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

Submission #4859: byrz & CyclopsDragon's GC Super Monkey Ball "Expert through Master" in 12:55.18

Console: Nintendo GameCube
Game name: Super Monkey Ball
Game version: USA
ROM filename: smb.gcm
Branch: Expert through Master
Emulator: Dolphin 4.0.2 64-bit
Movie length: 12:55.18
FrameCount: 46511
Re-record count: 275856
Author's real name:
Author's nickname: byrz & CyclopsDragon
Submitter: byrz
Submitted at: 2015-09-27 12:00:51
Text last edited at: 2015-10-06 20:21:09
Text last edited by: Fog
Download: Download (13702 bytes)
Status: published
Click to view the actual publication
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

Older version of DTM detected: Run time may be incorrect

Author's comments and explanations:


Please watch in 720p 60fps. We are not embedding this video because TASVideos uses a youtube player that doesn't support this quality.

Super Monkey Ball is a physics based arcade-“platformer” released in 2001 for Nintendo GameCube. The game features 3 difficulties, Beginner, Advanced and Expert, each with their own set of levels greatly varying in difficulty. The difficulty chosen for this TAS is expert, also considered as the main category of this game, covering the largest amount of content out of any difficulty. The game engine has no real glitches, rendering the TAS to be purely about optimizing the movement to beat the floors as quickly as possible.

This TAS is a major improvement over the currently published TAS of this category, Super Monkey Ball “Expert through Master”, playing all the way through the 50 Expert, 10 Expert Extra and 10 Master levels. The final in-game time achieved in this TAS is 4:01.733, improving 938 frames or 15.633 seconds over the previous TAS.


  • Emulator used: Dolphin 4.0.2 64-bit
  • Aims for the fastest in-game time
  • Warp goals are used to skip levels
  • Death is used to complete bonus stages faster

Emulator version

We have decided to use this version because it’s the one we are used to, and no newer version has known advantages for TASing this game.


The previous TAS of this category aimed for the fastest completion time. We have since decided that in-game time is the timing method that matters most in Super Monkey Ball TASing, in terms of measuring quality and how optimal the TAS is. It also translates well between different emulator versions with varying loading times; for example, the newer versions have loading times more accurate compared to original hardware, making them longer. The in-game time is measured by converting the completion time of each level into the number of frames, then converting the total framecount back to normal time, each frame representing 0.01666… (1/60) seconds. We would also like to point out that a run being optimal in in-game time is not mutually exclusive with a run being optimal in real time or TASVideos timing and that we also did optimize every aspect outside of in-game time, emulator glitches and sync issues being the only occasional limiting factor.


A few of the levels in this game have more than one goal, in which case the additional goals are either green or red. In expert mode, green goals skip over 1 floor and red goals skip over 2 floors. In this TAS, 5 floors are skipped with these warp goals, which means 65 out of the 70 total possible floors in the difficulty are beaten.


Falling out in bonus levels instantly finishes the level, making it a much faster way of completing bonus levels. This trick is used in all of the 5 bonus levels in the difficulty. This doesn’t cost any lives either, so it’s debatable whether they count as “deaths”.

Techniques used

This game is purely about movement, the analog stick being the only control there is to move the monkey around. Even with an engine this simple and restricted, there are many techniques you can use to save time.


Boosting is a movement technique used to accelerate the monkey on flat surfaces. It is performed by alternating between the two up-diagonal inputs (up-left, up-right) on varying paces. It is most commonly used right when a floor starts, being much faster than simply holding straight up. Boosting on ground is useful all the way until your speed is around 51 mph or more. At that point just following the shortest possible route is faster. Boosting is also especially helpful in most upward slopes.


In mid-air, the monkey accelerates much faster than when rolling on the ground. Clipping is used to gain this precious air time. However, clipping too high up slows you down more than it helps so it’s mostly useful when you clip in angles such low that you stay barely above ground. Clipping is also used to launch the monkey into the air to pull off shortcuts in some floors. Clips can be performed from many kind of edges, corners and bumps.


