infomon
Smash Scientist
Hmmmmm.... tough question, dj.
Before your post I would have just assumed that clash bubbles are not a reliable way to determine hitbox overlap, for the conclusion you've reached: they appear at strange places, lol.
However, looking at your images, I think you've convinced me you were doing it right, lol. First I'll assert that when two hitboxes overlap, either one or both may clash causing a clash-bubble. (Actually, a hitbox can clash against a shield, starman-invincible hurtboxes, etc.) Also, I think the clash-bubble may appear anywhere along the overlap.
But your images further suggest, IMO, that the clash occurs along one of the edges of this overlap. In the Lucas image, I think the clash bubble is either at the rightmost edge of Snake's hitbox, or the rightmost edge of Lucas' hitbox. In the Sonic image..... yeah, haven't you ever hit someone backwards with Sonic's Fsmash? Well it happens You can't do it to a standing opponent because their hurtboxes won't want to overlap, but you can walk someone into Sonic while he's charging the Fsmash and see that the hitbox is as disjoint (toward Sonic's back) as that image suggests.
It's still complicated to determine where exactly the clash bubble will appear, and from there solve the inverse problem: given position of the clash bubble, determine where the hitboxes and/or hurtboxes are. (Hmmmm, neat tangential idea: using starman-invincibility might help you get images of characters' hurtboxes, since they'll remain in place while the clash bubble hits them)
First problem: are there any attacks with hitboxes that are not [convex polygons]? Or if there's a single attack with multiple hitboxes that count as the same one (ie. clashing with one counts as clashing with the other). Those would be tricky.
But let's assume that it's two convex polygon hitboxes overlap and it's a clash. The intersection of the two polygons is itself a convex polygon (hmmmmm wait is that correct? anyone have a theorem for me?? yaaay geometry ) Regardless, I assert that the clash bubble appears along the exterior of that intersection polygon... the problem is determining where. Like, even in the Snake vs. Lucas image, maybe the hitboxes are more disjoint than shown, and the bubble's along the top of the intersection polygon (note that it's above Lucas' stick).
*shrug* I dunno how we'd be able to figure that out..... but at least we can use them to give us bounds on where the hitboxes overlap.
Sorry for wall-of-rambling-text
Edit: What exactly do you mean by the "disjoint" of an attack? I guess you want to know the amount that the hitbox separates from the hurtbox? This is equivalent to just wondering where the hitboxes and hurtboxes are, in all of a character's different states/frames. The best way IMO would be to make a hitbox/hurtbox hack..... which would require some serious dev.work, but it's been suggested so much that I wonder if one of the haxx0r devs are trying to put this together? Once I get some time to play with the hax I can try my hand at some Brawl-disassembly......... mwahaa.....
Before your post I would have just assumed that clash bubbles are not a reliable way to determine hitbox overlap, for the conclusion you've reached: they appear at strange places, lol.
However, looking at your images, I think you've convinced me you were doing it right, lol. First I'll assert that when two hitboxes overlap, either one or both may clash causing a clash-bubble. (Actually, a hitbox can clash against a shield, starman-invincible hurtboxes, etc.) Also, I think the clash-bubble may appear anywhere along the overlap.
But your images further suggest, IMO, that the clash occurs along one of the edges of this overlap. In the Lucas image, I think the clash bubble is either at the rightmost edge of Snake's hitbox, or the rightmost edge of Lucas' hitbox. In the Sonic image..... yeah, haven't you ever hit someone backwards with Sonic's Fsmash? Well it happens You can't do it to a standing opponent because their hurtboxes won't want to overlap, but you can walk someone into Sonic while he's charging the Fsmash and see that the hitbox is as disjoint (toward Sonic's back) as that image suggests.
It's still complicated to determine where exactly the clash bubble will appear, and from there solve the inverse problem: given position of the clash bubble, determine where the hitboxes and/or hurtboxes are. (Hmmmm, neat tangential idea: using starman-invincibility might help you get images of characters' hurtboxes, since they'll remain in place while the clash bubble hits them)
First problem: are there any attacks with hitboxes that are not [convex polygons]? Or if there's a single attack with multiple hitboxes that count as the same one (ie. clashing with one counts as clashing with the other). Those would be tricky.
But let's assume that it's two convex polygon hitboxes overlap and it's a clash. The intersection of the two polygons is itself a convex polygon (hmmmmm wait is that correct? anyone have a theorem for me?? yaaay geometry ) Regardless, I assert that the clash bubble appears along the exterior of that intersection polygon... the problem is determining where. Like, even in the Snake vs. Lucas image, maybe the hitboxes are more disjoint than shown, and the bubble's along the top of the intersection polygon (note that it's above Lucas' stick).
*shrug* I dunno how we'd be able to figure that out..... but at least we can use them to give us bounds on where the hitboxes overlap.
Sorry for wall-of-rambling-text
Edit: What exactly do you mean by the "disjoint" of an attack? I guess you want to know the amount that the hitbox separates from the hurtbox? This is equivalent to just wondering where the hitboxes and hurtboxes are, in all of a character's different states/frames. The best way IMO would be to make a hitbox/hurtbox hack..... which would require some serious dev.work, but it's been suggested so much that I wonder if one of the haxx0r devs are trying to put this together? Once I get some time to play with the hax I can try my hand at some Brawl-disassembly......... mwahaa.....