Submission #9244: Nami's N64 Tony Hawk's Pro Skater in 04:34.90

Nintendo 64
baseline
(Submitted: Tony Hawk's Pro Skater (U) [!].z64 NTSC 1.0)
BizHawk 2.4.2
16494
60
11605
PowerOn
Submitted by Nami on 8/22/2024 4:22 AM
Submission Comments

Game Objectives

  • Emulator used: BizHawk 2.4.2
  • Game version: NTSC 1.0
  • Aims for fastest time
  • Get to the credits
  • Obtain 26 tapes
  • Earn 3 bronze medals
  • (no abusing replay glitch)
Genre: Sport

General Comments

Metanotes

With some experimentation, the older version 1.0 saves a few frames in loading times. There are no differences known between this and the 1.1/Rev A that I used for my former any% TAS, aside from the title screen saying "Tony Hawk's Pro Skater" instead of "Tony Hawk's Pro Skateboarding," and a different interface on the screen to delete files. I refer to this run as 'any%' multiple times. Although this game has major skips possible that are banned outside their own categories (see the Replay Glitch run), the community still calls this run any% for legacy purposes. Even if you have a Memory Pak and/or Rumble Pak in either controller, there is always a message noting the potential use of them depending on which is in Controller 1, so there's no way to ever skip that part of the menu. The any% run skips 4 tapes over the whole game. Since speedrunning this game began, the optimal tape route was to skip 2 tapes in School (Grind 5 benches and Collect S-K-A-T-E) and 2 tapes in Streets (Destroy 5 Cop Cars and Collect S-K-A-T-E). That changes now, as this new TAS skips a single tape in School (Collect S-K-A-T-E), 2 tapes in Downtown (Destroy 5 Signs and Collect S-K-A-T-E), and a single tape in Streets (Collect S-K-A-T-E). Along with this TAS, I simultaneously made a new TAS using the old route. Both used nearly identical inputs on the remaining levels, and this new route beats the improved old route by roughly 1 and a half seconds. Current rules on SRC allow IL runners to End Run during a combo, which removes the timer but allows them to control the skater and obtain objectives until they land, counting the IGT as the time when they used End Run. I'm forgoing this strategy since it severely hurts the RTA time and is not truly 'faster' in any way. While this game usually runs at 30fps, due to odd-numbered lag frames offsetting it and some other quirks, I will refer to any specific framecounts in terms of 60fps.

Metagame Strats

There are some instances where the first and second controller can both manipulate the menus, which lets me save a few frames when I can alternate button presses between them every frame. It is worth switching controls to Autokick Off - not only does this setting let me accelerate faster at the beginning of levels, but it also removes a natural speed decay so it's easier to preserve momentum. It is also worth switching characters to Rune Glifberg. Despite the couple of seconds this takes, he makes up for it since he is tied for the highest base speed and his Christ Air trick has a high base score that both lets him move the fastest in many instances and also allows for strategies to get scores in the competitions faster than others. Another strat is soft resetting by holding Z+A for about 3.5 seconds to return the main menu. This saves time over traditional menu'ing when ending non-competition levels. Unfortunately, due to the nature of BizHawk, loading times are still extremely slower than they are on console. Soft resetting also gets to the menu from power on faster than any other method. With some searching, I managed to discover some values in memory tied to speed. It turned out that, generally, the higher the base score (i.e. before multipliers) of your tricks, the faster you go when you land (only in score increments of 100). This is not a perfect 1:1, however, as many tiny factors affect speed in this game, but it made it easier to estimate how best to optimize certain sections. Sometimes interrupting a grab is worse than letting go early even though holding it longer is more points, for example. Many points of this run require tight turns as well, and there are many instances where speed varies wildly over a few frames, so each and every input matters greatly. A tricky part of optimizing speed in this game is that higher scoring tricks tend to take longer, so it's not always apparent when it's worth taking a higher jump for them versus landing earlier. Add that tricks stale the more you use them, so Christ Air is not always the best choice after it's been thrown out a couple times already. Unless it's at too harsh of an angle, landing on a grind accelerates you. It decelerates you each frame you stay on it, so I almost always ollie off immediately unless I need to set up a specific jump from it. After landing a trick, there is a second or so where speed can vary a decent amount before settling to something consistent. I often use the tactic of finding the highest speed during this time and jumping to preserve that higher speed until I land again. Wallrides and successive grinds are another way to keep momentum up when possible.

