Table of contents
Introduction
Ninja Gaiden is a game that invites you to go full violence: you get a sword, you get weapons, you break things, you kill enemies. However, it is also possible to beat the game without harming anything but the end-of-stage bosses. Even the final boss --the Demon-- can be treated nicely, only killing it by hitting its heart, and not his head, tail or shrimps.
This movie accomplishes all of the above and it does so at maximum speed. The result is a highly RNG and physics-manipulated movie that optimizes global routing and local movements in attempts to save every possible frame until the very end.
Authorship
The authors of this movie are listed in the order of how many saved frames they contributed. Although this is, in some cases, an approximation, the submitters believe it reflects pretty well their respective contributions. Some of the authors are credited in absentia.
Category Rules
The authors define the pacifist category as follows:
- No Enemies, except for bosses, shall be damaged
- No Scenery shall be damaged
- No Projectiles shall be damaged
- Demon head, tail, and all its 'shrimps' shall remain untouched.
Note: the first version of this movie was made under a stricter rule, where the button B cannot be pressed (except for Up+B which is innocuous) during non-boss stages. This rule had to be discarded when resyncing to BizHawk, as pressing B was the only way to fix some occurrences of NES's DPCM glitch.
Story Time
This movie has story that is almost a decade long:
Around 2011 I saw sinister1’s RTA attempts at NG1 “pacifist” and got inspired, and started supporting this category on the forums. In 2012 out of boredom I started TASing it, but wasn’t sure if I’m interested enough to keep going.
jprofit22 and Marx helped me which made me finish the first complete test run. While working on redoing that, I started reverse-engineering the problematic aspects of the game which resulted in scripts on GameResources/NES/NinjaGaiden.
MESHUGGAH joined and quickly became the god of the category, I was unable to improve his times anymore, so I just kept debugging.
I discovered one of the most important aspects of the game for this category: how you can delay spawns by moving back and forth at certain spots, completely despawning some enemies if you want. It affected the behavior of random enemies, which in “pacifist” is critically important to manipulate, and you can even clip through enemies and projectiles without taking damage.
MESHUGGAH kept grinding for a couple years and made his full test run which was 1169 frames faster than mine.
After redoing things even harder and saving 68 frames over that up to 6-1, MESHUGGAH decided to significantly change the rules of the category but hasn’t released any WIPs since then.
Since I was unable to contribute in any way anymore, I just kept waiting for him to get inspired to work on it again, but it ended up not happening. And then [4751] NES Ninja Gaiden by eien86 & Scumtron in 10:50.33 happened. And I thought it was my chance to see this thing finally done, and I was right!!!
The story concludes in July 2022 as follows:
My work on this category started shortly after Scum and I were done with the any% variant, having improved that one by a single frame. feos had asked me if I was planning to do the pacifist run, pointing me to the unfinished work by MESHUGGAH. It seemed like a good choice because I had all the knowledge for NG1 already and the bot ready to go. Little did I know that pacifist is a whole different monster. I can divide this adventure in two stages:
At first, I assumed I could just re-do the entire movie, only using the MESHUGGAH's movie as reference. After a week of painful progress, winning and losing many frames to RNG and enemy manipulation, I ended up on stage 3-1 15 frames ahead (see: https://tasvideos.org/UserFiles/Info/637933728969148329). This is where my progress came to a halt. After two days of work, I could not make 3-1 work without losing all 15 frames. At this point I decided that I was tackling this project wrong, and decided to pursue another strategy.
The second stage consisted of me figuring out that my job was to finish the previous work, not to create something anew. After this realization, the project became fun again and I could dedicate my efforts to completing the last few stages without pressure. Even then, I managed to find many (supposed) improvements to the way these stages were solved. What I'm proud of the most, is the fact that I totally squeezed the Ninja HP stock (a precious resource), ending Stage 6 with only 1 HP left. This strategy could have backfired, but in the end avoiding death saved a ton of frames compared to previous attempts.
This is by no means a perfect movie. Its main goal is to represent a highly optimized NesHawk-resynced baseline for future attempts. I am 99.9% convinced that many more frames can be saved without too much effort (other than resyncing). So this movie is an invitation to both old and new players to hone their TASing skills and see if they can beat my execution of 6-1 onwards, or MESHUGGAH and company's execution, 6-1 backwards.
The current work starts from the WIP movie from September 2016, and improves it on the following aspects:
- It was resync from FCEUX to QuickNES prior to re-routing
- Found a faster kill on Stage 5's boss
- Re-planning, and re-optimizing all stages from 6-1 onwards
- Re-worked Demon kill, where Ryu clips inside it and damages only the Demon's heart.
- Work on entertainment factor, including Train Mode in 4-2, post stage 5 and 6 boss kills
- A final resync from QuickNES to NesHawk for increased emulation accuracy.