In order to cross over certain curvy bridges while also preserving a high velocity, you can perform a grind on the inner edge of the curve. This technique is usually very hard to perform but saves a nice amount of time. Grinding does not save time if the radius of the curve is so big you could cross the bridge full speed anyway.

Air boosting

Slowly alternating up-left and up-right midair increases the distance you cover before landing. However, it makes you fall slower, so it should only be used when really necessary.

Fast falling

Opposite to air boosting, using no input or light analog inputs in midair increases your falling speed, saving time in falling stages. Completely releasing the analog stick is the fastest possible way to fall down. Doing light analog inputs or straight down input is also faster than regular “strong” inputs.


Normally, you are supposed to break the goal tape by physically moving through it. However, there is a very small window where the tape can also break if you go through the “sidegoal” above the tape, next to the party ball in a specific way.

Glitch goaling

The party ball located in the goal of each floor has some interesting behaviour. Sometimes it can pull you around in unexpected ways if you touch it with certain speed and angles. In some cases, this behaviour is abused to make goaling possible at all. The approach angle might otherwise be too extreme to allow breaking the tape. Glitch goals are so sketchy that the replay function of the game will not even break the goal tape in replays if you finished with a glitch goal.

Collision manipulation

This technique is more of a TAS-only technique. It consists of testing many kinds of analog input combinations to get desirable collision from any solid surface, in order to get perfect clips and shortcuts.

Author comments


I have TASed Monkey Ball 1 and 2 since summer 2013. I’ve done countless individual level TASes and six full TAS runs of the games prior to this project. The previous TAS of this category from spring 2014 was a big success, and at the time it was made, it was considered near-perfect. However, by finishing 4 more TAS runs (beginner and advanced for SMB1 and 2) I realized I have gained even more experience in TASing this game. In November 2014, I decided to start an improvement project for this category, without any idea that more than one second could be saved over the first TAS.

Over 10 months of hard work (+ a summer break) has proved that the category still had insane potential for improving apparent perfection. We saved over 15 seconds by improving 55 out of 65 possible floors; at no point could anybody have predicted half of the improvements we managed. This TAS is perhaps the single piece of work I’m most proud of I have ever worked on, it was absolutely worth the close to a year’s effort and hard work it took to finish this monster.

CyclopsDragon joined this project after slightly less than one third of the TAS was finished. His effort has been a tremendous help through the project and even though his inputs in the final recording are more sparse than mine, he worked just as hard as I did for the most part, sometimes even more.


I started TASing Monkey Ball games for fun in the summer of 2014, focusing more on stunts than times and scores. Byrz brought me onto this project in January of 2015, and after working with him for a while, I realised how difficult it was to keep up with his ridiculous standards of optimization, since his understanding of the original Super Monkey Ball engine is much deeper than mine. This whole TAS project has been a really rewarding experience, between bouncing ideas off of each other and random discoveries of inexplicable, yet helpful collisions. I’m very happy to be able to say that I was able to help with this project, because the end result is absolutely astounding.