Opening

I realized that there was an extra input available after the 1st soft reset. By shortening it by a frame, the soft reset still activated, and I could add an 'up' input on the second controller earlier here and it would work. I use the same improvement after exiting other non-comp levels to press 'A' on the second controller to progress through the main menu faster. I press 'B' on the same frame as 'A' to turn Autokick Off to exit the menu, and then 'B' the next frame on the second controller to buffer returning to the main menu as well. I hold 'R' while choosing the first level to set up a starting angle.

Stage by Stage Comments

Warehouse Woodland Hills

By delaying holding c-down til after at least a frame of turning, I delay crouching which would hurt my acceleration but can still start buffering an ollie. Landing a Nollie Madonna speeds me up as much as possible at this point, and makes it so I can take a lower jump down the roll-in to finish building my special meter and land on a steeper section. While it's possible to perform a Christ Air after this first grind, I would need a higher jump and that combined with the small extra speed makes it impossible to turn sharp enough to get the 2nd boxes and E while still setting up the angle to make it to the platform with the 3rd boxes. Jumping from this platform to get this S preserves a little more speed than falling off of it did, and sets up a better, lower jump for the transfer over the 4th boxes into the K. I use a Front-Back Kickflip for this gap since I have time and so Christ Air is worth more points over the taxi than it was in my previous TAS. While the Kickflip up to the 5th boxes is worth less than the old 360 Flip, it requires a shorter ollie so I can get back to the ground quicker and turn for the jump to the Hidden Tape. Ultimately, I saved 22 frames over my old Warehouse.

School Miami

In a similar fashion to Nahoc's 5:08 TAS, I grind the bench and then table in quick succession, each of which accelerates me. Unlike his, I grind the right side, which sets up a straighter path towards the wallride. I ollie off the wallride late so that I can go as low as possible while still landing on the awning and turning without sacrificing speed. I also ride the very right side of the awning, because each time I bounce down it, I get a speedboost from landing on the slope that's greater than if I were to ride and jump down it normally. Here's where we get to one of the aforementioned macro changes to the route, as I continue the level to grind all the picnic tables. I jump to the incline at the end of the awning just right to have the speed to grab the tape without ollie'ing off and land on the ramp below. Although the big drop here slows me down, the Kickflip McTwist and analog turn into a slope boost keep me going fast enough to the left to jump to the 2nd table. I forgo tricks in the grind sequence or I otherwise would be going too fast to connect the grind from the 3rd table to the ledge properly. While this section looks simple, it is one of the hardest too optimize since I need to turn around almost completely after the 3rd table. If I roll off, I'll slam into the wall before I can turn, but if I jump I end in the grass if I don't grind the ledge. I could also intentionally slow down before the grind, but this loses more time than taking the wider turn with the grind but preserving as much speed as possible. Once next to the 4th table, I can turn as tight as possible to get a good grind and set up a jump over the building. This ollie off the quarterpipe varies wildly - one frame earlier or later doesn't even let me transfer off the quarterpipe. I searched some slightly different setups to find the lowest transfer so I don't waste precious time in the air. This direct route over the building from table 4 to 5 is one of the main reasons this new route works better.

