Any sample size over 30 can be used to make a normal distrubution on a bell curve. We can say with a great level of confidence that the best sellers on Amazon will be in the correct order when they release because the sample size is so large.
Ex. Destiny PS4 #1 Destiny Xbox One #2 ...etc.
There are problems with using Amazon as an indicator for sales. There are Biases such as people with internet tend to order from Amazon and may purchase primarily online games. preorder bonuses for certain games at other stores etc.
I would say Amazon is a good indicator of sales 95% of the time. Just look at Amazon hot sellers vs. The actual sales charts! 9/10 games are in there right place. Is Amazon a good indicator?