Individual levels

Floor Old TAS New TAS Frames saved Comments
E1 28.35 28.35 0 Looking deceivingly easy, pulling off this time is not a simple task
E2 26.45 26.46 1 Saving this frame was very unexpected. Previous TASes of this floor started from .40, climbing up one frame at a time up to this point, each frame save being more unexpected than the previous
E5 58.06 58.10 2 Pretty simple time save that wasn’t found at the time making the first TAS
E6 55.73 55.73 0 This is one of the only floors of the TAS where we had to use the inputs from the previous TAS because it turned out to be an absolutely insane time to match
E7 51.05 52.10 63 Completely crazy new strat was pulled off. This strat was thought to be simply impossible, with previous attempts at TASing it falling short. With the magic of perfect movement optimization, we barely managed to pull this off
E8 27.01 27.01 0 One of the simpler floors
E9 55.75 55.83 5 We managed to save a bunch of frames with a sexy sidegoal finish
E10 27.46 27.50 2 This simple 2 second bonus stage took a huge amount of rerecords compared to its size, but a 2 frame improvement was pulled off just before giving up looking for time saves
E11 23.46 23.48 1 Completely different route, one frame faster. The left railing has strange collision, giving you way more speed than it should. Briefly visiting it allowed this strat to happen
E12 55.63 55.66 2 Managed to save 2 frames with this clean run
E13 57.41 57.43 1 Surprising frame to be saved in such a simple floor
E14 57.35 57.41 4 By abusing these poor bumpers, we pulled off a new strat no one even imagined before
E15 28.38 28.38 0 Funny looking goal where you get stuck in the goal post until being forced through the goal tape
E16 27.85 27.86 1 Another unexpected time save for a short floor like this
E17 27.91 27.91 0 Cycle-based floor with a very precise bumper hit
E18 21.83 22.01 11 Extremely simple looking floor, yet there are many little tricks used to make it faster. Amount of time saved was somewhat insane
E19 57.60 57.65 3 Who would have thought that simply mirroring the normal strat would end up 3 frames faster?
E20 27.85 27.90 3 Not much to say about this, a minor movement difference produced a nice time save
E21 56.23 56.28 3 Never expected to see this improvement
E22 54.66 54.78 7 Prime example of how the game handles collision at very high velocities... if you manipulate it right, that is. Crazy time
E23 27.91 27.93 1 Yet another magical frame save for a short and simple floor
E24 43.01 51.86 531 Holy crap. A HUGE time save, saving more than half of all the time saved in the TAS, thanks to a huge new shortcut being found after researching and TASing this floor for a long time
E25 24.70 25.01 19 Dwarfed by the previous floor, this time save is still considered huge. Instead of skipping over the 2nd and 3rd “dip” and then flying straight to the goal from the 4th one, we managed to use the 1st and 2nd ones combined with a nice clip from a peak for this crazy run
E26 27.33 27.36 2 We found out how much fast falling matters in this floor and then found a nice way of inputting the moves for this time save
E27 57.00 57.03 2 By finding a slightly nicer set of clips, we managed to cut off 2 frames
E28 56.58 56.83 15 Instantly dropping off to get clips from every edge on the way to the goal ended up a really nice time save
E29 55.01 55.20 11 One of the cycle-based floors where going later allowed us to preserve more velocity, bringing us a crazy time save
E30 57.65 57.68 2 Frames were saved for no apparent reason other than lucky collision
E31 28.38 28.38 0 Very short and simple
E32 47.56 47.68 7 The longest level in the TAS, pretty nice time save. Grinding only helps in the second curve, in the first one it would only slow you down. The shortcut used requires a very strange collision
E33 27.83 27.85 1 Simple shortcut, unexpected frame save
E34 26.91 26.91 0 This floor is a massive taunt. Due to the cycle, you can only get the right collision on specific timing, even if you can make it there one frame earlier you are awarded with a worse collision. In addition to this, the monkey is incredibly close to breaking the tape a framer earlier
E35 56.16 56.23 4 This time save is kind of nuts. The floor has some of the wonkiest collisions to abuse
E36 26.38 26.50 7 One of the cooler strats. The ending of this one includes a collision that defies all the logic in the universe. Just discovering the collision was an extremely lucky thing to happen
E37 27.26 27.28 1 No one could have ever imagined that there was a frame to be saved in this floor, but there was
E38 27.58 27.61 2 Very nice clean time save that no one expected
E39 26.45 26.56 7 A new speed bump was found in this floor, allowing this time save to happen
E40 27.00 27.00 0 Finally, the last bonus floor in this game. No one likes you
E41 28.18 28.20 1 No idea how this floor had one more frame to be saved. Sometimes things get crazy with the party ball giving these sketchy finishes
