Tool-assisted game movies
When human skills are just not enough

Submission #5617: HappyLee's NES Super Mario Bros. "PAL" in 04:55.16

Console: Nintendo Entertainment System
Game name: Super Mario Bros.
Game version: Europe
ROM filename: Super Mario Bros. (E).nes
Branch: PAL
Emulator: FCEUX 2.2.1
Movie length: 04:55.16
FrameCount: 14760
Re-record count: 11754
Author's real name: Li Tianda
Author's nickname: HappyLee
Submitter: HappyLee
Submitted at: 2017-07-26 18:16:33
Text last edited at: 2017-09-14 20:42:30
Text last edited by: Nach
Download: Download (3420 bytes)
Status: decision: rejected
Submission instructions
Discuss this submission (also rating / voting)
List all submissions by this submitter
List pages on this site that refer to this submission
View submission text history
Back to the submission list
Author's comments and explanations:
This run is definitely not with the intention of obsoleting the published SMB any% TAS! This movie aims to present what the fastest and the best SMB PAL TAS would look like. It improves MUGG's submission for 66 frames, and tubby's TAS for 46 frames.

(Link to video)
Download 1024x960 HD encode (MKV, 8.95 MB)

The Europe version of SMB is an official game run in PAL mode. The physics are almost identical, but the speed values are set differently, causing more potential glitches. TASes on this version are only faster due to a different version of flagpole glitch, which allows Mario to skip the castles without the help of other enemies or entering the ground.

New Trick: Falling into the Ground

It's probably no secret that Mario can sometimes fall into the ground after stomping on an enemy in SMB Europe version, but to do this without the help of anything but a lift is something new. The lift is still required for manipulating Y position before jumping. This trick is used in 1-2, saving 18 frames (a frame rule).

Time Saver: Faster Acceleration

It's faster to accelerate on the floor in this version. This simple new way of acceleration saves 18 frames in 8-3, and 10 frames in 8-4 (along with other arrangements).

This run also differs from the normal SMB any% TAS on details. For example, only in this TAS is Mario able to kick some shells in 8-1, and to show the 1-UP mushroom in 8-2, to walljump on the higher floor and to swim through the ceiling in 8-4.

I'm submitting it here mostly to show people what the best SMB PAL TAS would look like, regardless of whether it has reached TASVideos' standard for publication.

Nach: Let me start off by saying that judging this was one of the most difficult to judge TASs. The verdict I'm presenting here is based off of the current rules and knowledge I have regarding this run. It is subject to be revisited if anything significant changes. It should also be noted that no matter what the decision here is, a large chunk of people will not be happy with it. I will however lay out some additional info not discussed in the thread which factored into my decision.

Before I dive in, let me also iterate that this was an entertaining run, and there is little to dispute that, certainly by the audience at large.

NTSC vs. PAL theory

In terms of PAL games in general, different platforms, different companies, and different games all exhibit varying levels of quality. Obviously if a PAL game is the original then it can be easily considered the main version of a game. For some platforms, there are also no difference game-wise if something is running in NTSC or PAL mode. However, for platforms designed to be timed and framed into old television sets, there are important differences between the two modes.

Once there is a difference between the two, games designed for NTSC which are not modified for PAL generally exhibit some very weird behavior. As one example, I've seen fighting games where the key combos to execute various moves barely work when playing in PAL mode, the timing is altered enough that the game doesn't recognize the key presses the same way. As many PAL ports are like this in some fashion, it's ample reason to reject them, Just play the original which works normally.

Game variants on TASVideos

When we look at PAL ports, we must understand that these games are adaptions or variants of the original. Although there are many kinds of variants. Some variants are ports to a later platform. One kind of variant such as those seen in Mario Bros. has completely different levels (even though all the levels are repetitive). Some variants like those in the Street Fighter 2 series are the same game but with changes with varying levels of importance. Castlevania II: Belmont's Revenge exists in two variants where the entire game is the same, except one has a boomerang as a secondary weapon, and the other has a throw-able ax. The Pokémon series has game variants at each generation, generally limited to monster selection, where a dozen out of 150+ are different (which may not differ at all with certain glitches exploited), but barely has any affect on how a well planned run plays. Other differences are ports from the NES to SNES to Gameboy Advance and so on. The deciding factor in how these are dealt with on the site always boils down to how identical are the engines, and how unique and interesting is the gameplay that each variant offers over others.

