~ Announcement ~

Starting from today, the ranking in the Q3 leaderboard will be decided based on times without loads. The times with loads will remain in their own column. Loads include the 'awaiting gamestate/connection' screens and the loading of the map assets, bots etc before the spawn. Note that the menuing required to go to the next map is not included in the loads and is still part of the final time.

~ What's the idea? ~

Recently, we started to notice that the loads' duration differ massively from one player to another. After some investigation, it turns out that beyond the CPU specs of the players, the OS actually has a massive impact. For whatever reason, Windows 7 has a ~25s advantage over Windows 10, due to the fact that Win 10 gets stuck on 'awaiting gamestate' screens. Windows 8 is somewhere in between.

On top of some computers being able to load the maps much faster, we all agreed (on Discord) that such an unfair factor mostly out of our control should be removed when comparing runs. After careful consideration, it was decided that an autosplitter with load removal (courtesy of liveth/pekafacer) will be used.

~ Autosplitter ~

If you want your runs to have a time without loads, please use the livesplit autosplitter script that you can find in the Resources section. So the competition and ranking are as fair as possible, we'd like for everyone to use that same tool to get their no load time. https://www.speedrun.com/tools/Q3A_ASLR_644it.rar

liveth explains in the following post how to set it up if you're unsure: https://www.speedrun.com/post/eenu7

~ What about old runs? ~

So that all the past top runs without a 'no load' time don't become obsolete/overshadowed, Ewil and I are going to retime them to some extent when we have a chance, and manually determine 'no load' times. We can't promise to retime the entire leaderboard obviously, and priority will be given to runs near the top.

~ Can I just submit times with loads? ~

Yes, using the autosplitter is not mandatory to be in the LB. Note that if you don't use it and don't have a time without loads, it will impact your ranking though as any run that used the AS will be marked as faster than yours. If you don't care, feel free to submit with loads if you prefer.

That being said, if you're going for a top time or consider going for one, please use it so we can compete on equal grounds and accurately have your run at the top if it should be there!

~ Bonus information/tech for loads ~

funny_ha_ha found out that the gamestate screen can be made much shorter, even on Windows 10, by kicking the bots at the end of the map before the pedestal.

This can be done by binding "kick allbots" to a key and using it after the last frag. Make sure to not use it before reaching the fraglimit though or you will be unable to complete the map.

Note that as the gamestate screen is already removed from times without loads, this tech has no impact whatsoever on those, and on the ranking. But if you're interested in also pushing your time with loads, give it a try!

I noticed that due to the double 5:55, a mod went and added milliseconds to all the runs in the LB.

The problem is that all of them were incorrect because they were just based on what livesplit was saying (sometimes when the text screen was shown (i noticed that in the case of Baruch at least), sometimes just when the timer was stopped like for me or the newer 5:55. In the first case, it didn't take into account the gap at the start, and in the case of the 2 other runs I mentioned, it didn't take into account the gap both at the start and at the end.

In short, do not add milliseconds unless the run was framecounted (we need to figure out a clear start & end point for that in fact) or unless there's an autosplitter out there that we could use (that would be great but there isn't one right now).

It's true that the new 5:55 is slightly faster and the LB should reflect that, but adding wrong/quite off milliseconds is not the answer.

The Switch version of SMB3 is now allowed for all 4 categories of Super Mario Bros. 3 (any%, any% no WW, any% warpless, 100%).

Note that the Switch has a framerate very slightly slower than the NES and you'll lose a few seconds over the course of a 100% run due to it (the difference is quite negligible unless you're going for the WR).

In the case of any%, note that the wrong warp setups possible on NES (that you can see all over the LB currently) are the standard we use to determine which version should be allowed or not. For example it was found out that WiiVC behaved differently in that regard and as a consequence was banned for this category.

What it means is that with the informations we gathered so far, it seems that the Switch behaves similarly to the NES, and NES setups have been found to work on the Switch. This however might be reassessed in the future if new data comes up, with Switch specific possibilities for example. If that were the case, the version might become banned for this category alone, but we would still grandfather in older runs using "NES compatible setups" so don't be worried.

All in all, if you're planning on playing any% on the Switch version and are noticing something in 7-1 that seems to differ from the NES, or a setup possibility that nobody else seems to use, please contact us (on Discord) and we'll look into it if necessary.

