By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Farsala on 07 December 2012
Edit Thread
Pineapple said:

How did you guys make your predictions?

This year, I simply used a statistical method to get my predictions.

(1) I looked at the sales from the past 2 years, and calculated the percentage of the total over the 10 weeks that was sold in each week. So my predictions have the machines trending exactly as they did the average of the last 2 years.

(2) Then, to predict the total amount sold over the 10 weeks, I looked at the total sold up to mid October in 2012, and then compared that to the 2011 and 2010 numbers. In other words, with this method I assumed that the trend that was held through the entire year up to now would keep. That's a slightly flawed assumption, so I added in another method.

(3) I looked at the numbers from the start of September to mid October 2012, and compared those to the 2011 and 2010 numbers in the same period. Then I found the ratio, and assumed that this trend would keep into the holiday season.

(4)In the cases of the X360, Ps3 and 3DS, I weighted (2) and (3) equally, simply using the average between the two. I multiplied that average ration with the total sales in the last 10 weeks of the year for 2010 and 2011. That gave two expected total sales for 2012, and I took the average of those as my prediction for total sold for the year.

(4) With the Wii and Vita, I weighted (3) twice as strongly as (2), as they trended downard for the year. I then did the same as above to make my prediction for total sold. For the Wii, I multiplied the results with 0.85, because I felt the method gave it a too high prediction.

With (1) and (4) I now had both a prediction for the relative sales between the different weeks, and a prediction for total sold. Thus, I simply divided the total sold between the weeks as dictated by (1).

 

And that's the method in its whole. It's worth noting that with the 3DS and Vita, I added together the DS and 3DS data and the Vita and PsP data, because otherwise I wouldn't have enough data to make a prediction. In other words, I made the assumption that the 3DS demographics is identical to the DS one and the the Vita demographics are identical to the Psp one. Naturally, this is a flawed assumption, but one that had to be made for this method to work. I'm thus very surprised that the Vita is the machine I'm predicting best.

 

This method is completely different from what I did in past years, which was mostly just following my guts.

Pretty much did what you did except in a much faster and exxagerated way.