Taking SMB2 as an example, the SNES variant adds on a save game feature which can be abused which can change the warped route considerably. Same for the Gameboy Advance variant, which further has other game changes. Due to these considerable changes in what one would see in a TAS for them, we have accepted them all.

In the case of Pokémon, since the engine/quality of the game between say Blue and Red is identical, and the observable changes in a TAS are insignificant, any new record with one will always obsolete the other.

In the case of various Street Fighter games, there is a large similarity to the TASs being produced. The audience at large doesn't notice much other than some Street Fighter characters are more or less beating up the same set of Street Fighter characters, using many of the same moves. In these kinds of cases, we have the best version of the game obsolete the others. Best version often is based on figuring out which has the broadest set of move possibilities, most fluid version of the fighting engine, and so on.

We haven't had multiple variants of Castlevania II: Belmont's Revenge submitted yet, but if someone submits a boomerang heavy run with sizable differences from the existing ax run, I can see accepting them side by side. It's sort of like accepting various X and Zero runs side by side for the later Megaman X PSX games (note: I rejected some MMX5 runs for being too similar to others).

In terms of Mario Bros. since a full variety run of levels while similar is quite different, we have accepted both.

NTSC vs. PAL in practice

If a PAL port offered the exact same engine/quality as the original, it could make sense to have it obsolete the original (and this could make sense to occur in games that are not meant for old television sets). If a PAL port offers a somewhat different engine, the question becomes whether it deserves to be added to the list of accepted variants alongside the original. The answer to that hinges on do the engine differences necessitate very different ways to play the game, and do those differences register with the audience at large. In most cases, PAL runs should be rejected, but based on the various aforementioned criteria, there are cases where PAL runs will definitely be accepted.

Nintendo unlike other companies has always aimed to do a decent job porting NTSC games to PAL. Nintendo is often one of the only companies where you will see the PAL game having various timings corrected to ensure that the game-play closely matches that of the NTSC version. Nintendo is often one of the only companies that adjusts the resolution of the displayed game to match the different resolution PAL televisions are capable of. Nintendo often also does some localization, converting currency, weight, and measurements to be those used locally. The attention to detail by Nintendo in PAL porting started with early NES games, and improved as Nintendo ported more and more and with newer platforms.

This game in particular

For its time in history and in comparison to a bunch of other early NES PAL games, Super Mario Bros. PAL is actually a relatively decent port (although maintains several noticeable differences from the NTSC original in terms of movement and other factors). Since the game is non-original but a decent port (relatively during its debut), it definitely qualifies for consideration as to whether it should be published alongside the original as another game variant.

This game happens to also be a game I'm quite familiar with. I played many of its variants on NES (since the 80s!), SNES, and Gameboy Color. I also dabbled in its programming and made various hacks on NES and SNES versions. In my opinion, I find this game qualifies for having many branches made of it. I can also see the SNES variant qualifies for certain branches as an acceptable TAS to show off a run without as many glitches being possible, and the Gameboy Color variant for some of its challenges that earlier versions do not offer. The question of course is, is there value in this PAL variant that we have lacking from all our other variants and branches thereof?

The first thing I want to shoot down is the idea that SMB PAL is faster than SMB NTSC. There are quite a few parts of the game that are non-playable. These include score countdown, castle animations, pipe transitions, 1-2, 2-2, 4-2, and 7-2 initial cut screens, level banners, vine climbing, and Bowser drowning to our princess is in another castle. When comparing across versions we need to take all this into account and figure out actual game-play time. NES SMB processes the non-playable segments of the games in multiples of 21 frames and 18 frames for NTSC and PAL respectively. Nintendo altered the number from 21 to 18 because 21/60 and 18/50 is 0.35 and 0.36, which should provide a close gaming experience on the port. In actuality, using more precise numbers, NTSC has frames which are ~0.0166 seconds long, and PAL ~0.0199 seconds. This means the non-playable parts are processed in multiples of ~0.3494 seconds and ~0.3599 seconds. Since these non-playable segments run on boundaries that are multiples of these, it means that the NTSC version allows slightly more time to get in activity before the game will round upwards. Conversely, if you just went a bit over a multiple, the PAL version will proceed to the next multiple sooner.