Mall New York

Unlike Warehouse, I crouch the first frame possible to turn sharper - otherwise I would move forward and fall off the left side of the slope, landing on the higher floor. This way, stuttering the turn lets me land on the rail below. Because of the way grinds, downhills, and landing tricks work, it's best to wait a little before jumping off the first grind to maximize speed. Up the escalator is the first instance of using a wallride to preserve speed longer. Unlike both my previous TASes which grind once then wait a frame and jump high to reach the floor above, here I jump low to grind a second time. Though I end up taking a wider berth, jumping over the platform here takes a more direct route to land further on the flagpole and ultimately saves time. After another big wallride up to the 2nd directory, I come to what may be the most difficult-to-optimize segment of this game. While my old any% and current 100% TASes may look similar, they all work rather differently, and since 100% has different stats at this point I cannot simply copy it. The old any% grinded the ledge, jumped to the right, then jumped through the water (slowing me down). The 100% TAS forgoes the grind to move slower but set up a better angle onto the grind below to stay out of the water longer and jump all the way to the Hidden Tape. This time, I painstakingly searched to find a way to grind the ledge, jump off it, bump on top of the ledge to move forward a bit, and land on the left side of the pool for a slope boost there while still having the room to jump to the grind. I couldn't find a way to jump all the way to the Hidden Tape this time, possibly because of the lower stats, but grinding the rafter may be faster anyway. Jumping to wallride after grinding the escalator helps preserve speed better when jumping to the flatter ground again than staying on the grind or jumping and regrinding does. Another wallride into the downhill grind prevents me from slowing down as much as I should have around the corner. While my previous runs took some interesting paths over the art structure housing the T, here I found it better to grind more and stay on the grinds longer to have better, faster setups out of them. I end up landing at the tail end of a ramp to get a slope boost before jumping to get the E. This way, I can avoid the ramps and water there which would slow me down. Considering this saved roughly half a second over my 100% Mall despite that one having better stats, I may need to revisit 100% and see if I can implement anything I learned here.

Skatepark Chicago

I use a faster trick before the wall here to stay lower. I delay pausing and ending the run so that I can crash into the wall lower, which lets me progress through the menu faster, saving a total of 31 frames over both heats compared to my old Bronze.

Downtown Minneapolis

Another major change, but since I only need the High Score, Pro Score, and Hidden Tape on this route, this level is fairly short. I'm holding soft reset before I even leave the grind. With the right speed and angle, I can fastplant wallride and grab the tape without even grinding the roof. It was fun to squeeze two special tricks into one combo.

Downhill Jam Phoenix

I had an issue where loading into this level gave me additional lag frames which messed up my beginning acceleration. I ended up delaying entering the level from the menu by a frame, and could then copy+paste the exact inputs from my 100% since the stats are the same at this point. It ended up being much faster at the beginning than anything I attempted with the detrimental lag frames, far more than making up for the single additional frame to load in.

Burnside Portland

By starting with a 360 Flip instead of a Madonna, I don't need to go high and wait to fall back down to grind - I can hit the coping on the way up. I turn right to be able to Crooked Grind instead of Boardslide. I use a Benihana before the next grind since it's the next fastest trick - any other grab would require a higher jump. The only faster tricks are 360 Flip (which I already used once so will be stale), Kickflip, and Heelflip, but they all offer significantly fewer points - not enough for this combo to get a high enough judge's score. I Smith Grind first and Crooked Grind second in the 2nd heat, since I need to stay on the latter grind an extra 2 frames to get a higher score and the Crooked gets more than the Smith. The order didn't matter the first heat since I used them both for a single frame each. I then jump higher at the end to have time for a Varial, which gives the most points here but usually takes too long to be useful elsewhere in the run. I jump left off the grind instead of right to land on the middle of the higher ramp after score canceling and ending the run for a faster menu. All this combined saves a grand total of 12 frames over both heats compared to my old Bronze.