Hey so I received messages from garadas regarding beco's runs having been reuploaded on Youtube:

garadas21: hey there! i found two of beco's gimmick run on youtube garadas21: both WRs garadas21: i saw that you're one of the mods garadas21: could you add them both? tnx! garadas21: any%: garadas21: best ending:

Imo the records should definitely be added to the LB again but that being said, seeing as the runs are on someone else's youtube (the guy is not taking credit for the runs to be clear, he just reuploaded them unlisted so they're not lost), it would probably be a nice idea to send a message to becored to see if he wants to reupload them himself on one of his accounts, whether it's twitch or youtube, and submit them again himself.

People request from time to time that the All-Stars version gets its own leaderboard, so I'm making this thread to try and clarify what the main differences are between the various versions of SMB3, how the versions rank in terms of speed, and why we've decided in the past to keep all of them in a unified leaderboard.

~ Version Differences ~

NES (NTSC US) + fastest loading time for stages

• Mario loses only one powerup increment per hit

• slower castles than japanese versions (text speed)

NES (PAL)

• Mario loses only one powerup increment per hit - game runs at 50 hz instead of 60 hz (so the game speed is only 5/6 of the NTSC speed) - shorter jumps than NTSC, making many strats impossible

FC (NTSC J) + fastest castles (gains several seconds per castle compared to US versions) due to text speed - Mario goes small mario when hit, regardless of powerup (makes a few strats impossible) - extra fade-in at the start of every "stage" (includes overworld pipes among other things) wastes around 1s each time

• extra wall & pipe at the end of 5-1, wastes a few seconds

SNES (All-Stars US) + lagless, which works in your favor most of the time but can potentially hinder a couple of despawning strats

• can use fireballs with two different buttons: allows faster mashing, and using fireballs without releasing B (run)
• Mario loses only one powerup increment per hit
• 4-4 is easier & slightly faster (small detail) - extra "Mario Start!" screen at the start of every stage, wastes around 86 frames each time - doors are slightly slower

• slower fall after grabbing the wand
• can't fireball while offscreen, hinders a few strats

SFC (All-Stars J) + lagless, which works in your favor most of the time but can potentially hinder a couple of despawning strats + semi fast castles (combination of faster text & slower fall)

• Mario loses only one powerup increment per hit
• can use fireballs with two different buttons: allows faster mashing, and using fireballs without releasing B (run)
• 4-4 is easier & slightly faster (small detail) - extra "Mario Start!" screen at the start of every stage, wastes around 86 frames each time - doors are slightly slower

• can't fireball while offscreen, hinders a few strats

Extra: Wii VC - slightly slower framerate than the NES, wastes around 20s for a hundo run

Note that this isn't meant to be an exhaustive list of the differences. For example, Mario appears in the overworld 1s faster in All-Stars at the start of the game compared to NES. Also, the 1-3 whistle Mushroom House is faster on J. The differences listed should be the biggest though.

~ Version Ranking ~

It's impossible to quantify exactly what the difference is between let's say NES & All-Stars. For example, you can't put a number on the "lagless" aspect of All-Stars.

That being said, the time gained/lost due to certain aspects is big enough that we can determine a ranking:

NES > FC > SFC > SNES >>>>>> PAL

The NES Wii VC version would be somewhere between NES & FC, closer to FC.

Also note that the FC version is the most difficult version due to being less forgiving for hits. In that regard, PAL's difficulty is hard to assess because PAL is the only version that requires its own set of strats for almost every stage.

~ Unified Leaderboard ~

-> So, if there's a clear time disparity between some of the versions, why not separate LBs?

The main idea behind the unified LB is that almost every strat is the same in every version (ignoring PAL), the routes are also pretty much identical, and all the physics/core mechanics are the same. Meaning that the time difference is mostly from loading times. Having an entirely separate category/LB just because of loading times seems really overkill to us. Using filters is usually more adequate for cases like this (like for LBs with multiple language versions).

-> But other Mario games do it!

Other 2D Mario games with separate LBs for All-Stars usually have major gameplay differences between the two. For example Super Mario Bros. 2 (USA) allows you to Save & Quit, which allows you to do a completely different route in any%. In SMB3, the gameplay & routes being almost identical, it's a different case (although obviously subjective to some degree).

But there's also other reasons. For example, if you're gonna have a separate LB for All-Stars, you then need to separate SNES & SFC. Because the time difference during castles is big enough that SFC is clearly better than SNES.