In order to get a better handle on this, I went to time the actual playable segments between the fastest NTSC and this PAL run (note, there may be rounding errors, and it's possible I was a frame off either way for some calculations):

1-1 12.2305 12.083
1-2 21.583 21.15
4-1 23.983 23.967
4-2 17.9495 17.567
8-1 40.0825 40.233
8-2 24.8655 23.383
8-3 22.6985 22.767
8-4 32.5275 32.601
Total 195.92 193.751

Based on this NTSC is slower by ~2.169 seconds (about 130 frames in NTSC). However, there is a flaw with this logic. These runs aim for overall fastest real time, and thereby performs some actions which are slightly slower in the playable segments in order to abuse how the non-playable part is played as well as avoid 3 or 6 castle fireworks animations. However, the NTSC run goes significantly out of its way in 8-2 to abuse this trade off, by ~2.379 seconds in my calculation. If the run would discount non-playable segments to achieve the fastest possible any-variant time, we'd instead be looking at:

8-2 22.4865 23.383
Total 193.541 193.751

In this case, the NTSC version is faster by 0.21 seconds (about a dozen frames)!

NTSC improves further if we decide that the mid-level non-playable segments must be included in 1-1, because unlike other levels, going through that here is a decision that can be avoided. In that case the 1-1 times become:

1-1 18.1655 18.433

Gaining the NTSC run an additional 0.415 seconds (about 25 frames). All in all, PAL being necessarily faster in terms of game-play is doubtful.


Armed with all the aforementioned information, how do we look at this? I decided to ask other judges for their opinions for the different possibilities, raised a few counterpoints with them, then assessed how they changed their opinion. I will not list their names because I should be the sole person receiving any fallout for the judgment on this run. What follows is how I characterize the opinions they conveyed to me.

Before I mentioned (counter)points:

Judge Obsolete New Variant Reject
A Absurd Yes! No
B Absurd Yes Maybe
C Yes No way! Maybe
D Absurd Yes! No


Judge Obsolete New Variant Reject
B Absurd No way! Yes!
C Maybe No way! Yes
D Absurd Maybe Yes
E Absurd No Yes!

(One judge was unique in each group)

When I initially saw this run, knowing the differences right off the bat between variants and our aims, it seemed clear to me that obsoletion was lunacy. However my knee-jerk reaction was that I love this run, the engine is a bit different, let's just accept this as another variant. However, those are not good reasons to accept something, we have rules.

Thinking about how this run actually differs from the NTSC when viewing, it's not by much. More than that, there's nothing that really necessitates a difference. Just because one run decided to randomly jump at some point does not make it different from a run which does not. It has to be different as a branch in a significant manner, not just how it was played back in a particular run or mere moments of it. The new glitch, while new, does not look so different going through the wall than going through the wall otherwise. Also, I'm not convinced every run of this PAL branch would require this glitch being abused. So looking at changes across the run, they seem minor, and 4/5 judges I spoke to are now in favor of rejecting.

After assessing everything yesterday for one last time, I was conflicted on what to do. After sleeping on it, seeing no new convincing posts one way or the other, and considering the different factors listed above further, one side in my mind now slightly outweighs the other. In conclusion, while some PAL games are acceptable, and other branches for SMB PAL may be acceptable, this TAS does not seem to be acceptable with what we know right now and how we handle these sorts of things. Rejecting.

Nach: Since some people had a hard time following the above points, I put together a decision tree.
Nach: The last judge (Judge A) has since wrote back to me that in light of additional data/(counter)points, they now also favor rejection.


Nach: When we accept improvements across game versions, we only do so when there are actual improvements in the game-play by the player(s). The quality of the existing published NTSC run and this submission are practically the same. This submission did not improve upon the existing NTSC publication in any meaningful way. All time-related improvements are due to subtle version differences that the player has no control over. Since there is no improvement upon the existing publication once the version differences are factored out, this submission is not considered an improvement.

The game-play in this submission is similar to existing publications, and there does not seem to be substantial differences to warrant this submission to be published alongside them. After speaking to five judges regarding the similarities, they are all in favor of rejection. Rejecting.

Similar submissions (by title and categories where applicable):