Submission #5203: jagg2zero's N64 F-Zero X "Jack Cup, Time Trial" in 07:09.37

(Link to video)
Nintendo 64
Jack Cup, Time Trial
BizHawk 1.11.6
25762
60
61745
Unknown
F-Zero X (USA).z64
Submitted by jagg2zero on 8/23/2016 8:41:20 AM
Submission Comments
F-Zero X is a futuristic racing game released in 1998. It is known for its high-speed gameplay, large vehicle roster and ability to maintain a constant 60fps even when racing with 29 other AI opponents on any track at any speed.
The goal of this TAS is to beat the first 6 tracks in Time Trial mode as fast as possible according to the in-game timer using every glitch and exploit currently known. I used Mupen64 0.5.0 re-recording v8 and converted the movie to BizHawk 1.11.6. Because this TAS wasn't done in BizHawk, the re-record count wasn't carried over during the conversion; BizHawk displays 229 re-records while the actual TAS took 61,745 re-records. An in-game cheat code (L, Z, R, C-Up, C-Down, C-Left, C-Right, Start) is used just after the title screen to instantly unlock all machines and courses so that there's no need to go through GP mode to unlock everything.
This TAS saves 1'10"614 (in-game track totals) over the previous Jack Cup TAS by Lord Tom #3815: Lord_Tom's N64 F-Zero X "Jack Cup, Time Trial" in 08:17.72. Below is a table showing a comparison for every track's time:
CourseLord Tom's TASMy TASDifferenceOther TASes
Mute City1'03"0740'33"214-0'29"8600'50"993
Silence0'59"7570'57"962-0'01"795
Sand Ocean0'48"5700'29"943-0'18"627
Devil's Forest1'02"4230'52"609-0'09"814
Big Blue0'44"9640'39"685-0'05"279
Port Town0'58"9500'53"711-0'05"2390'58"858
Total5'37"7384'27"124-1'10"614
Lord Tom made an improved TAS of Mute City with a time of 0'50"993; The difference between our latest TASes is now -0'17"779.
WMJ made a Port Town TAS that's slightly faster than Lord Tom's with a time of 0'58"858.

Techniques:

Here's a review of all the currently named advanced techniques used in this TAS. For the most part, it is not fully known which techniques are glitches and which are intended mechanics.

Sliding

By deliberately losing grip while cornering, the machine will slide sideways around the corner. If the machine has neutral acceleration/max speed settings, sliding around the corner will not change the machine's speed too much while machines with settings closer to max speed will lose speed while sliding. If the machine has acceleration settings, however, the machine will actually gain speed while sliding. Because of this, sliding is best done with machines that have maximum acceleration settings and low/minimal grip. This technique can be combined with a boost, turning it into a boost-slide. A slide/boost-slide can also be done while airborne.

Railslide

Even though machines with maximum acceleration settings can quickly gain speed by sliding around corners, they are consequently slower while driving in a straight line. However, on (straight) roads that have rails, it is possible to gain the speed benefit of sliding at the cost of energy by constantly crashing into the rail at a 45 degree angle or thereabouts. This technique can be combined with a boost, turning it into a boost-railslide.

Double-Tap / DT (or DT-R / DT-Z)

A side attack can be executed by quickly pressing R or Z twice. Normally used to damage other F-Zero machines, side-attacks are also useful for very sharp turning and they're a crucial element for the technique below.

Double-Tap Dive / DTD (or DTD-R / DTD-Z)

A series of DTs used in the air while the machine is pointing downwards and at a tilted/slanted angle. Used for gaining very large amounts of speed well beyond the capabilities of conventional methods.

Air-Ground Glitch / AGG

By landing on a rail and then leaving the track, the game thinks the machine is on the ground while it is actually in the air. This grants the machine grip while in the air which can be used to fly upwards to pretty much no end by tilting the machine on its side and pointing the nose upward. The typical speeds a machine would travel at while AGG floating are 550-700km/h.
The general rule is that the more extreme of an angle the machine is facing while in AGG, the faster it will gain height. However, if the angle is too extreme, the machine's grip will break, and it will start to plummet towards the ground as a result.
Machines with higher grip ratings can be pushed to a more extreme angle before it "stalls", so they are generally more preferable to lower grip machines in tracks with lots of AGG abuse. Arguably more important than grip is the machine's weight: The lighter the machine is, the faster it will gain height and vice versa.
While in an AGG float (or on the ground), the A button can be released for a frame to gain a large amount grip for 0.5 seconds without sacrificing any speed whatsoever. This is useful when preparing to change directions while AGG floating because the machine's grip is guaranteed to break if it makes any sudden movements while floating at an extreme angle. It is also useful for pushing the machine into a more extreme angle than it can otherwise handle if there is a need to gain height very quickly at the cost of covering horizontal distance and eventually some efficiency.
The general purpose of AGG is to gain a large amount of height, then perform a massive DTD. In some (but not all) places, this can save a lot of time in the long run.

