@Guga: don't worry, there's been a shitstorm on IRC, which we might paste into the thread. (I'm mostly arguing against the category, rather than the lack of a verification movie, which is fixable. It is, of course, important to check that the SRAM is legitimate, especially because there have been past submissions in which it wasn't *glares at RingRush*, but that's something that's irrelevant to the main thrust of my point.)
EDIT: Swordless didn't want the IRC conversation posted, so I'll re-make my case in the thread here instead. I'm complaining about the category used by this run.
Basically, what's happening here is that the run's copying information from another save file, and using it to complete the game, while apparently aiming for speed.
First, if this were actually allowed within the game mechanics, it'd clearly be rejected as a run. It'd be the equivalent of loading up an SM64 file that was already completed, running up to the top of the tower, and beating Bowser to complete the game. Or loading up a Pokémon save file that was already complete, flying to the Elite Four, and beating the Champion with a squad that had already been ground up to level 100 with competitively relevant moves. OK, so technically you could call this "newgame+ completion", but it's not interesting; it skips the part of the game that you actually care about. And in particular, you can't consider this a sensible category for a speedrun, any more than "fastest death" is a sensible category; it's not even playing the majority of the game.
In this case, a similar effect is happening via glitch. But we can look at similar glitches from other games, both tool-assisted and unassisted:
- In the Super Metroid publication [1978] SNES Super Metroid "X-Ray glitch" by Cpadolf in 21:25.12, a glitch that relies on accessing uninitialized RAM is used. In order to set up the contents of the RAM, the run uses a second save file, and loads it and dies just before performing the glitch. Now, this isn't exactly the same sort of glitch as is used in this run, but noticeably, the second save file is produced during the run, via use of an in-game save, and the time it takes to produce that save is counted against the run. (And as this is the "fastest at all costs" category for Super Metroid, a very heavily competed game, it would make no sense to spend that time if it wasn't necessary.) The run is also considerably more entertaining completing the game on its own merits, without having to produce a second save file via magic (or dirty SRAM).
- Donkey Kong 64 (which is sadly nonTASable because Mupen) has a glitch widely known as the "Funky Weapons Glitch", which is very similar to what's done in this submission: what FWG allows you to do is to start a new game, but with information from what you were doing immediately preceding starting that game to carry over. If you have a save file that's already complete, you can load and quit that save file while doing the glitch in order to have your newly started save file already have keys 3 and 8 (which is what's required to unlock the final boss), then merely buy moves and go defeat the final boss. The unassisted glitched speedruns of DK64 do not do this; instead, they copy over information from the game's multiplayer mode, in order to start with the weapons that are available in multiplayer, and still have to play through the game itself as a result (although that game is so heavily sequence-broken by now that it doesn't look much like a normal playthrough). And the reason, clearly, is that copying over completion from another file, even if it's done via glitch, defeats the point of completing the game; you already did the work of completing the game, but you're not counting it against the timing, so you're basically just misleading people about the clear time.
- The Legend of Zelda: Skyward Sword has a glitch (that also exists in several other Zelda games) known as the Back in Time glitch that makes the title screen playable. However, unlike other Zelda games, it's possible to use menus under Back in Time to the extent that it's possible to load a game from the title screen, and save the game from an in-game save point inside the (glitched to be playable) title screen, simultaneously. This causes the game to get confused and merge some information between the two save files. If the second save file is one in which much of the game has already been completed, it's possible to copy over cutscenes like this, which is valuable because cutscenes can have effects on the game (e.g. unlocking areas much later in the game), and a large proportion of the game can be skipped like that. However, Skyward Sword runners restrict themselves to using save files that they created during the run itself (which is still useful to unlock a couple of specific cutscenes, and to confuse the game as to Link's X/Y position as a particularly convoluted method of glitching through walls; interestingly, these uses of the glitch are typically called "speedrun-legal"). Again, this is for the same reasons as before; if you've already done the work of completing the game, and are merely copying your completion record across (whether via an explicit "copy saves" command, or via glitch), you're hiding a lot of time that's part of completing the run from the savefile.
So, what we've established so far is:
- Newgame+ is not an interesting speedrun category in games that don't actually have a newgame+; at most, it corresponds to "load a savefile in which the game is already complete and go beat the final boss again", and it's even less interesting in games with no final boss.
- Using information from another save file via glitch is no different category-wise from using it via a mechanism explicitly present in the game.
The conclusion here is that this submission is, if considered as a speedrun, lying about its time. It doesn't complete Diddy Kong Racing in less than 5 minutes; it takes an already-completed save file, and copies across information from it via glitch, leading to a game that's completed. (Such an apparent speedrun submission that lies about its time has happened before:
#3519: RingRush's PSX Croc: Legend of the Gobbos "glitched" in 01:10.12 was an April Fool's video that worked via glitching the graphics so that it was unclear what was being done, and simply exited the save file and loaded a 100% save file near the end of the game while nobody could see that that was what was happening. It was a test of the system to see if the judge would catch the cheating.) This run is more honest about how the fast game end is accomplished, and is more interesting because it uses a glitch rather than loading the save file, but the time shown as a "completion time" is just as fraudulent. As such, this method of completing the game, because it requires the game to already be completed and then to do something extra, is clearly slower than just doing it normally, and thus isn't suitable for a speedrun. (If people that "fastest recompletions of a game using an already complete save file" is an interesting speedrun category, then they can feel free to make such runs for other games; it strikes me as hauntingly similar to the Speed Deaths Archive, and that was pretty fun too, with some interesting submissions, but the joke wore itself out eventually.)
(Another example is the run linked from the postscript to the publication comments of
[1978] SNES Super Metroid "X-Ray glitch" by Cpadolf in 21:25.12; I'm not aware of the exact circumstances of that run, but discussions with others lead me to believe that it glitches back to the start of the game from a file that's already 100%, and completes the game faster from the "start of the game" because it already has all the items. But the game has of course been played up to then in order to obtain all the items in the first place, so it makes no sense to consider this some sort of "fastest Super Metroid completion".)
The other potential category that this run could be considered in would be as a glitchfest; although the fact that it completes the game is pretty much irrelevant, it does show off an interesting glitch. However, glitchfests should show off more of the game's glitches than this one does; a glitchfest that just shows off one glitch is about as interesting as a speedrun that only completes one level. That is, not without merit, and interesting to watch, but not worthy of publication.
As such, I conclude: this isn't a real category, it's more of a WIP. Merely showing off one glitch, and not using the fastest route, does not a publishable TAS make.