Billy Mitchell's Direct Feed Recordings are MAME Generated

[JCHarrist]

This is the result of extensive testing shot at 120 FPS with a GoPro Hero Silver 4 of all Donkey Kong transitions on both arcade and MAME .106.

The TKG4-14 board was connected to a Mike's Arcade Nintendo to JAMMA adapter then to a Gonbes 8220 RGB to VGA adapter fed to 2 separate LCD monitors. The small LCD was for reference as the game was being played while 19" LCD was being shot by the GoPro. The goal was to remove the CRT from the equation as the scanlines at these shutter speeds make interpreting results difficult.







The MAME setup is a homemade cabinet with a circa 2010 PC running MAME version .106.





Here is all 4 transitions shot at 720P 120 FPS Narrow and played back side by side at 1/100th speed.



Even at 120 FPS there is still "ghosting" effects that can cause things from a previous or later frame to display faintly but the differences between arcade and MAME are very apparent.

MAME draws the board faster in every case and there are key "fingerprint" frames that are absolutely unique to each platform.


The barrel board key frame is this girder pattern which everyone should be familiar with by now. Arcade will always show 5 girder segments and MAME shows 3 with the "Kong" girder extending all the way to left side of the screen.





On the pie boards the easiest identifier is the two upper long ladders. Arcade draws one at a time while MAME draws them simultaneously.





The elevator board is similar to the pie board in that arcade draws the long ladders one at a time and MAME draws them together. Other obvious differences as well.





On the rivet boards, arcade draws the right side ladders first and MAME draws them all but the lower right at the same time.





These results are 100% repeatable and screen and camera orientation make no significant difference.

[SHLONKY]

This is the result of extensive testing shot at 120 FPS with a GoPro Hero Silver 4 of all Donkey Kong transitions on both arcade and MAME .106.

The TKG4-14 board was connected to a Mike's Arcade Nintendo to JAMMA adapter then to a Gonbes 8220 RGB to VGA adapter fed to 2 separate LCD monitors. The small LCD was for reference as the game was being played while 19" LCD was being shot by the GoPro. The goal was to remove the CRT from the equation as the scanlines at these shutter speeds make interpreting results difficult.

These results are 100% repeatable and screen and camera orientation make no significant difference.

good work! maybe you should get this pinned to the first page!

[YesAffinity]

^^Submit to the TG thread too if you haven't already.  Further documentation and refinement of the primary basis of the claim.  Good work!

[Sqrlmonger]

Before I spend more time on this, how much interest would there be in more statistical analysis of scores?

I have already taken the KongTracker database and imported it into a small c# program to produce a proof of concept:

Code: [Select]

