For my part, I usually prefer the prose. Formulas are easier to read for people who are used to and have practice with mathematical notation, but even then they are easier to read for those people when they are sufficiently complete, and formulas seem to often come with huge leaps and unstated assumptions.
Especially in computer science papers I tend to see extensive use of formulas over prose or - preferably - code or pseudo-code as big warning signs: Often it turns out to mean the author is glossing over a massive amount of hugely important details. E.g. a common problem I saw when studying was papers that would describe processes, but omit any indications of sensible ranges for parameters that were essential to getting good results (I was working on techniques for image processing to improve OCR results), or greatly obscure details of algorithms that would have taken no more lines to write out in working code.