Emulation
Rom Information
- Name: Ninja Gaiden (U) [!]
- SHA1: CA513F841D75EFEB33BB8099FB02BEEB39F6BB9C
- MD5: 0C2CCCDA6BA6CAB7BEDE0FF05E7F6852
Emulator
- EmuHawk 2.8.0 (Core: NesHawk)
Resynchronized from the initially submitted [EmuHawk 2.8.0 + QuickNES] movie, using Scumtron's DPCM glitch detection tool, available here.
Routing Bot
- Bot: Jaffar
- Routing Core: QuickNES (Average Exploration Performance: 1.6M States/s)
- Platform: AMD Ryzen Threadripper 3990X Processor (64 cores, 128 threads) + 256Gb RAM
Timing
The game is structured into acts representing the story arcs in the game with bosses at the end; stages, the different playable parts of an act, and; substages, the division of a stage into multiple NES vertically stacked 'screens' to allow for the illusion of verticality within a given stage. The nomenclature goes as follows: III-2a, represents act 3, stage 2, substage a. Timing is structured as follows:
- Substage timing starts from the very first frame where Ryu is playable (typically 2 frames before he appears on screen)
- Substage transitions happen when the substage number (0x006E) changes.
- Stage transitions happen when game mode* (0x002D) has a non-zero value. Alternatively, the preserved value of the A register at the start of the NMI routine that happily serves this purpose quite well.
- Act transitions happen when a Boss HP (0x0066) equals 0.
Since there is no preceding movie, we compare our movie against use a combination of MESHUGGAH's et al. best work in progress files:
- We ignore the first frame difference, since this is attributed to using different emulators
- For Acts I-V, we use their last partial WIP
- For Act VI, we use their previous full WIP
This is not an ideal comparison, but gives the best impression of the improvements made in this work.
New Old Diff
Stage Initial Total Initial Total Stage Total
Boot 0 173 0 174 -1 0
I 1 a 173 2028 174 2028 0 0
Transition 2201 277 2202 277 0 0
2 a 2478 198 2479 198 0 0
Transition 2676 740 2677 740 0 0
II 1 a 3416 973 3417 973 0 0
Transition 4389 23 4390 23 0 0
b 4412 814 4413 814 0 0
Transition 5226 22 5227 22 0 0
c 5248 97 5249 97 0 0
Transition 5345 22 5346 22 0 0
d 5367 604 5368 604 0 0
Transition 5971 277 5972 277 0 0
2 a 6248 652 6249 652 0 0
Transition 6900 22 6901 22 0 0
b 6922 614 6923 614 0 0
Transition 7536 52 7537 52 0 0
c 7588 1396 7589 1396 0 0
Transition 8984 277 8985 277 0 0
3 a 9261 191 9262 191 0 0
Transition 9452 741 9453 741 0 0
III 1 a 10193 1353 10194 1353 0 0
Transition 11546 278 11547 278 0 0
2 a 11824 2066 11825 2066 0 0
Transition 13890 277 13891 277 0 0
3 a 14167 93 14168 93 0 0
Transition 14260 744 14261 744 0 0
IV 1 a 15004 1484 15005 1484 0 0
Transition 16488 23 16489 23 0 0
b 16511 75 16512 75 0 0
Transition 16586 22 16587 22 0 0
c 16608 408 16609 408 0 0
Transition 17016 161 17017 161 0 0
2 a 17177 2024 17178 2024 0 0
Transition 19201 277 19202 277 0 0
3 a 19478 118 19479 118 0 0
Transition 19596 22 19597 22 0 0
b 19618 622 19619 622 0 0
Transition 20240 22 20241 22 0 0
c 20262 789 20263 789 0 0
Transition 21051 22 21052 22 0 0
d 21073 572 21074 572 0 0
Transition 21645 277 21646 277 0 0
4 a 21922 39 21923 39 0 0
Transition 21961 747 21962 747 0 0
V 1 a 22708 2053 22709 2053 0 0
Transition 24761 278 24762 278 0 0
2 a 25039 459 25040 459 0 0
Transition 25498 22 25499 22 0 0
b 25520 465 25521 465 0 0
Transition 25985 622 25986 622 0 0
c 26607 -113 26608 -113 0 0
Transition 26494 22 26495 22 0 0
d 26516 516 26517 516 0 0
Transition 27032 277 27033 277 0 0
3 a 27309 411 27310 411 0 0
Transition 27720 23 27721 23 0 0
b 27743 462 27744 462 0 0
Transition 28205 23 28206 23 0 0
c 28228 427 28229 427 0 0
Transition 28655 23 28656 23 0 0
d 28678 463 28679 463 0 0
Transition 29141 23 29142 23 0 0
e 29164 219 29165 219 0 0
Transition 29383 22 29384 22 0 0
f 29405 204 29406 204 0 0
Transition 29609 22 29610 22 0 0
g 29631 235 29632 235 0 0
Transition 29866 157 29867 157 0 0
4 a 30023 189 30024 204 -15 -15
Transition 30212 741 30228 738 +3 -12
VI 1 a 30953 1434 30966 1366 +68 +56
Transition 32387 277 32332 277 0 +56
2 a 32664 1054 32609 1040 +14 +70
Transition 33718 22 33649 22 0 +70
b 33740 629 33671 704 -75 -5
Transition 34369 23 34375 22 +1 -4
c 34392 75 34397 306 -231 -235
Transition 34467 22 34703 22 0 -235
d 34489 465 34725 491 -26 -261
Transition 34954 23 35216 23 0 -261
