By using this site, you agree to our Privacy Policy and our Terms of Use. Close
TWRoO on 05 September 2012
Edit Thread
Chark said:

TWRoO said:

(.........)

Alternately you could continue using weeks but use a rolling average found from the past few weeks... which means you can still predict the future sales in weekly increments. So you still get the nice curve upwards through November and December, with the tooth of Black Friday in it, but it still irons out any of Vitas anomalous weeks. (It won't help iron out any anomalous weeks the other platforms had in 2011, but the fact you are doing an average of multiple platforms should be helping that already)


I could switch to months but the damage has been done already, besides the massive fluctuations keep this thread lively. Just wait till it shoots up and all the crow will be served to the other camp.

Well switching to using an average across a number of weeks is simple enough.

I mean your formula now is probably something like this:

[week n] = ( [week n-1] / [average of week n-53] * 100 ) / 100 * [average of week n-52]

The bit between the rounded brackets is what works out the percentage Vita is selling compared to the 2011 average... as long as the "week n-1" period corresponds with the "week n-53" period it can be any length you want. I think a period of 1-3 months is best because it doesn't get skewed too much by anomalous weeks but also doesn't go too far back in time including possibly irrelevant data (particularly in the Vita's case as going back more than 6 months would include launch sales)

And regarding crow eating, if the prediction graph shoots up because of an anomalous week, it's only going to come down again in the following week or two.
Looking at next weeks sales for example with Japan at ~50k and the usual 20k in the West: Assuming I am still correctly interpreting your formula your graph is going to be predicting the Vita to more than double the current prediction for the rest of the year, bringing the end of year total to around 5.5 million. Then if past trends in Japan are anything to go by, the following week Vita will be sub 30k in Japan, then sub 15k again the week after and your graph will be showing an end of year about 4.1 or 4.2 million again. (at least assuming there aren't any other special bundles in Japan I am unaware of)