Do you think it's more effective to use displacement mapping to customise characters or to make a seperate model for each character?Like if I want my character to have adjustable levels of muscularity,I can use a displacement map to do it,but I would have to store different meshes created when using the map on the original mesh,depending on the muscularity level for each player that does it,basically resulting in a bunch of new meshes just like the second option.

Displacement mapping ends up costing more in cycles, as you are for every draw call, for every vertex reapplying the delta to the mesh. You don't need to store a different mesh for each object *because* of this - you just change the displacement map texture with each draw call.

I found it cheaper to store the transformed positions/normals/tangents in a separate vertex buffer space per character as the displacement mapping cost does build up. However when you take LOD into consideration, you can save a fair bit of memory if the different LODs all map back to the same displacement map.