Amazon US December monthly bestsellers and general Amazon-based discussion

[binary solo]

celador said:
I certainly don't think Amazon threads should be abandoned after one incorrect result, even if it is as wrong as an Xbox win tomorrow would be.

If it is wrong in December as well, or maybe early next year with no explainable circumstances, then it is time to consider how useful the thread is. You could use other retailers, but it risks becoming diluted or complicated. If three stores have PS4 winning, and two X1, and X1 wins the month, are they really wrong overall? It would be difficult to tell because you would have to think about how close each retailer was, what different deals they all had, how much market share they have etc.

But with multiple data sources, and an understanding of market share for each retailer, then a highly accurate predictive pattern can be established. If the 2 retailers showing Xb one as #1 have a larger market share than the 3 retailers showng PS4 at #1, then on balance our hyopthetical thread should be calling a Xb one win for the month. If we start with no knowledge of market share and simply see how the chips fall over the course of a few months then we can see whether a partern emerges or not, and even contrast with whether a multi-chart tracking is more or less reliable than tracking Amazon alone.

it is possible to flip a coin and get heads 12 times in a row, unlikely but possible. So it is possible that all the times Amazon was right it was always a fluke and just one of those rare occurrences of chance appearing to establish a predictable pattern. There will need to be a few more months of Amazon being wrong to be able to confirm whether Amazon really is hopeless at predicting the NPD outcome. If Amazon is a reliable predictor of the NPD winner it should predict the Nov 2015 winner, because most of the factors driving hardware sales throughout the momth favoured the console that ended up highest on the Amazon monthly chart. Even Infoscout validated Amazon, though Infoscout's methodology has such a high margin for error that a 1% difference basically means NPD is 50/50.