Many games just make the aim very inprecise if you run or jump.(More generally, your mode of movement) I think that probably is a much better solution.
As for jumping, i get the idea that jumping is to high because the avatars dont climb. Perhaps jumping should not be too high, but there should be a sort of climbing.

I dont like opponents strafing-hopping around each other for a minute before one of them dies; kills should be swift. But it does not have to be 'realistic' for me.

On my programming
I have tried making games. I probably could make one if i had the time, but , for one, i switch projects too quickly. Although i can make all kinds of aspects of a game, find it very hard to put the pieces together. Because of that, i have no "finished goods".

I have made: (Well, i have tons and tons of useless stuff, and just these might be interesting.)

• Colliding 2d objects consisting springs and triangles.(triangles also have forces between the points.) These simulations where stable. (for suitable timesteps, Euler approximation method.)

• A 2d-bitmap - to point stable colision simulation.Was planning to make a 2d game in which objects consisted of a bitmap and some points in a way that only point-bitmap collisions where necesary.
  I suggested it to Wormux.(Good that they did not adopt it, that version.) Also tried to get a contribution in changing the dreadful angle-math, which was converting back and forth angles-vectors! And I think i got them to change the surface-normal obtaining method.

• I made and tested a quad-tree for colision detection.(One that does not have to state at the top node every time.) Not made a efficient height map with it, though. Have been thinking of using it to cleverly make random generated smaller features in a map. And how to do these features consistently. (The small features can be deleted and re-generated again being the same as it was before.)

• I actually made and tested for a very particular situation a simulation-timestep-varier, that made timesteps powers of two, based on what size the timesteps need to be for stability and precision of the simulation.(In a game, the first one is the main concern)
  It has some problems, for instance, the particular test i gave it did not conserve impulse, because two particles at different timestep-sizes interacting do not exert the same average force.

• This one was easy so far.(Practically wrote itself) I made the structure of lisp with the idea of making a lisp-like language. I am currently trying to work on this. It can already load files, but not execute them. (which would be much harder) It is also wonderfully short. I am planning to not make any atomic functions or types. It wont be Turing Complete by itself, instead it will be able to load c-functions. (Some of these functions would require knowledge of internal workings; while basically means people would have to be supplied atomic functions anyway.) (Have to figure out how to load these from files, rather then having to recompile.) A rather big caveit is still that it can not deal with circular conscells, only trees. I would preferably force only tree-like conscells somehow. Dont know if that can even be done while being reasonable.
  If i got interpreted working, the idea is to make a compiler.(probably compiling to C.) BTW this sounds like lush, i havent looked at the code behind lush, but i am sure it comes from different ideas then those i am using. (Or the manual is misrepresenting it.) I got annoyed with lush when it didnt do what i wanted :P. I guess it is childish, i guess i should try find my fault or show what i dislike to the devs of lush.

Not exactly in my defense, i have read only one book on programming, one on C++. So far, I actually rather dislike OO, probably because i dont understand it properly. It just seems to make stuff too hard on the user often, whining to while compiling often. In most of the projects only use simple derivation.

Anyway, i have a 8:30-17:30 job now,(until 1 nov, but i will probably get another job soon after.) so i dont really have too much time. Also my lispy project does sound very interesting. On games, i think what would be cool, is a GPL game similar to BF2; with vehicles, infantry. (again, preferably not realistic)
There is something that could be turned into this: Rigs Of Rods. My computer is too slow for it though. Of course, the simulation might need adapting for speed.(Not sure if they are set for precision or realism now.) The heightmap would probably need change, and buildings should be more central. (And i also might have an idea to efficiently improve geometric detail.)