What's new:

Most of the improvements from Lord Tom's TAS come from heavy exploitation of a specific element of the Double-Tap Dive. This element is technically as old as the DTD technique itself and has been used to a smaller degree by many real-time runners (including myself), but the potential it had to cut down time (at least in a TAS) was only discovered somewhat recently.
If the machine is rolled onto its side at a ~90 degree angle while in the air and the nose is pointing upwards, DTing will cause the machine to gain height AND speed, though only for a short while and only in certain places. This can help set up for a DTD that will travel much faster and further than previously thought possible.
This is where it gets better/broken/worse: If this technique is used while the Air-Ground Glitch is activated, the amount of height and speed that can be gained as well as the amount of time this technique can be used are all increased massively! Now, speeds of up to ~2600km/h can be gained/sustained (the typical speeds being 1600-2100km/h), all while gaining more than enough height to do a follow-up DTD that easily reaches 3000km/h!
This technique seems to take advantage of the exact same type of "physics" that make Double-Tap Diving possible except the machine isn't actually diving, but ascending. Since this technique hasn't been (formally) named yet, I'm referring to it as a "Double-Tap Ascend" / "DTA" for the time being (yes, yes, very creative naming on my part).
For several tracks (most notably Mute City and Sand Ocean), I end up shaving off colossal amounts of time by going through this 3-4 step cycle:
  1. DTD onto a rail.
  2. Leave the rail to activate AGG.
  3. DTA to gain lots of height and maintain as much of the previous DTD's speed as possible.
  4. Repeat 1-3 as much as possible.
Even in tracks where AGG isn't used like Big Blue or Silence, Double-Tap Ascending is still useful for shaving off "merely" large amounts of time.
What I'm trying to say is that this new technique is absolutely massive.

The tracks:

Mute City

Machine Choice: Twin Noritta (Body: E. Boost: A. Grip: C. 780kg).
Reason: The Blood Hawk's low grip is best suited for tracks with lots of sliding and the Twin Noritta's low weight and moderate grip are best suited for tracks with lots of AGG abuse. In this new strat, the first ~3.6 seconds are dedicated to railsliding and the rest are dedicated to AGG abuse, so the TN is is now a lot more preferable to the BH.
It is now faster to drive backwards at the very start of the run and set up an AGG chain at the end of the road despite the fact that about 8 seconds (or ~1/4 of the entire run) is spent behind the finish line, technically not even on the first lap yet.
I used a more efficient loopskip that is faster to set up, faster to dive through, has a great setup for the next AGG chain and is even console-friendly.
I do a backwards-facing DTD for the final lap's loopskip because it has a faster preparation and it reaches 3000km/h quicker thanks to a sharper diving angle. The only problem with facing backwards is that once I'm on the ground, I lose speed quickly due to the game's lack of a reverse gear/button but since there's a very short distance to cover between the landing and the finish line, the time lost is much smaller than the time saved. A win-win for entertainment and speed :D

Silence

Machine Choice: Twin Noritta (Body: E. Boost: A. Grip: C. 780kg).
Reason: Settings close to max speed (aka "Jumper" settings) allow me just enough speed to leave the road and DTD on lap 1, whereas DTDing on lap 1 is impossible with max acceleration / "slider" settings. The TN is considered to be the best machine to use with jumper settings due to its light weight, unmatched acceleration, reasonable grip and A-level boost.
There are no viable AGG opportunities for 3lap runs here for a quite a few reasons, the biggest one being most of the track is either sideways or upside-down making AGG chaining impossible. Even if that were not the case, setting up a good AGG-chain would take an almost comically large amount of time. So it's mostly normal driving on this track.
I saved ~0.22 seconds by optimising lap 1 in a few places and by squeezing more speed out of the first DTD. But thanks to this extra speed from the first DTD, I was just barely able to add in a second DTD. This second DTD is initially 0.20 seconds slower, but the trade-off is that I now start lap 2 traveling 116km/h faster, which ultimately saves me ~0.15 seconds across both laps 1 and 2 over not using the second DTD, bringing the total time saved so far to ~0.37 seconds.
The main timesaver, however, is the addition of a(n AGG-less) DTA in-between the first and second DTDs at the end of the second lap. Since the part of the track that I flew off is heavily angled this way / , the machine is rolled onto its left side once I become airborne, which is a great position to use a DTA. If you look closely, you'll see that I gain height and speed simultaneously (1725km/h to 1934km/h). This setup results in a higher, faster and deeper DTD that saves over a second.
During the 51-53 second mark, I drove in what looks like a fairly sub-optimal manner, but this was to stop the machine from completely flying off the road (this would be fatal to the run) while preserving as much speed as possible between the tunnel and the final DTD.
Finally, I crashed into the wall at the very end of the run to make the machine explode while crossing the finish line. This doesn't save or lose any time; it's simply there for entertainment purposes.

