For those who are curious, there are now seven runners with 4m54s records. Andrewg had a long reign but isn't very active these days and has fallen to like 47th. Nifski is generally considered the strongest runner now.
For most records, there's three categories: tool assisted, emulated, and original hardware. TAS techniques are often hard to replicate with live human inputs, but emulator and original hardware records tend to converge but usually there's movement on the emulator record first.
I imagine not all Technology Assisted Speed-run techniques are even possible on original hardware, even if you’re pumping electrical input directly into the controller port, since the TAS ‘implementation’ or ‘driver’ would depend on the cycle-correctness of the emulator used to record the inputs, the literal physics in terms of analog input, etc....?
What a fascinating topic. Are there any good starting points you can think of to explore it more?
This is pretty awesome, for example: “TAS replay device hidden in NES controller” https://www.reddit.com/r/speedrun/comments/kewjhd/tas_replay...
I think you might conceivably have trouble with running TAS replays on consoles on the Atari 2600, since the inputs on that are just direct pins and afaik, there's no way for the controller to know it's been read. The NES input is a serial read, so a smart controller can use that to sync up playback. That said, I did see an Atari 2600 console verified TAS for Dragster, so there must be a way to sync.
Once you've synchronized the controller playback to the console, the only limitation on playback would be how accurate the emulator is, and/or if there's anything that's not really reproducible.
On the NES, there's no source of real randomness, but some (many?) of the games don't clear all the ram at boot, so you could have residual data if the system hasn't been off for long enough (there's some fun things you can do with cartridge swaps on real hardware); that one is solvable by making sure your TAS goes from an appropriate RAM off state, and that you leave the console off for long enough that RAM decays all the way. Also on the NES, the CPU and PPU have different clock rates (they both run from the master clock, but with different dividers) and there's different ways the clocks can align, which can result in different behavior especially at the limits. My understanding is that most emulators pick one alignment and emulate that, but if your TAS timing is very tricky, you might have to run multiple trials before you got the alignment you want. Clock alignment is a deep topic, but this should get you started [2].
Newer consoles may have more features that are effectively non-deterministic, which could make TAS harder. Addressable ROMs have essentially fixed access times, but optical media, spinning hard drives and flash memories typically don't; this can lead to unpredictable delays that make TAS on real hardware more difficult.
See also TASBot [3] and TAStm32 [4]
[1] https://tasvideos.org/Movies-ace-verified
[2] https://forums.nesdev.org/viewtopic.php?t=6186
I'm not sure if this quote has aged like wine or milk. Possibly both.
For context, we now know the Billy Mitchell score at issue was almost certainly made on an inaccurate version of MAME. It has emulation artifacts, and no sound, which MAME couldn't emulate well back then. Twin Galaxies does accept emulation scores, but they have to be verified in a different way, and they go on a different leaderboard. The way Billy recorded it without the additional emulator verification would have made it trivially easy to play an input file and claim a just-slightly-higher score for pure ego reasons.
For legal reasons, I'm not saying Billy Mitchell actually did this, but I am saying he sued Twin Galaxies when they removed the score at issue.
There's a few other particularly suspect high scores that were purged from Twin Galaxies; notably Todd Rodgers' literally impossible time on 2600 Dragster. That score and Todd's explanation for it is now a speedrunning in-joke.
ljf•9mo ago
Some of the previous discussion is here https://news.ycombinator.com/item?id=8142269