Facilitating translation convergence
It's worth testing whether offering points for editing our own translations after we have read the work of other users facilitates translation convergence. Ranking the work of other participants nearly always suggests corrections to me. Sometimes I make those corrections, but most often the revisions remain in my head. The Doulingo project would probably benefit if users actually edited their work. Offering points for doing so might help insure that happens.
I'd like to see Duolingo succeed. For that to happen they need to converge on a single translation that can be sold to website owners. From what I have seen so far of the "best translations" that may turn out to be a hard sell. Even "best translations" often need editing.
My goal has been to come up with a translation that comes closest to my understanding of "standard English" and would be understandable to most readers in the United States. Sometimes this means that I depart from a strict translation--for instance, adding a word not in the original to indicate that a name refers to a sports team or a race track. Americans would not normally know even the names of even the most popular Spanish sports teams or race tracks, just as Spaniards would likely be puzzled by the term Los Angeles Clippers in Spanish translation.
Sometimes viewing other translations helps me see where I have been wrong in my own version, attributing a verb to the wrong noun due to unfamiliar word order, for example. I like to finish with a translation I'm happy with, as fluent and as faithful to the intention of the original as I can get. Even so, I try to come up with a slightly different option. I guess this is contrary to convergent translation.
I think the encouragement for users to edit their translations to make them better comes from the fact that people rate translations. If you finish your translation and see that the "best translation" is actually better than yours, it is in your interest to change any mistakes in yours so that other users will give you a "very good" rating. Of course I imagine in some instances there's a bit of vagueness to what is the "best translation", especially for longer sections of text. No idea what Duolingo does when you have multiple, similar but slightly different Best Translations but I imagine they compete with each other and the more users translate that sentence and vote on it, the closer Duolingo gets to a perfect translation. The sentences you have seen with poor "Best Translations" could have either a low number of translations to draw from, or be a more complex bit of text, which will take longer/ a higher number of translations to iron out the wrinkles.