In that regard, PAL would obviously "deserve" a separate category way more than any other version, considering you need your own set of strats for it.

All in all it would end up looking like this: Any% Warpless (NES) / Any% Warpless (FC) / Any% Warpless (SFC) / Any% Warpless (SNES) / Any% Warpless (PAL) And now imagine that for every category.

Additionally, the bulk of the competition being on NES, we strongly feel that fragmenting the community and isolating a few runners is unneeded.

Lastly, as mentioned earlier, you can (and should) use filters if you want to see the competition/best times on a specific version. For example select Filter -> Platforms -> SNES if needed.

The unified leaderboard is not meant to be unfair or insulting to players using different versions. We're just trying to be coherent and efficient.

Hey, instead of cutscene trigger at the end, can't we just end the time on the final frag against xaero? Way easier to pinpoint (and i've been verifying thinking it stopped there until now honestly) and makes the most sense as it's the last thing you do in the run.

Not sure if this is the appropriate place to request this but would it be possible to give an option in the LB settings to mark milliseconds as optional (not saying it should always say optional next to milliseconds for any LB, but that the option is there for some LBs if mods want it).

Basically my problem is this: I added the possibility to add milliseconds to the times submitted for SMB3 because it can prove important for short categories. Obviously, it should only be used if the run was actually framecounted, and whole seconds are perfectly fine otherwise, particularly for long categories. But since I added it, there's been many submissions with milliseconds that are completely wrong because people just entered whatever livesplit was saying at the end of the run. Just trying to think of a way for people to stop doing this (if they didn't framecount).

If not an "optional" message, maybe hide the milliseconds by default in the submission page and show it only if the runner clicks something, i don't know.

OK so your speedrun.com account (0100100001101001) being brand new, can you tell us who you are, if this is your first speedrun etc.

Additionally, where did you learn that setup? Most of what you do is known by some players but hasn't been showcased in any public run yet (because of the difficulty to pull it off). Please detail everything you do in 7-1 as much as possible.

Also, we're gonna have to ask you to give us some videos of attempts. You don't seem to be streaming, your Youtube account only has this video, so the only thing we can rely on is your offline sessions. You say on your Youtube videos you grinded for weeks, so please show something for it.

Thanks in advance for helping with the verif!

In any% (and only in that category), the time now ends on the 1st frame Mario is visible inside the chamber (and not on the 1st frame the chamber itself is visible). The entire LB has been retimed (by me) and edited. You can see all the changes here:

http://pastebin.com/7PZXCfhN

If you think I made a mistake, don't hesitate to contact me or another mod and we'll look into it. As far as I can tell, the only "problematic" run is Fernando's because it's so borderline it should be framecounted from the source (but even then I think the new time should be accurate).

If you're wondering why we decided to change that (there was discussions about it on Discord), I put in that pastebin the message I initially posted on Discord about all that:

http://pastebin.com/dgvcchWf

Among the people who expressed their opinion on the subject, there was an overwhelming support for 1st Mario frame, a couple of people in favor of last input, and nobody wanted to keep the old system (it was shit so that's no surprise). If you want to see the full discussion, I should be able to pastebin it (also if you're not in Discord and want an invite, just ask, most runners are there already).

Additionally, I now added the possibility to add milliseconds to the times in the LB. PLEASE ONLY USE THAT IF YOU FRAMECOUNTED YOUR RUN. Don't just write down what livesplit says if you manually started and stopped the timer. Rounding down to a full second isn't a problem right now but it could be one in the future for short categories in particular. If you don't want to bother framecounting your run (and honestly, don't bother unless you're going for WR in one of the two short categories), you don't have to worry about it, times in whole seconds are perfectly fine.

Edit: oh and one thing, we kept the "If the game crashes the run is invalid" line in the rules, but one of the purposes of stopping the timer on Mario is you only see him if the game didn't crash, so we basically stop the timer when the run is confirmed to be valid.

Hi Fernando,

So regarding your run: please remember to always write the emu (& version) in the description when submitting. Your previous run was on fceux so I'm assuming it still is, but please confirm that point.

Also your timer seems to be a later addition done with Vegas, and you get a 3:11 final time on it assuming it's accurate. If it is, why did you submit 3:12? I edited your time into 3:11 for now because we don't round up the times, we just take the whole second (and for short runs at the top we should add decimals in the future in fact, particularly for any%).

So yeah, if you can confirm the emu & the time it would be great! Thanks