            Other      Billy Mitchell
Board #   Averages    1.047M   1.050M
1            8,724     8,700    8,900
2            8,296     6,700    8,200
3            8,876     9,700   10,500
4            7,081     6,300    7,600
5            8,546     7,900    9,800
6            9,935     9,400    9,700
7            8,272     9,000    9,000
8            7,403     7,400    7,800
9            8,964    10,600   11,000
10          11,714     9,700   12,300
11           8,270     8,400    9,000
12          11,379    10,700   12,600
13           6,659     7,600    6,200
14           8,035     7,700    8,400
15          11,726    10,500   11,400
16           8,114     7,100    7,900
17          11,662    11,200   10,500
18           6,720     7,300    7,400
19          11,653    11,300   10,100
20           7,592     8,300    6,700
21          11,631    11,800   11,400
22           8,275     8,200    9,200
23          11,642    12,300   12,100
24           6,694     7,800    7,500
25          11,776    10,200   10,600
26           7,740     8,100    7,900
27          11,757    10,900   10,600
28           8,023     7,300    8,700
29          11,579    11,400    9,900
30           6,773     7,500    7,000
31          11,490    11,000   12,500
32           7,324     9,000    8,500
33          11,674    11,700   10,900
34           8,221     6,900    9,700
35          11,526    11,000   12,200
36           6,640     7,600    6,700
37          11,447    11,700   11,100
38           7,249     7,600    5,100
39          11,614    10,100   11,000
40           8,207     6,600    9,100
41          11,521    10,600   12,100
42           6,659     8,000    6,800
43          11,663     8,700   12,100
44           7,554     4,900    7,300
45          11,470     9,300   12,200
46           7,863     8,100    8,000
47          11,500    11,400   11,200
48           6,498     6,900    5,800
49          11,286     8,600    8,900
50           7,412     8,700    7,900
51          11,589    11,600   10,900
52           8,321     8,300   10,200
53          11,542    11,100   10,100
54           6,602     7,300    6,600
55          11,510    11,900   11,200
56           7,668     7,400    7,900
57          11,581     8,800    9,600
58           7,804     7,600    9,700
59          11,465    11,700   12,000
60           6,657     8,000    5,900
61          11,384    11,000   11,400
62           7,291     6,500    8,800
63          11,529    10,200   11,000
64           8,102     8,200    7,400
65          11,470    11,300   11,300
66           6,639     7,300    7,400
67          11,404    10,500   11,800
68           7,256     9,700    9,400
69          11,427    11,300   10,800
70           8,014     7,600    5,500
71          11,703    12,000   10,700
72           6,641     7,300    7,400
73          11,464    11,200   10,800
74           7,483     9,200    8,300
75          11,460    11,300   11,300
76           8,146     8,400    8,700
77          11,484     9,100   10,700
78           6,462     6,700    6,100
79          11,379    11,300   10,500
80           7,236     6,700    7,200
81          11,361    11,200   11,100
82           8,064     8,300    8,100
83          11,512    10,500   11,600
84           6,545     7,500    7,100
85          11,224    10,900   12,200
86           7,430    10,200    5,500
87          11,335     8,500   11,700
88           8,059     8,200    7,700
89          11,226     8,800   11,100
90           6,647     7,100    7,100
91          11,358     8,500   11,100
92           7,443     7,900    4,500
93          11,306     9,300   11,300
94           7,970     7,300    9,500
95          11,158     8,900   10,000
96           6,416     7,200    6,900
97          11,240     8,700   10,900
98           7,161     6,600    7,500
99          11,220     9,200   12,500
100          7,879     8,200    9,900
101         11,130     9,000   11,000
102          6,528     6,200    6,700
103         11,092     8,400   10,300
104          7,446     5,400    6,300
105         11,031     8,800    9,800
106          8,220     6,500   10,400
107         11,156     8,400    9,500
108          6,339     7,200    6,700
109         10,929    11,700   11,300
110          7,394     6,100    9,800
111         11,056    10,600        0
112          7,928     8,400        0
113         10,942    10,900        0
114          6,361     7,200        0
115         10,919    12,100        0
116          7,461     8,600        0

Note:  The "Other Averages" above are based solely on 1M point games with killscreens that had complete data available in kongtracker and were not played by Billy Mitchell. I can modify this criteria with minimal effort, so I'm open to suggestions here if other criteria might be more suitable (like excluding games with more modern point pressing techniques for example) and this approach is deemed worthwhile in general.

Would more work on this that could offer some statistical analysis like identifying variance or z-scores as well as looking at boards by type (e.g. barrel, rivet, elevator, pie) be of interest or do you think this is not worth the effort?

[xelnia]

Would more work on this that could offer some statistical analysis like identifying variance or z-scores as well as looking at boards by type (e.g. barrel, rivet, elevator, pie) be of interest or do you think this is not worth the effort?

I'm a big fan of statistical analysis. So, in general, I think it's a worthwhile effort. In regards to Billy's scores specifically, I think it's probably a dead end. Every WR or PB is an outlier, all the top players have some kind of outlier in their top games (Robbie and pies, Dean and barrels, Wes and springs, John and rivets, etc.). So, it's easy to find "something" if you look hard enough. I imagine it would be incredibly difficult, but to get any sort of meaningful statistical analysis of Billy's games, you'd have to quantify the odds of him surviving every decision he makes in his games. As it stands, it doesn't pass the eye test.

