I'll be honest - I'm spitballing here, as I havent made an isometric game yet. My instinct would be that the cursor, a visual UI element, only directly interact

I'll be honest - I'm spitballing here, as I havent made an isometric game yet. My instinct would be that the cursor, a visual UI element, only directly interacts in visual space - ie the screen tile is the screen tile, with no height information. When rendering though, you're projecting logical information to visual information. So when you render an object, you fill some struct that correlates screen tiles with last game objects' IDs that were drawn over it. The cursor logic just queries that structure. edit: thats assuming you cant select anything that's been occluded. If you need to select things that've been occluded, lmao i dunno