The problem I find with collab and splitting 50/50 is that the programmer since he is the tester and coder tends to make the game his way, Like I tell him no double jump, yet he ads it, I tell him I want him to move faster and he doesn't do it. It's just a big pain to work with someone who doesn't follow the idea that we defined from the beggining, yet hopefully I will be able to work with someone who is.
If you're trying to micromanage the coder, or get him to code 'to spec', it's not really a collaboration, it's more of a manager-subordinate relationship.
On small projects, you've just got to work with a coder that you can trust to make things 'feel right'. Sometimes they'll have very good technical reasons for changing certain things. With your example, he may have already tried making the character faster - and realised that allows him to jump too far, or not see far enough ahead - then reverted it.
Also, be realistic. You won't find a coder that'll make an epic MMORPG on the basis of some concept art and a design doc... But if you've got a good idea for a relatively simple project (e.g. 2D platformer or puzzle game), and some pretty art, you'll probably find a competent programmer without too much trouble.
(As a coder, finding good artists to work on a collaboration is rather tough, unless you know one IRL, or are willing to work with someone very inexperienced. Those with the skills/talent generally don't want X% of possible future revenue, they want cold hard cash up front!)