Having more data does trump a better algorithm, but it's not that simple.
Data collection and analysis is about having the right data not more of it. Correlations without a hypothesis are just correlations, interesting but of no intrinsic value. It's why the China Study is not only stupid, but wrong.
Your goal should always be to find data that is germane to your goal both good and bad and evaluate your hypothesis within that data set.
Data is always telling you something and very often it's outside the parameters of whatever your algorithm/model du jour is.