Billy Mitchell's Direct Feed Recordings are MAME Generated

[YesAffinity]

All right, here's the first stab:

Arcade, with composite out via JROK encoder
VCR composite in, recording to VHS tape
VHS playback via Iscan VP30, 640x480 @ 60hz locked
Iscan VP30 HDMI out to Avermedia C127 capture card
Avermedia C127 capture to hard drive, 640x480 @ 60fps

I first looked at the 60 fps video which should be uploaded to youtube by 1 a.m., and can be found here: https://youtu.be/qk-RW9K_vF0.  There was clear presence of the sliding door effect that is expected from true arcade hardware, but is shown to not be present in Billy's videos.

I then re-encoded to 30fps, using MPEG-2 encoding.  This would likely have been the compression method employed when capturing from the assumed VHS tapes, circa 2007.  I used Handbrake to transcode the video, and turned off all superfluous options, such as deinterlacing.  Below are frame-by-frame progressions at the lead ins to different boards.  Although there are obviously less frames, and the sliding door effect is less obvious, it can still be seen as explained by Jeremy and Mr. Sock Master.  The sideways view is exactly how it is recorded from the cabinet.  I also rotated them for easier viewing.

















Conclusion of this effort: as best as I'm able to replicate what captured video from an arcade cabinet at the time and with some assumptions as to how the video was being recorded, the results support the assertions.

Capturing directly from MAME hardware is proving significantly more challenging, at least for what I have immediately available to me.  My JPAC/ArcadeVGA MAME rig will not output video simultaneously to the arcade monitor and the composite output.  It's one or the other.  I am going to try outputting to composite from the rig, recording that to VHS, and playing with video being sent to a LCD screen piggy-backed off the VCR.  More on that tomorrow.

[bh]

Crossposting from TG for the sake of completness:

As I mentioned earlier, the RNG is unusually kind to Billy in the 1,050,200 point WR game. On barrel stages he had 105 blue barrel smashes and 38 flame smashes for a total of 78,500 points, a >99th percentile performance for the RNG.

At my request Jeremy Young provided me with transcripts of Wes Copeland's 1,170,500 and Robbie Lakeman's 1,172,100. Over the entire game, Wes had 99 blue smashes and 85 flame smashes for 91,000 points. This puts him in the 62nd percentile. Robbie made a paltry 95 blue and 72 flame smashes for a measly 80,500 points. This is a 21st percentile RNG result. Tough luck, Lakeman!

This is interesting because it shows that top players world record with bad RNG performance. Billy knocked down the record with an amazing RNG. This is exactly the sort of thing you'd expect to see with a segmented run. Taken with the video evidence of MAME usage it's downright incriminating.

I'm slowly working on the 1,047,200 game, but I expect to see another super lucky RNG. Does a full tape of the 1,062,800 exist? I'm not picking over video artifacts, so really any copy would do.

My next step is to get more DK stats. Then I'll do a regression on either smash points vs. total score or barrel smashes vs. total score. Next I'll determine if any scores are outliers in both number of barrels AND points per barrel.

[Scoundrl]

I might have some footage of Billy playing at the KO3. It would be interesting to see  these numbers vs known good numbers.

[YesAffinity]

Taking a different approach with the same VHS recording as the pictures and video above, I dug out an old USB composite/s-video capture device.  Using the same VHS recording of a direct feed composite output from DK arcade hardware, I then played back directly from the VCR to the USB capture device.  The video was encoded with MPEG-2, and captured at 29.97 fps.  In my mind, this is probably the closest to how a VHS recorded arcade feed would have been captured to a PC and then uploaded.  This does the transcoding direct from the VCR to the PC.

Same result.  You can see the screens transitioning via the "sliding door" effect.



Fully video here, if anyone wants it:

[Sock Master]

Taking a different approach with the same VHS recording as the pictures and video above, I dug out an old USB composite/s-video capture device.  Using the same VHS recording of a direct feed composite output from DK arcade hardware, I then played back directly from the VCR to the USB capture device.  The video was encoded with MPEG-2, and captured at 29.97 fps.  In my mind, this is probably the closest to how a VHS recorded arcade feed would have been captured to a PC and then uploaded.  This does the transcoding direct from the VCR to the PC.

Same result.  You can see the screens transitioning via the "sliding door" effect.



Fully video here, if anyone wants it:

Wow. just wow. I can't express how ridiculously awesome it is that you went through all this trouble to create a duplication of the process and steps that it would have taken to get DK direct feed, onto actual VHS tape, then onto Youtube.

Just watching it with all the VHS wobbles and blurriness... it's just awesome to see.  I can't wait to see the MAME equivalent

[Scoundrl]

The videos I have do not show any of Billy's game play. It was a nice trip down memory lane watching all the camaraderie in Denver though.

[LMDAVE]

I played next to Billy at either Kong off 2 or 3, cant remember, but I do remember him getting a 82K level three on his first man. Now I know what you're saying, that's just level 3. But, to get 80K+ on level 3 you have to have the DK skills to pull it off. Getting large scores in DK is applying your use of skills and control of mario, but have to know the tricks and "how to" maximize your score, then you have to have the concentration to pull it off through 21 levels.