[Sqrlmonger]

Would more work on this that could offer some statistical analysis like identifying variance or z-scores as well as looking at boards by type (e.g. barrel, rivet, elevator, pie) be of interest or do you think this is not worth the effort?

I'm a big fan of statistical analysis. So, in general, I think it's a worthwhile effort. In regards to Billy's scores specifically, I think it's probably a dead end. Every WR or PB is an outlier, all the top players have some kind of outlier in their top games (Robbie and pies, Dean and barrels, Wes and springs, John and rivets, etc.). So, it's easy to find "something" if you look hard enough. I imagine it would be incredibly difficult, but to get any sort of meaningful statistical analysis of Billy's games, you'd have to quantify the odds of him surviving every decision he makes in his games. As it stands, it doesn't pass the eye test.

Thanks for the feedback, and I do agree just a single outlier is not even close to sufficient to make any sort of claims re: Billy's scores.  Having said that, if the notion is that MAME was used to construct some of these scores, I do wonder how the random nature of a legitimate game would not be altered in some way. Though again, it may be very difficult to find how it is altered in a way that is statistically significant.

More specifically, the approach I had envisioned was to avoid the need to quantify individual decisions by focusing on the degree to which some board scores were outliers.  Analyzing these outliers in a vacuum is certainly doomed to failure for lack of context, but in comparison to the rate of outliers of peer games of the era we should be able to address that need for context. 

With all of that said, I do lack domain knowledge specific to DK so I may not be aware of specific challenges that you have already encountered in your previous efforts (like those from your link).

edit: I do want to add the specific concern you raise about WR games being outliers by definition, is, I believe, addressed to a large degree by focusing on 1m+ point games as so-called "peer games" all of which, I would think, could be considered outliers.  I concede the pool may need more refinement, but I do think that refinement is possible.

[JCHarrist]

Upon closer examination of the 1.05M tape, it appears to me that the game was paused when the recording was started.

This is the standard Donkey Kong attract mode transition into the title screen. Note that after Jumpman dies and does his death tumble, the attract screen transitions into the title screen where the DONKEY KONG logo is immediately flashing red and blue. It flashes red every other frame 24 times before returning to solid blue for the remainder of the screen.




Now take a look at the start of the 1.05M tape. As the DONKEY KONG logo comes into view, it is solid blue for 42 frames before it starts flashing , and then only flashes 15 times. This would seem to be a clear indication that the game playback was paused when the recording started and unpaused 1.4 seconds later.

[f_symbols]

Is that good? 

 Camp-   It was burn-in on the monitor.

[SHLONKY]

Now take a look at the start of the 1.05M tape. As the DONKEY KONG logo comes into view, it is solid blue for 42 frames before it starts flashing , and then only flashes 15 times. This would seem to be a clear indication that the game playback was paused when the recording started and unpaused 1.4 seconds later.

In my first post here i put forward a few points that would suggest the original recording was not done on vcr.

[maximumsteve]

 Looking at the 2 links, there just isn’t enough to make any kind of evaluations. Here is why...

The Image that is captured is shaky... Then you also see the word play in the corner which is also shaky but NOT a matching shake with the image. And whatever device used to capture to computer would not produce a “play” text when recording.

So I believe we are watching the recording OF a recording of a recording.  What was captured to the computer was played. And what was played was already a copy. So having the original ready at start for the recorder to copy so you see the words play. And the final Recording to computer is a copy of a copy.

[PreacherPatriot76]

Looking at the 2 links, there just isn’t enough to make any kind of evaluations. Here is why...

The Image that is captured is shaky... Then you also see the word play in the corner which is also shaky but NOT a matching shake with the image. And whatever device used to capture to computer would not produce a “play” text when recording.

So I believe we are watching the recording OF a recording of a recording.  What was captured to the computer was played. And what was played was already a copy. So having the original ready at start for the recorder to copy so you see the words play. And the final Recording to computer is a copy of a copy.

That won't cause any changes to how the images are displayed. That all comes down to the monitor used in the cabinet vs. the computer monitor used in MAME gaming. The monitor used in the Donkey Kong cabinet is a Sanyo 5JOUTB22 19-inch 90° that uses video from the display PCB that passes into an image inverter before it hits the display. MAME doesn't do this image inversion and the appearance of the elements that make up an image is all dependent upon the video card in the computer. It will only display the image as the video card processes it and it does it out of order compared to the native cabinet hardware.