Sand Ocean

Machine Choice: Twin Noritta (Body: E. Boost: A. Grip: C. 780kg).
Reason: Same reasons as Mute City 1 but with even less railslide time.
Like Mute City, this track gets obliterated. This time with 2 AGG chains per lap. Barely any time is spent on the track, corners get cut and the average speed is through the roof here! This is also the first ever track in F-Zero X to get a sub 30 second run without the use of a speed code/cheat/hack!
One small thing to note is that during the last DTD of lap 2 (at 21"92 to be exact), I deliberately pause for a frame before starting the last DT. This briefly causes me to go from 3000km/h to 2976km/h which, even though I made it back up to 3000km/h very quickly, is sub-optimal in most cases but by starting the next DT a frame later, I end up landing on the road at a slightly better angle and speed which ultimately saves me a little bit of time in the long run.
I also did a suicide finish at the end of the run which, in this case, is slightly faster than trying to land onto the road.

Devil's Forest

Machine Choice: Blood Hawk (Body: B. Boost: A. Grip: E. 1170kg).
Reason: The TN is bad for sliding thanks to its higher grip and it has the weakest body for the boost-railslides. Even though AGG abuse kicks in at lap 2, it would actually have worse boostlaps because it would lose too much energy from landing on the rails while AGG chaining. Even if that weren't the case, the 1st lap disadvantage far outweighs the 2nd-3rd lap advantages.
Hooooo boy. This track used to be all about clean driving with careful energy management. Now it too falls victim to AGG chains.
The only points where the machine can leave the track and activate AGG are the 2 hills just before the finish line. Unfortunately, it's not possible to leave the track without boost power so as soon as I trigger lap 2, I turn around and boost-railslide to activate AGG further down the road. This is now the 3rd track (out of 4 so far!) where I drive the opposite way at some point to start an AGG chain.
On the outside part of the 2 hairpin corners, there is a brown strip of dirt. Driving through this dirt will slow the machine down. In the air, this is not a problem since the dirt a ground-based hazard, but thing about AGG is that the game thinks the machine is on the ground even though it's in the air. Therefore, while AGG is active, all beneficial/hindering ground-based hazards affect the machine either when it flies over them, or when it is flying past the side of the road while there is a hazard on the edge of the road (basically, the hazards extend horizontally away from the road if they're on the edge of the road). The exceptions to this rule are mines, jump-plates and other F-Zero machines because they're physical objects rather than road segments with special properties.
Thankfully, being in a boosted state cancels out the dirt-zone's effects and I don't mean "evens out" the effects, but straight up nullifies it. So at the beginning of lap 2, I boost twice while I'm near the first hairpin corner to prevent speed-loss from the dirt strip even though boosting while DT/DTDing in itself does pretty much nothing in the ways of increasing speed. The dirt isn't a problem for the second hairpin or for the entirety of lap 3 since I don't have AGG activated during those corners.

Big Blue

Machine Choice: Twin Noritta (Body: E. Boost: A. Grip: C. 780kg).
Reason: The only real advantage that Blood Hawk has on Big Blue is better sliding at the very beginning...which is more-or-less canceled out by the Twin Noritta's faster acceleration thanks to its lighter weight. Speaking of weight, the TN has better/sharper aerial maneuvering and more efficient floating than the BH since there is less weight to move around.
I made some improvements in lap 1 by starting the first 2 DTDs a bit sooner than usual. Normally, starting a DTD too soon can be slower in the long run, but I was able to minimize or completely eliminate any long-term speed loss. Also, instead of boostsliding off the very end of the cylinder, I DT/DTA off of it at a specific angle so that I can get a much better DTD, similar to what I did at the end of lap 2 in Silence 1.
For laps 2 and 3, I have a few different strats to choose from (using the first lap's strat as a baseline):
  1. Swap the 2nd DTD for an aerial boostslide, land on the cylinder while still sliding, leave the cylinder and DTD onto the road.
  2. Repeat the 1st lap's strat.
  3. Swap the 2nd DTD for a boostslide, but this time, float a bit higher and then cancel the boostslide, DTD onto the cylinder, then use the extra speed to do a larger final DTD than lands halfway into the chicane.
Strat 1 was used in the previous TAS, but it's now the slowest strat of the 3 to choose from.
Strat 2 would have been the slowest, but thanks to recent changes, it's now slightly faster than the above option.
Strat 3 is what I went with in the end for both laps. Not only does the final DTD save way more time than it loses through its setup, I end up starting the 3rd lap with way more speed than I did in the 2nd, saving even more time! The only downside is that I can't refill my energy meter in the due to skipping over the only road strips that can refill it. Thankfully, not only is the amount of extra energy lost is so tiny that it's barely worth mentioning, but the base amount of energy I already had was pretty sufficient despite the fact that I only had enough energy to cut the final corner in the 3rd lap instead of both laps 2 and 3...
Like Silence, I crashed into the rail at the end to make the machine explode purely for entertainment reasons at no cost/gain to time.