E42 58.63 58.63 0 This red warp goal is supposed to be hard to reach, right?
E45 23.88 24.03 9 Very hard strat to pull off. We also found that you can squeeze some crazy speed out of the left edge of the second bridge
E46 23.73 23.83 6 Some impressive collision manipulation at high speeds brought us a nice time save
E48 26.03 26.08 3 Very annoying level to TAS, attempting the shortcut gives you crap 99.5% of time. Happy with this time save, though
E49 28.48 28.51 2 Last time we saved one frame by abusing one bumper. This time we saved two frames by abusing two bumpers
E50 25.36 25.53 10 Crazy how many more frames we squeezed out of this insane strat. People generally seem to like seeing this strat
EX1 26.00 26.03 2 Thoroughly testing different clips on this level helped us save these frames
EX2 26.76 26.80 2 Another little nice time save
EX3 26.71 27.88 70 Wow. Never imagined you can actually finish this floor in a cycle this early. Definitely one of the most unexpected new strats in this TAS. Also got a very cool finish with a record breaking septuple “Yay!” from Aiai (unconfirmed)
EX4 27.56 27.60 2 Managed to finish with a sidegoal by touching the revolving platform very lightly on the way down
EX5 27.01 27.10 5 Another very unexpected new strat for a nice time save
EX6 25.76 25.78 1 This was yet another of those floors where the previous time was considered maxed, but we wanted to prove it wrong
EX7 23.55 23.81 16 Another cycle based floor where going a bit later actually helped us preserve much higher velocity with an amazingly good setup
EX8 27.68 27.70 1 The unexpected frame saves go on and on. Aiai decided to disregard the fact that he is crossing over a gap, going into the goal without clipping
EX9 21.48 21.53 3 This floor has the longest flat straight surface in this run. Optimizing that simple platform allowed us to save these frames. The ending of this strat is very tough to pull off
EX10 26.76 26.91 9 A new way to get hit towards the goal at very high velocity was found. We managed to do it with a fancy new angle as well
M1 47.56 48.43 52 We found a very strange collision on the far-right corner of the moving blocks on this floor. This was abused along with very low clips from the corners of blocks to produce this insanely huge time save
M2 58.46 58.48 1 A frame was saved on this level by dropping off very slightly earlier than before, also making goaling barely possible
M3 54.41 54.45 2 We found a specific way to climb the hill that allowed us to preserve slightly more velocity, awarding us with these frames
M4 54.05 54.23 11 After abusing this level’s strange collision, we clipped high off of a platform and hit the back side of the rolled-up carpet on a frame where the angle redirected us straight into the goal
M5 57.55 57.58 2 We used a new strat involving very precise collision to get hit straight towards the goal rather than rebounding off of a platform on the left
M6 58.71 58.71 0 Getting this time was the most trivial task of the whole TAS
M7 58.56 58.58 1 We recently found out that an unknown Japanese player got this time in 2008 with an extremely lucky goal. Matching that time was very challenging, even in TAS
M8 57.50 57.56 4 We used a new strategy with an earlier hit into the air that makes it just barely possible to squeeze by the goal post
M9 58.23 58.25 1 Very similar to the previous TAS of this level, we altered the ending slightly to save this frame
M10 57.08 57.10 1 Similarly to the previous floor, we managed to improve the old setup with careful optimization

Known improvements:

In terms of in-game time, there are absolutely no currently-known improvements for this category, rendering the TAS to be optimal in terms of gameplay execution.

Around 19 frames were lost in TASVideos timing due to syncing issues when hex-editing past floors and an emulator glitch where the emulator randomly fails to detect the input device for a single frame. To fix the sync issues, we had to delay few level transitions by 1-5 frames (another way would be redoing months of work). To get around the input drop glitch, we had to visit the pause menu for 3 frames to force the input drop not happen during gameplay, which would have cost movement velocity. The glitch happened twice during the TAS (floors E9 and E46). These frame losses did not cost any in-game time, so it doesn’t affect the gameplay quality of the TAS.

Suggested screenshot:

Frame 4048

Samsara: Judging~

Samsara: This was a fantastic improvement! After watching the published run and this run back to back, the improvements were clear. As many have stated in the thread, Floor 24 was a particular highlight, but every other stage looked awesome as well.

Feedback has been unanimously positive, and it's all well-warranted. Technical quality is astounding, with well over half the stages being improved, even if it's just a frame or two. Therefore, accepting as an improvement to the published run!

Fog: Processing...

Similar submissions (by title and categories where applicable):