The unfair ranking system has put me in first place in the global leaderboard, so I feel obligated to take part in this discussion.
Before the optimization challenges were a part of this site, I was placed around the top 50, which took me about six months to achieve.
The last month was a period of fascinating self-improvement, because I have never had anything to do with code-golfing before.
When it comes to the scoring system, I think that one simple constraint – one player - best solution – is enough.
Some languages are just better suited for this task than others, and they definitely should be used.
The current system of keeping multiple solutions from a single player can be kept, so one could see how his solution is placed globally, but the only ranked and scored solution should be the best one for the player.
The interface, next to the current “All” and “Following” tabs, could be extended with “All-best”, and “Following-best”.
Players who have the same score should take the same place in the ranking, which would encourage new players to participate.
For Thor and Paranoid optimization I’ve collectively spent more time than a “very hard” task. The maximum amount of points for an optimization task should be greater than a very hard puzzle reward, because there is always incentive to look for a solution that is just a byte shorter, whereas you never come back to a very hard puzzle after you complete it.
I also suggest adding a “Translator” category - as an extension of the “Translator” achievement - where players would be ranked by the amount of languages they used to solve a task.