|
Missed opportunity to take one of his Punch Out records instead. How hard could it be?
|
#
¿
Sep 18, 2023 12:59
|
|
|
|
| # ¿ May 17, 2024 12:07 |
|
|
Zamujasa posted:it can make sense in very limited contexts, like offering a bounty if someone can come up with a reproducible time-saving glitch; iirc there's still an unclaimed $1k for anyone who can figure out how to reproduce the legendary mario 64 ttc upwarp, the one speculated to be a random fuckin cosmic ray Soft errors/bitflips are fun. There's a famous study that estimated soft error rates in google servers to be somewhere around 10% per memory module and year in the late 2000s. They're common enough that some weird glitch has assuredly happened in a speedrun due to a stray alpha particle over the years, but rare enough the chances of any particular weird glitch in a particular speedrun attempt being caused by a bitflip rather than a logic error or hardware fault are effectively 0. That's the sort of probability combination that humans aren't great at working with. Programmers have an unfortunate tendency to blame weird unexplained behavior on soft errors as a result, even when we shouldn't. Working on a big cloud compute system I have probably seen a cosmic ray error in a crash log, but the number of errors that we chalked up to bitflips because we didn't understand how they could logically happen is definitely much higher than the number of likely impactful bitflips in our system over the years. On an N64 with no error correction I'd still say a hard error is more likely than a soft error, even had it not been a unit where you have to shim the cartridge in place. It's also really difficult to rule out a repeatable logic error, though. Bugs can get weird, and proving their absence is hard (see: Atari Dragster).
|
|
#
¿
Jan 19, 2024 06:59
|
|
|
Rubellavator posted:I don't remember when this was ever the case. The idea for a long time was that there simply weren't any cheaters because why would you? And then people discovered spliced oot runs. Not sure that's really true. I know there was a lot of angry posting about cheating after the Morimoto SMB3 TAS video got traction in 2003. I recall the stolen valor of TAS videos (and nesvideos) getting posted and taken as legit runs by the people watching was cited by a few runners as part of their reason for quitting the hobby, I think including the then current Super Metroid 100% WR holder (or, well, segmented-run-on-SDA holder). And that's without anyone intentionally attempting to pass off TASes as real time speedruns! People were well aware of the ability to lie with video (or demo files!), runners felt threatened by even the possibility, and best I remember SDA took preemptive steps to prevent it like asking for sent-in recordings to be bookended by non-console footage in addition to their usual (slow, often criticized) human peer verification before publication. Splicing was I think initially not so much a concern since segmented runs in the mold of QDQ were the any% prestige category so indeed why would you? but various other forms of cheating were definitely a common concern. The belief on SDA was just that their process was sufficiently robust to filter out any attempts. Then increased access to recording tools and online streaming started clashing with the slowness and quality requirements of SDA's robust process, then it turned out spliced single-segment runs got past SDA's reviews for years, so it goes, etc. Anyhow, people have been simultaneously concerned about stolen gaming valor and unable to identify it at least since Activision went ahead and certified the Dragster 5.51s record in 1982.
|
|
#
¿
Feb 25, 2024 06:36
|
|