Thank you for posting this for discussion. When I first viewed this run I got the impression that something was "off." There are many instances where it appears that inputs are being entered much faster than I'd expect from a controller. I suspected that either A) emulation had been slowed down at certain points to allow for TAS-like inputs, B.) keyboard inputs had been mapped to the controller in such a way that multiple inputs were registering from one button press or C) that an actual keyboard was being used to allow inputs not normally possible with controller.

One overall red flag is that in the audio commentary, no button presses can be heard at all. Even w/ a headset I would expect to overhear the occasional button presses in spots where there is no vocal commentary. This suggested to me that audio was recorded over the video after the fact, which implies that the actual gameplay was not live and could have been tweaked with emulator tricks beforehand, outside of real-time.

The gameplay itself raises questions in several spots. The first is a very difficult backwards shot at 0:48 that the runner does not appear to be very surprised by. There is a frame-perfect avoidance of the hopper (after the 1st Kraid clip) at 2:00. I've never seen this approach of purely AVOIDING the hopper before, as the logical approach would be to shoot the hopper and gain a damage-delay to ensure some degree of success. There are then the frame-perfect tricks of the Ripper boost at 6:06, post-ice quickmorph jump at 9:37, two very tricky and unlikely metroid avoidances/freezes at 13:16 and 13:19, and a perfect escape door at 16:15. While none of these is impossible on their own, the combination of seeing all of these tricks in one run, again w/out much surprise expressed by the runner, is questionable and suggest that perhaps emulation has been slowed down to get a successful result at these difficult junctions, which is essentially TASing, OR that something has been done to allow faster inputs. I see the very best runners pull these tricks of all the time, but seeing them all together is what strikes me as odd. Similarly, the dramatic and very unlikely sequence of events in the bridge skip at 11:50 suggests the same.

The lynchpin of this really is though, the Kraid lock at 4:06. Locking up Kraid so that he cannot move is a TAS trick famously performed by Lord Tom, and requires very specific inputs. My understanding, and someone please correct me if I'm wrong, is that this requires a very specific pattern of shots, timed to one shot for every X frames, to achieve. It is not currently known to be possible for a human player with a controller. As this is a completely revolutionary technique, I would want to see the runner show live video of this with the screen and controller in view so we can confirm it is legitimately performed with a controller and without TAS-assistance. Here is a case of extraordinary claims requiring extraordinary evidence, or at least something better than a screen dump.

There were some comments made in the discord channel that also raised eyebrows. One was the apparent admission that 2 (presumably duplicate) inputs were being mapped to one button. I would think that this was grounds for disqualification on its face. I also noted that the runner said they were using EmuHawk for emulation. I'm not familiar with this particular emulator and would prefer to see runs done on FCEUX, which is universally considered to have the most faithful emulation. Real Nintendo hardware would of course be ideal

I do believe this runner has a grasp of some of the skills necessary to pull off this route, but there are too many questions about the execution right now for me to believe this is 100% authentic. The Kraid-lock especially is just impossible without TAS assistance as far as anyone knows, and really must be demonstrated under circumstances that can verify it's being done in real-time without TAS-like emulator functions. The sequence of multiple Very Difficult and/or Lucky stunts is also suspicious. I would invite Freya to stream some runs live, with both the TV/monitor and controller in view of the camera, to demonstrate (and allow others to verify) that these difficult tricks are being performed legitimately.

ETA: I forgot to mention that because of the suspicion of abnormal keyboard-to-controller mapping, that any live runs conducted would need to clearly show the input map settings, for verification purposes.

I went in a tested a number of things for I-Frames and shot frequency.

Normal enemies have exactly 10 frames of invulnerability. I tested on the Red Rippers in Brinstar, since they have a bajillion HP. The only guaranteed way to stunlock is to hit them on every 11th frame. Hitting one frame early would cause them to get a substantial period to move, hitting one frame late had unusual results, but it usually resulted in one frame of movement. Occasionally a stunlock could occur on the 12th frame, but it seemed position based at best.

As for bosses, their invulnerability period is only 3 frames, requiring them to be hit on every 4th frame. Same as with normal enemies, one frame early, and they won't take damage or go back into stunlock, and one frame late gives them one frame to move. I tested this on Ridley, though it likely works the same for Kraid.

To hit exactly on every 4th frame you'd have to be mashing at 15hz, no more, no less. With 2 fingers hitting the same button, you'd need 7.5hz per finger. This is consistent with Lord Tom's findings. Now, 7.5hz isn't hard to do on it's own, but keeping it consistently at that speed is unlikely over long periods. 15hz is out of reach for most people, unless mapping 2 buttons to B, which Freya claimed was the case with the 3DS VC version, since I'm guessing both B and A are mapped twice.

As for shot frequency, for the in-game auto-fire. The first shot happens, then there's a noticeable delay, then the auto-fire kicks in, firing once every 8 frames (7.5hz). Using 2 buttons mapped to B, and holding both of them had the exact same result. It would be possible to alternate, but the game is picky and if you hit either button even a frame before letting go of the other, it won't register the input.

As for emulator choice, FCEUX is not the most accurate. You can reference this page: http://tasvideos.org/EmulatorResources/NESAccuracyTests.html

I personally use BizHawk with the NesHawk core (likely what Freya is using). Thus far I've noticed no anomalies vs. FCEUX, and to be honest, it feels like there's less input lag, but I could be imagining that. BizHawk also has an input viewer that can be toggled.

So no, she doesn't need to start running on FCEUX.

Thanks for the emulator information, DD, that's a super useful link. Looks like BizHawk is an ideal choice actually, I had just never heard of it until now. I'm glad that this can be narrowed down to just the inputs. Awesome stun lock research, dude. I think from other info we've gleaned from discord, this is shaping up to be about 1) use of turbo and 2) mapping of multiple inputs to one button.

I've heard Freya is interested in continuing to run the game, and I hope she will stick with it. DD: What specific recommendations do you have for Freya that will help put this incident behind her and allow her to move on to comfortably submitting future runs?

I think it was already said in Discord that she was using Turbo, by her friend. So long as she doesn't use that she should mostly be fine. She should also check and make sure "Allow L/R + U/D" is unchecked, as that will make it impossible for those specific actions.

Her friend also said she'll have access to a NES at some point, so hopefully that happens soon.