Streets San Francisco

While I don't think I fully optimized my 3-tape route here (unsubmitted but it's on my personal channel in low quality - main difference from old run is skipping the roof climb and jumping from the planter), I don't think it's possible to improve it enough to make up anywhere near the ~1.5 sec difference it would need to beat this submission. Since I skipped 2 tapes back in Downtown and 1 in School still, I need to make up for it by destroying all the cop cars here. Manipulating my ollie buffering and timing by overturning left and then turning slightly back to the right, I can spin a 360 and hold a Madonna long enough to get my special meter filled without going too high or fast over the cop car which would avoid destroying it. I use a grind, slope boosts, and a Christ Air to max my speed, then turn a corner and jump to the 2nd cop car. I land on it so I can jump high enough to land on the ledge next to the handicap ramp, then use a series of jumps and grinds to sprint to the next 2 cars. With a Christ Air and a particular turn, my ollie gets interrupted with me landing on the roof of the 3rd car. This slows me down just enough to trigger the destruction of the 4th car without touching it. The slower speed is also good since I need to turn around now, where I hit another ledge to grind a couple times, then spend very precise jumps for the next setup. After maintaining as much speed as possible while moving to a good position for it, I turn and Fastplant Christ Air landing just before the planter. Blink and you'll miss it - I Fastplant up and onto the ramp of the planter, then do a late Fastplant off of it. It's hard to distinguish the first from the second without slowing down the footage. Thanks to the setup of speed and landing the Christ Air, I'm able to jump over the pagoda and grab the Hidden Tape during a McTwist. I level out using the analog stick to retain some speed, but quickly Christ Air to regain more. By using a lower ollie to grind one of the lower steps, I can quickly grind the higher one in succession. Although the structure here is awkwardly designed so that I have to keep jumping left to right, the angles end up working out just perfectly to propel me towards the final car. Normally, speed decays on a grind, but with everything leading up to this my speed is maxed and stays consistent for now, so I stay on the grind as long as possible and jump as far off it as I can. A slight angle adjustment just before I land gets me to hit the last cop car as soon as possible.

Roswell New Mexico

I turn and jump just high enough to spin a 360 and hold Madonna long enough for a full special bar and land on a slope boost. Another short combo on the rail can be topped off with a Christ Air now that I have special to give me enough points for bronze. I have to stay on the coping for a couple extra frames during the Smith Grind, but this isn't for a higher score but to set up the angle for afterward. That is to land in front of a rail that I can trip over by inputting a Tailgrab while I'm still in the air after I pause and End Run. Like Skatepark, falling here shortens the time for the menu to appear. I saved 5 frames over my old Bronze.

Closing Notes

Along with the new route and improvements to parts of the old one that remained the same, I saved somewhere around 10 seconds over my old any% run, which feels amazing given how short and optimized this game already was. Not long after I finished my last TAS, BSGrind found a couple ways to jump from the planter over the pagoda to get the Hidden Tape. I and a couple other users experimented with a couple others. It's been over 3 years ago since I first theorized this route with the 4 cop cars, but I worried that the new jump would also help the 3-tape run too much for this to be worth it. The more I theorized and tested, the more it was clear that both runs would be within about 1 second of each other. The only way to know for sure was to create 2 entirely new TASes: one using the old route with improvements, and one with the new route plus those same improvements. That task was daunting, but I'm proud to have finished it. Even if it proved worse, I still want to say: always look for new ideas. Never assume that common knowledge is absolute just because it's always been that way. This run has seen some incredible changes the last few years, even though it's a 25 year old game, and it's mostly because I and some others decided to try things no one had before. Thank you for reading and watching!

nymx: Claiming for judging.
nymx: Removing the goal, since it is implied. Also, thanks to McBobX for the encode.
This is very good work, over your previous publication. There sure are a lot of mechanical exploits to take advantage of.
fsvgm777: Processing.
Also replacing the movie file with one that removes 107 frames of blank input at the end.
Last Edited by fsvgm777 on 9/6/2024 10:51 AM
Page History Latest diff List referrers