in higer pool, we can't guess without data. Is it more far to the average in your data ?
Finally, there's the case of supply windfall 15%. It used to be lower than average until I've got a ton of these recently. So what does your data say ?
Note : I think CR33% and SW100% are in the higher pool since other CRs and >25% SWs are high-frequency too.
I mean, we can speculate whether something can fall into low frequency or high frequency pools, based on some circumstantial evidence (like other recipes in the same group). But ultimately what matters is what data shows, and for those recipes that are close to the cut off that's exactly what it means - that it may ultimately fall into either bucket, once we get more data. The farther it is from the cutoff, the less likely for it to happen (but may happen nonetheless).
My point was that even with relatively low number of observations for my cities, our bucketing results match pretty much exactly so far. That's not to say that these are not wrong in both at the same time, for some recipes. But it would be much more concerning if we would get very different views on what is high frequency and what is low frequency. That would indicate that either it is player specific, the whole process is not stable or a multitude of other things.
As of now, there is nothing in our data that disproves the 2-pool hypothesis. But it's not quite exactly the same as saying that the data proves it
