Mishu, I just realized that I never answered your question about if it is exactly 45.5 seconds, sorry.

The time I calculated was 45.52 seconds, but that was with my method of looking at the video by .04 seconds at a time, which causes a few random rounding errors in the frame counts. Over a sample size of 75, small random errors of + or - .04 seconds average themselves out, so I think it's a very good estimate.

Anyway, there is no way to ever get an exact frame count unless I have 60 fps source footage of each console, but that's not going to happen. So, I think 45.5 is as close as we need. And since it probably never will be exactly 45.500 seconds, we can retime any close runs. I think that is the simplest, and most effective, solution.

https://docs.google.com/spreadsheets/d/1KoN_0OmwC8PkuEnl2x1niu9zDQcaoObrHGlrbM1TAZE/edit?usp=sharing not private anymore, sorry. Also koffing, I don't agree with what acebentura is saying. I think loading times are the only difference between versions.

I might make a video too after a bit more testing. We can compare strats and stuff

Using a specific setup in 1-1 like you said, then using my strat for the shell in 1-2, I just got shell from a fresh save file three or four times in a row... holy crap. I'll have to make a video too and show my exact setup. Great job on this koffing, maybe you can find something that's 100% consistent.

Edit: I've gotten it so many times that I've lost count. This is insane.

Hmm. I'll have to test some also.

https://docs.google.com/spreadsheets/d/1KoN_0OmwC8PkuEnl2x1niu9zDQcaoObrHGlrbM1TAZE/edit?usp=sharing here's the data. I'm pretty sure you can use decimals, I have my pb timed to the tenth of a second and planned to submit it as that.

The item you get from the box isn't RNG. It quickly cycles through the powerups starting from when it gets loaded. Think of it like the box in the red toad houses, but you can't see which item it's currently on. This is why my consistent 1-2 shell strat is consistent. In the original post, I mentioned that there is only one variable the player controls with my strat- when you jump. The reason it's consistent is that since you hold right and b the whole time, mario falls down the pit, loads the box, gets up, and runs to the right in exactly the same number of frames every time, then the only timing you need is the jump.

@LightOfEon I don't have any experience with the emulators, do you know if all of them have the same loading times? I know there are 2 or 3 different popular emulators.

Don't worry about it, I love data, research, experiments and stuff like this. I'll put all the data in a spreadsheet later then post it here.

If you're wondering about the timings, it starts on the first frame the screen is completely black and ends on the last frame it's completely black.

Making a new thread because the other one is long, and I think this deserves its own thread.

The past few days, I've been going through a few videos frame by frame to count the exact loading times, and now I have an exact number that I am certain of. VC's loading time advantage over DS is 45.5 seconds. I have lots of data if anyone wants it.

Since we finally have a solid number for this, I propose we change the rules/timing so VC runs add the 45.5s timing difference on, and we can time it exactly if two runs are close. If anyone disagrees/has any questions, please reply here. Thanks.

More information:

45.5 seconds is for the route that koffing, ewaller, and I use with 75 loading zones (not 73- I forgot about the opening cutscene). The timing difference would change slightly for a different number of loading zones, but it's not enough to require a retiming unless it would be WR (I'm assuming any future WRs would use the same route we do with 75 loading zones though).

Something else I saw was that not every loading zone had the same timing difference: -entering a level: consistently .55 seconds faster -exiting a level: consistently 1 second faster -pipes: consistently .45 seconds faster -doors: inconsistent loading times. Sometimes faster, sometimes the same.

Lachy, do you mean you want it to be a main category?

doesn't the cannon in world 2 bring you to world 5, not 8?

I don't know if you can set loading times in emu, but I know emu would be the easiest to go through frame by frame and count up the loading times. Also, I'm not sure if the different emus all have the same loading time.

I'll try using ewaller's run and my run, going through and counting up all the frames of the transitions, to get an accurate time, but please can other people try comparing some runs with the procedure I describe here: http://www.speedrun.com/nsmbds/thread/t82dg/3#u2v1y

Multiple people have said that we should time more runs to get an accurate number, but so far no one else has actually done it or suggested a better way.

You're probably right, do you have any ideas about a new way to time things?

Decided on some of the other categories: -downvote all exits because it's too close to all levels and 100% -low% for main category because it sounds really fun

Just want to note 2 things about my estimate: -it's only comparing VC and DS, I haven't tested anything with emu. -this is for 73 loading zones. Going to the red toad house in world 1 adds 2 loading zones, not doing the 8-F2 autoscroller skip takes away 2 loading zones, and deaths add 2 loading zones each.

I guess I should say how I did the timing, if anyone wants to do it on their own or see if there is an error in how I did it.

I could think of four things that affect the final time: execution, overworld RNG, the flagpole, and loading times.

For execution, I used the ingame timer and estimated to the tenth of an in-game second. For example, landing on the flagpole right when it switches from 451 to 450 would count as 450.9. I found the total difference for all the levels and multiplied by .75 to get it in seconds.

RNG was difficult to get exact because hb/box are often off-screen, so I had to estimate based on the number of wing flaps/footsteps.

The flagpole was easy; each time you land backwards it saves 1/3 of a second, so I added those up.

The you can use those numbers and the total time difference to find the loading time advantage.

The most accurate way to find it would be to go through all 73 loading zones a count the number of frames each one takes, but that's not really feasible going off just youtube videos.

If you spot an error in my procedure/think of a better way, please tell me.

Since most people in here seem to agree about changing the timing, should we have a vote on how? Maybe the options could be something like:

1. add 48s to all VC runs
2. edit out loading times for all runs
3. add 48s to all VC runs, but edit out loading times if two runs are close (+/- 3 seconds?)
4. going by in-game timer, or some variation of this
5. something else I'm not thinking of
6. don't change how it is currently

Feel free to suggest other options.

My vote would be #3.

I vote for cannonless being main and NSE being misc, because I think cannonless would be the next way after the any% route that people would play the game. Still undecided on the other categories.

Northern here. It should be starting in a few hours for me, good luck haha.

Video uploaded:

There's a HUGE snowstorm about to hit me and my power might go out, sorry if I can't answer questions.