Imagine the worst case scenario, a fighting game where each moves' frames are different in size. Some kicks have an exceptionally far reach and some even further (think Dhalsim)

![enter image description here][1]

The bounding box may be very different in each frame and so are the frame's size and the relative position point. There is no mystic way to do it. Normally to handle complex sprites' animations, one of the teams' members uses a tool to define a bounding rectangle around each frame and place a relative positioning dot in each frame. This data is saved to an xml file often but is created with a specialized tool. [1]: https://i.sstatic.net/1QLmG.png

Imagine the worst case scenario, a fighting game where each moves' frames are different in size. Some kicks have an exceptionally far reach and some even further (think Dhalsim)

The bounding box may be very different in each frame and so are the frame's size and the relative position point. There is no mystic way to do it. Normally to handle complex sprites' animations, one of the teams' members uses a tool to define a bounding rectangle around each frame and place a relative positioning dot in each frame. This data is saved to an xml file often but is created with a specialized tool.