Time accuracy
6 years ago
Sweden

Hello there,

One fact has been bugging me and I have to ask. Why do we track time as low as milliseconds(ms) becuase we CAN NOT verify that. If you record your gameplay at 60fps, which means there is a new frame every 16.6ms the RTA time shown in the gameplay would be incremented 16ms every frame. That means we don't know when the run was actually finished.

As an example, the current record in Any%: 4m 56s 462ms

The ms part of the time may really be anything from (446-462)ms, we don't know.

Would appreciate if someone shed some light on this for me.

Canada

Okay, so you're saying that just because framerate can't be precise enough to calculate milliseconds by the unit, we should ditch that altogether and calculate by entire seconds, which is even less precise?

This is silly, and a non-issue.

Edited by the author 6 years ago
Sweden

That's not what I'm saying, maybe measure at a 10th of a second? or 100th of a second? But not a 1000th of a second.

Edited by the author 6 years ago
Ontario, Canada

Actually, we have a spreadsheet for timing runs based on the full fractional framerate of the NES, based on bowser hammers. Accurate timing is only necessary for times around the 5:00.3XX and below.

Colorado, USA

@Odezza We can and we DO verify upper tier runs to the thousandth of a second. We have spreadsheets and screenshots. How it works is when you enter the final room of 8-4, the patterns with Bowser's movement, his hammers, his fireballs and the podoboo is different on every frame. However, if you enter that final room on a specific frame, the pattern will always be the same. What has been done is screenshots of each pattern have been collected roughly from the 4:56-5:02 range. Each frame when you enter that final room gives you a different pattern. We reference the screenshots and/or spreadsheets and are able to get an exact time based on the patterns.

I've always wondered, what if you do fast accel in the last room, therefore saving 1 frame, would the bowser hammers be different? If not, how would you know if the fast accel actually saved time? Also what if you slow down in the last room for some reason?

Ontario, Canada

Yes, the bowser pattern would change, unless you did it after bowser loads.

United States

Timing is done based on visual frames. Given your example, times between 446 and 462 ms would not be valid times. The time ends at the moment a frame is shown, not any time between that and the next frame.

Sweden

@Quivico roopert83 just told us that's not true.

Oklahoma, USA

What @roopert83 and @Quivico said are in agreement. SMB uses the exact frame entering the final section of 8-4 to determine which pattern will be used for bowser's hammers. The pictures of what the hammers look like is referenced :

see:

  1. https://docs.google.com/spreadsheets/d/1RwGuBNiZ40oksnkvA855PWeiMHWYUvc0I_Pt1MvlZXI/
  2. https://docs.google.com/spreadsheets/d/1O9c0ITzt1JRhfwcUFvcQpG_Dh0G9X8xm6hJttVG0-oU/edit#gid=0
  3. https://www.speedrun.com/tools/time_n0yyf.rar

Based on that - we know what EXACT millisecond the final section is entered in. Then you just subtract the # of extra pixels it takes to hit the axe from what would be perfect, and you have your official finish time, down to the millisecond.

[removed paragraph for inaccurate information]

Because each frame is 16-17 milliseconds... you can't finish the game on every possible millisecond, only on every possible frame... which means that finish time can only be possible in increments of 16-17 milliseconds.

Edited by the author 6 years ago
Sweden

@Darpey Thank you for that explanation, and I guess that also means it doesn't matter if the timer is started at the same time as the game is started. The RTA timer is just there for the show =)

KingOfJonnyBoy likes this
Oklahoma, USA

@Odezza exactly - your timer could be completely screwed up... you also don't even need to HAVE a timer. You can do your final split BEFORE you even touch the axe (like I've accidentally done in one of my pb's) it doesn't really matter, that's all for the runner's reference and benefit. Every run will be retimed without that timer anyway.

Edited by the author 6 years ago
Odezza likes this
Colorado, USA

The on screen timer is not needed. It's nice and we do like to see it, but not a requirement. We will ensure that the submissions are timed properly regardless of what your splits say.

I do want to make a correction to the comment @Darpey made. It does matter what platform you play on. If we are using patterns to time your run and you're playing on VC for example, your time will be converted due to the difference in framerate.

Darpey likes this
Oklahoma, USA

Interesting, so if you're playing on the Wii VC, or the retrousb AVS, then your time WOULD reflect the slower framerate?

@roopert83 To clarify, if Kosmic did his 4:56.462 on the Wii VC using the exact same moves relative to the frames that he encountered... his time would have been slower, correct? I wasn't aware of that, but that totally makes sense.

Edited by the author 6 years ago
Colorado, USA

Kosmic's 4:56.462 would be 4:57.824 on WiiVC since the frame rate is 59.824. I just did the math really quickly and didn't double check it but you should get the idea.

Nevada, USA

Correct @Darpey

OpenEmu apparently runs at 60fps instead of 60.0988... fps so despite me having a pattern that was 4:59.989 on console, it's a 5:00.4xx on OpenEmu because of the frame rate difference.

So that's why I'm switching away from my mac and to my PC starting next week, so I can get away from OpenEmu

Edited by the author 6 years ago
KingOfJonnyBoy and Darpey like this
Game stats
Followers
7,784
Runs
8,790
Players
1,781
Latest news
Requirements for High-Level Any% Runs

Any% (NTSC) runs below 4:57.000 must now fulfill additional requirements in order to be verified.

  • The run's full session must be included in the submission description.
  • For emulator runs below 4:57.000, some form of input display must be visible for the duration of the run. A hand-cam or input
3 months ago