Why it's an approximation
The ¾-word rule holds for ordinary English. Code, numbers, URLs and most non-English languages use more tokens per word, and every lab's tokenizer splits text a little differently. For an exact count of your own text, use the token counter.