I am trying to simulate a soccer ball. I already know all the maths and physics (magnus effect, etc.).
Right now i have just an understanding problem with velocities:
the ball moves in a certain direction x,y (lets assume its not flying so no 3rd dimension) by having 2 velocities, each for x and y axis.
My problem is now that once the velocity in one direction gets zero (cuz of friction and other forces) the ball still moves on the other axis which then looks strange cuz it then starts to move only horizontaly/verticaly depending on which velocity gets zero.

Maybe someone could enlighten me ....

You should not think of it as separate velocities. The velocity is represented by a vector, and in the most simple case, the acceleration due to friction etc. is pointing in exactly the opposite direction.

Ok, let me get a little more precise:
I got a routine which calculates the ball movement with the typical math/physics equations and using an ODE solver (Runge-Kutta) to computer new ball position and velocities for a time-slice. So what i initially feed it in are, besides some other constant parameters, 6 parameters:
(x,y,z) - positions
(x,y,z) - velocities

It then updates the values for each iteration (which is also dependent on a time-slice).
I should say that these aren't my routines (lazy me), but from a physics-for-game-programmers book.
So far, the routines work great, except for the above problem.

I was thinking that the acceleration/velocity should be only one parameter (which is then damped by friction etc.) and this value is then used to move the ball, so the ball is always moving along its direction until it is stopped.
Somehow i am confused ...

I can provide the code if it will help.

Thanks for your comment.
I will check my routines. I am thinking about if i need an ODE solver or if i should go for an easier way.
I've also to distinguish between ball in air (projectile) and ball on the ground (ground friction) and also need to extend the routines for the ball to bounce (right now i use height_velocity = -height_velociry / 1.32).
The good thing about the routine i am right now using is that it gives me magnus-force, friction and air drag (also wind which i don't need).

You should try to describe your Physics update in terms of vectors (and perhaps points and rotations), without any explicit reference to the coordinates. Then you will probably never see anything special about the directions of the axes.

Thanks for the help.
The mistake i was doing was to check for each individual velocity to stop so i can prevent the ball moving too long (even slow).
What i am doing now is to check the magnitude of the velocity (sqrt(vx*vx+vy*vy+vz*vz)) for stoping the ball if it falls below a certain value.