I am aware that approval voting doesn't have a straightforward "honest" strategy, and mentioned this in the article. It effectively requires voters to choose a cutoff, which is an inherently tactical decision. Perhaps there's a better choice I could have made about where to choose that cutoff in the absence of expectations about likely winners. I could see an argument that instead of 50% of approval, it should be at something like 1/sqrt(n_candidates), assuming that the top two candidates are likely to appear in the upper part of your own honest rankings, but this too depends on some assumption the voter makes about how their own preferences will compare to other voters.