"Consider vocabulary, for example—in all languages, vocabulary follows a Zipf distribution, meaning it's a sharp exponential where a small fraction of the lexicon comes up all the time in natural utterances, and the vast majority of the lexicon is in a long tail of sometimes-words. Your first task as a learner is to get familiar with those ultra-frequent, common words. By definition, the words that come very often all the time will come up very often in your text, too, no matter how fictional or niche it is. This is why “spaced repetition” methods are not useful; when learning from natural input, you always exercise vocabulary at the optimum frequency of repetition."