Now that I've read the article, as a scientist by training and career I'm not impressed with the analysis. It's a poor sampling (n=7) with a subjective scoring system. Scoring on a 0-100 point scale is fine (but he only did a 6 point system: 0, 25, 33, 50, 75, 100%) and including not just product but also sales predictions is good but again with a small sample and subjective scoring, we can't take anything from it. It's not like Kuo is making tons of predictions (around 91 reported by MacRumors if we trust DeWitt's numbers), which means someone should look at all of them (and certainly at least 30 randomly chosen). I'll try to do it at some point.
My bias is to pay attention to Kuo's product predictions but not sales or revenue ones. I think Kuo has solid connections on the product line but any of his other predictions are more speculative. The cult of Mac analysis, while far from perfect, was much better. It was also more supportive of Kuo's accuracy (87+% based mostly on products and not revenue/sales).