Now my two cents on the entire thing. I am not affiliated with anyone in the Donkey Kong community, but I have been a gamer since Pong in the 1970's.

Why are you defending a man who lied about getting the first perfect score in Pac-Man? The first person to do that was Bill Bastable who did it in 1988. There was a newspaper article written up about it and he has the picture of the kill screen.

One thing I do not understand with a community is the defense of a man who is a known liar and a cheat. From the very beginning Billy Mitchell and Walter Day, both of them have been running a scam. With Billy Mitchell funding Walter Day the entire time is a severe conflict of interest. Would you be so accepting of a judge that was also the plaintiff in a court case? No you wouldn't. So why does Billy Mitchell get a pass?

Have you all heard of the Fruit of the Poisonous Tree legal doctrine?

The "fruit of the poisonous tree" doctrine is an offspring of the Exclusionary Rule. The exclusionary rule mandates that evidence obtained from an illegal arrest, unreasonable search, or coercive interrogation must be excluded from trial. Under the fruit of the poisonous tree doctrine, evidence is also excluded from trial if it was gained through evidence uncovered in an illegal arrest, unreasonable search, or coercive interrogation. Like the exclusionary rule, the fruit of the poisonous tree doctrine was established primarily to deter law enforcement from violating rights against unreasonable searches and seizures.

https://legal-dictionary.thefreedictionary.com/Fruit+of+the+Poisonous+Tree

How it applies to Billy Mitchell and this entire mess is that since Walter Day and Billy Mitchell have lied, excluded Billy Bastable from the proper place of being the first person to get a perfect Pac-Man score, and other highly unethical and immoral actions taints all of what both have done. Billy Mitchell funded Twin Galaxies and kept Walter Day afloat and that relationship goes back to the 1980's with Walter's arcade Twin Galaxies going under then being bailed out by Billy Mitchell and other pro-gamers. If that doesn't ring alarms then I don't know what will.

Good day and God bless y'all.

[Mary McManus]

I can hook MAME DK up to a 1986 WG K7000 or 1983 GO7-CBO monitor via direct RGB (no A/V or composite video conversion board involved). Only the Arcade VGA and Jpac from Ultra Marc are used.

The game will run at it's native resolution at its correct 15Khz frequency Anyone try this?

[maximumsteve]

The order differences can be seen in the example used in original accusations. Mame draws barrels and Kong at separate times and arcade does it together.
What some are holding on to is the arcade producing a swipe effect in drawing up the stage and mame doing a clean popup of the stage building. Many are missing the fact that the swipe is produced by the screen and NOT the pcb. And mame's captures are NOT produced from a screen recording but direct snapshots of the sequence from memory. This will be explained in more detail soon.

[xelnia]

The order differences can be seen in the example used in original accusations. Mame draws barrels and Kong at separate times and arcade does it together.
What some are holding on to is the arcade producing a swipe effect in drawing up the stage and mame doing a clean popup of the stage building. Many are missing the fact that the swipe is produced by the screen and NOT the pcb. And mame's captures are NOT produced from a screen recording but direct snapshots of the sequence from memory. This will be explained in more detail soon.

Neither arcade nor MAME build the entire level in one frame/pass...regardless of display method or MAME version. This is absolutely irrefutable. It's the differences in how the platforms build their scenes, step-by-step, that expose the original source. I'll give Carlos the benefit of the doubt and assume that he knows how CRTs work in general, but he clearly doesn't understand how DK hardware works specifically, and how it relates to MAME. At least three unique direct capture methods have been created and dozens of new examples of footage have shown my original assertions to be correct. So far, in 5 weeks, Billy has produced nothing.

[bensweeneyonbass]

snip:

At least three unique direct capture methods have been created and dozens of new examples of footage have shown my original assertions to be correct.

THIS

If I hear any more talk about how the monitor is producing this or that, I'm gonna crap myself.