Port Town

Machine Choice: Blood Hawk (Body: B. Boost: A. Grip: E. 1170kg).
Reason: Same reasons as Devil's Forest except the AGG abuse starts a bit sooner.
Sadly, Crazy Bear isn't used this time around because the much, much lighter Blood Hawk is better suited for the new AGG strat. There are two points towards the end of the track where the machine can lift off the track and activate AGG:
  1. The high jump.
  2. The small bump on the right side of the road before the refill area.
I went for the high jump because if I chose the small bump, I would end up slowing down a lot on the straight before I would reach the next corner. DTAs need corners to work, after all.
Strangely, yet hilariously, it's faster to do the entire AGG chain with the machine facing backwards! Video comparison between forwards and backwards:
https://www.youtube.com/watch?v=y4FyVwAU83A
As for why backwards is faster...
Reason #1 is that the setup for the first backwards-facing DTD is a lot smoother and quicker than the forwards-facing DTD: The forwards DTD fails to break 2500km/h due to the poor diving angle, while the backwards DTD can reach ~2800km/h and is already nearly 0.2 seconds ahead of the forwards run just after the landing.
Reason #2 isn't actually something in the backwards run's favor, but it does address a potentially huge problem: speed retention. Normally, simply coasting is always going to be way slower facing backwards than forwards because the only speed-gaining force in that case is the accelerator / A button. Eventually, the backwards-facing machine will stop completely and the forwards-facing machine will have a resting speed of a few hundred km/h. However, in this case, both the forwards and backwards-facing machines are either performing a DTD or a DTA at all times where 95% (figuratively speaking) of the speed-gaining force comes from the DTs instead. While this means that some speed retention is lost over time in a backwards-facing run, it does mean that a backwards-facing AGG chain is completely viable (and in this case, faster than a forwards-facing AGG chain due to the generous head-start).
Reason #3 explains the last huge problem: turning back around. Normally, turning around to face forwards again results in a colossal time and speed loss. This wouldn't be a problem if the AGG chain loops endlessly. However, an endless backwards-facing AGG chain loop isn't possible in Port Town because the 3 hairpins in the middle of the track are too sharp and successive, so I need to dive back onto the track at some point. The best strat that I found for both the forwards and backwards runs involves killing all of my speed on the 2nd hairpin just before diving back onto the track, which, due to the 180 degree nature of hairpins, is a fantastic place for the backwards run to turn around and integrate any problems of losing speed into a required part of the strat. Not only is the problem of turning around completely gone, the backwards strat is actually ~0.18 seconds faster than the forwards strat in this particular segment.
Having enough reasons in favor of a backwards-facing strat to the point where it's faster than a forwards-facing strat is very rare. It's entirely possible that there are other tracks that have faster backwards-facing strats lasting more than a single DTD, but it's also possible that Port Town is just an anomaly.
Finally, it's also worth noting that the AGG strat only saves 2-2.5 seconds per lap since I lose 1-1.5 seconds setting up the AGG chain and the track shape makes AGG exploitation difficult and slower than usual. So while Port Town gets broken like Mute City, Sand Ocean and Devil's Forest, the non-AGG strat can still produce a half-decent time in comparison to the AGG strat.

Thanks to:

Lord Tom – for all of his hard work on the previous TAS. His work serving as inspiration and as a fantastic benchmark despite new game-breaking tech rendering most of it obsolete.
The F-Zero Community – for...being the F-Zero Community, I guess :P Seriously, their dedication to these games, their articles, guides, videos, discovery of new techniques/strats and everything else is pretty amazing and helped in shaping me into being a pretty good F-Zero X real-time runner and TASer (if I do say so myself).
Anyone who makes podcasts and/or music (even if I don't actually listen to yours) – ensuring that I only have to listen to stuttering rock music and engine sounds or the sound of complete silence <5% of the time.

Samsara: File replaced with a version that contains the correct re-record count mentioned in the submission text. No changes were made to the input. Also, judging.
Samsara: Today I learned that the F in F-Zero stands for Flight Simulator. Accepting as an improvement to the published run. Also, jesus holy hell what has F-Zero X become that was incredible.
Note to publishers: Just make sure BH doesn't lose focus while the run is playing, since it seems to crash. As long as BH remains the focused window, it will properly sync to the end.
feos: Pub.
Last Edited by adelikat on 10/19/2023 12:25 AM
Page History Latest diff List referrers