[htfu]

Yup. As greek to me as the paper is at least it makes very clear what it sets out to achieve and why (and when) it differs. I suppose it's implicit but I feel article really ought to explain that in the demonstrated case of heavy bias, few attempts and fixed, coarse steps there is of course no advantage - apart from the stick in ground one could also best its resolution off 0b1000000 and ++/--.

It's a nice explainer on polarization but tries to be more than that and doesn't achieve it - but with further work (not in form of added caveats but rather a new approach to tying the two concepts together) I'm sure it could.