Do I think Billy has the skills to get a 1.1M eventually? Yeah, if he applied himself to trying to get back into it, probably if he follows the blue print that is out there now. I don't think he had that blue print before. That's what makes the probability and high blue hit values on the 1.062M (1.125 pace game) so hard to believe. I think Steve Wiebe would be able to cross 1.1M also if he spent some time on DKF or watched archived games and not just playing his same pattern game.

So, even if Billy did get back into DK to show a high score, still doesn't mean he didn't pass off MAME tapes in the past as arcade.

[YesAffinity]

Ok, well here we go.  I've taken a variety of different approaches with MAME, again attempting to recreate likely scenarios of what we see in the Billy M videos in question.  MAME .122 was used, with no rotation.  All joined images are frame-by-frame steppings, using VLC.

First, MAME PC is outputting composite direct feed to a VCR for recording.  I recorded the LCD screen connected to the VCR with my phone camera, at 640x480/30fp, while the VCR was recording.  MediaInfo identifies this video as being NTSC.

The pictures represent how the image appeared while standing in front of the screen.  This gives the appearance that the image is being drawn from the left side of the screen to the right side of the screen (or from the top of the play field to the bottom of the play field).  As we know, this is not consistent with how arcade hardware produces the image.

The video from which these frame grabs were taken:

 

[YesAffinity]

Phone camera live streams MAME CRT at 640x480/30.  The stream is captured by OBS at 60fps, to avoid OBS bottlenecking or having influence on the video.  The captured stream video is then transcoded w/ VLC at the video resolution, framerate and bitrate of the Billy M 1.05 video currently on youtube (https://www.youtube.com/watch?v=Y0ZKEGZpggI).  This was an attempt to:

1)   Somewhat recreate a camcorder recording
2)   Transcode it to the same format of the above linked video

Full video that screen grabs were taken from: https://www.youtube.com/watch?v=Ywu8y89CpeI&t=162s

Note the first grouping of frames is NOT transcoded, and is 60fps capture.  Interestingly, two of the frames absolutely display for a period of two frames (do not change from one frame to the next).

[YesAffinity]

The screens you see being externally recorded in the previous 2 posts were also direct fed via composite output from my MAME PC.  The direct feed was recorded to a VHS tape.  The VCR was then played back to a Happauge PVR USB capture device.  The stream was captured as MPEG-2, 720x480 @ 29.97 fps.

Full video that frame grabs were taken from:

[YesAffinity]

Lastly, I re-visited what my webcam had recorded from arcade play.  This is another effort to replicate a camcorder recording.  My webcam captured the DK arcade screen at 720x480/30fps.  This file was then transcoded to the resolution, bitrate, framerate and compression type of this video of Billy’s 1.05M run (https://www.youtube.com/watch?v=Y0ZKEGZpggI).  Although there are obviously many frames being outright omitted through this process, we still clearly see the sliding door effect.

original, straight from web cam to PC footage: https://youtu.be/5HftSTw9wC8

video file re-encoded: https://youtu.be/PUb7H0DMv5o

[YesAffinity]

That's about the best I can do with the equipment immediately available to me.  I'm still puzzling over the scenario that would marry MAME play, or even more challenging, MAME playback, with a live performance, direct feed, and all the other variables being contemplated as part of the larger picture.  For MAME play alone, Mike Haaland confirmed that both his JAM2NIN adapter and Nintendo cab harness-to-edge connector have been available since circa 2001.  This is one potential avenue for a 60-in-1 or other MAME device to have come into play in the equation.

Unfortunately, despite the video evidence, which represents the bullet in a murder case, we still need to find the gun.

At any rate, I hope my efforts prove useful in providing additional data points against which to analyze and verify.

I will upload the videos from which all of the above screen grabs came from, and updating the posts above with the links.

[Mary McManus]

I'm willing to bet the picture quality generated from MAME via the composite out from the video card  on the computer when shown on a regular TV was not as crisp and was fuzzy around the edges. It looks "O.K." but  quite noticeable when looking at the same pic from the  original hardware on an RGB monitor.  I know I have done this already.
 

[YesAffinity]

At Jace Hall's request, I also recorded video of MAME/CRT at 1080p60 with my phone camera.  No post processing to the video file unless youtube upload tweaks it.

I took the liberty of rotating the images included in the below collages.  MAME .inp recording was done with MAME rotation option set to '0'.  In reality, the score is on the viewers left and bottom girder on the viewers right, when standing in front of the monitor, like such:



Full video:

[YesAffinity]

For posterity, here are some more uploads, of "less processed" recordings or captures:

DK MAME composite direct feed to VCR - 60fps capture from VHS playback: https://youtu.be/Jrkaw9qhiTI

DK Arcade Sanyo 20EZ recorded with phone camera at 1080p60: https://www.youtube.com/watch?v=18lx9_cHsK4

DK MAME CRT recorded with phone at 720p30, no post processing: https://www.youtube.com/watch?v=0upbIdGW3XA