On scoring by ratio to shortest code
I like the philosophy here that you're competing relative to your language, but I don't see it working. The idea seems to be based on assumptions from Anarchy Golf that don't transfer:
- Most challenges have multiple submissions in each language.
- There is a fixed (though large) set of language options.
- The golfing is competitive enough that the best score well-approximates the optimal score.
On PPCG, most submissions are alone in their language and would get the best score of 1. The per-language competition is diluted by frequent challenges and a wide array of languages. Ironically, your proposed scoring method discourages such competition and encourages golfing in your made-up language that only you use.
If you're not first place in your language, you still really can't compare your score to other non-first-places. It depends on how good the first place is, which varies a lot and is outside your control. Maybe an experienced golfer tried really hard for the perfect golf in that language, or maybe not.
If you are first place, improving your answer doesn't improve your score, it just makes others' scores worse, which is strange.