e 34977 602 35239 612 -10 -271
Transition 35579 277 35851 277 0 -271
3 a 35856 437 36128 457 -20 -291
Transition 36293 22 36585 22 0 -291
b 36315 133 36607 133 0 -291
Transition 36448 23 36740 23 0 -291
c 36471 1333 36763 1345 -12 -303
Transition 37804 22 38108 22 0 -303
d 37826 793 38130 837 -44 -347
Transition 38619 158 38967 152 +6 -341
4 a 38777 82 39119 82 0 -341
Transition 38859 744 39201 747 -3 -344
b 39603 103 39948 102 +1 -343
Transition 39706 744 40050 744 0 -343
5 a 40450 230 40794
End 40680
You can watch a side-by-side encode of the table above here:
Technical Aspects
Tricks and Glitches
This category is fundamentally different from any%. In particular, it uses the following techniques extensively, which are seldom employed in the any% movie:
- Enemy Spawn Delay (ESD) trick: This one consists of pressing L/R alternatively just before an enemy appears. By doing this, the screen keeps scrolling, but the procedure that checks whether an enemy/item must appear is not checked. Therefore, one can use this to 'delay' an enemy spawn by a few pixels/subpixels. However, if done for enough frames, it is possible to outright prevent the appearance of said enemy. This technique allows clearing up the path out of hard-blocking enemies, without having to kill them.
- RNG Manipulation: The game contains a global timer which keeps advancing through stages. This means that delaying pressing S during a cutscene can make it possible to start the next stage with a more favorable RNG.
- Enemy Slot Manipulation: The game allows for having a maximum of 8 concurrent enemies on screen. When a new enemy spawns, most of its attributes are reset to its proper values. However, some values are not refreshed, leading to some crazy effects like Joecurr's pre-kill. Another value that is not refreshed is the enemy's X subpixel, which is then inherited from the previous enemy to the new enemy occupying the same slot. By altering the order in which enemies spawn or de-spawn, it is possible to assign new enemies a favorable X subpixel. This is a crucial aspect in many skips in this movie and any future movies in this category.
Acts I-V (excluding boss) Breakdown
For resources about how these acts were resolved, I suggest reading the following resources:
- User file history for Ninja Gaiden. Look for comments on files by MESHUGGAH, with the 'pacifisting' keyword.
- Ninja Gaiden forum thread.
- Ninja Gaiden game resources.
The submitted movie differs from the WIP, in that the Act V boss is killed 15 frames faster, by not waiting for a new jump cycle.
Act VI Breakdown
This act requires a detailed HP-usage routing. Ryu starts with 16 HP which is lost every time you get hit. Different enemies/projectiles remove different amounts of HP. Most enemies remove 1 HP, but birds in particular remove 3 HP (this is unfortunate because bird-skips are very useful).
Instead of planning HP usage, the previous WIP contained a purposeful death in VI-2c to replenish it. This loses a couple hundred frames. Instead, my approach was to save as much HP from the beginning (VI-1a) where I lost around 60 frames compared to the WIP. It turns out that this strategy paid off since I was able to get most skips without needing to abuse death.
Future work
This movie is pretty much a baseline for future TASers wanting to improve this category. Anyone aiming to do that should choose one of three approaches:
- Start from the beginning, using the improved movie as a starting point. The difficulties to face here is that of resyncing 4-1 onwards. This is by no means an easy task, as the amount of work is maximized. However, this approach is the one with the most potential reward. You will need to use Scumtron's bird manipulation tool to solve 5-1.
- Start from the ending backwards. This approach will seek to improve on my contributions (eien86). Although I used a bot to refine the execution, there is surely a lot of potential in routing the global strategy and HP management. This is possibly the easiest way to find an improvement and publish an obsoleting movie.
- Starting from 6-1 backwards. This approach will seek to improve MESHUGGAH et al. execution on the latter stages. Any frame found here would have a moderate amount of resyncing work, for the latter stages.
Acknowledgements and Attributions
- MESHUGGAH, feos, Marx, jprofit22: The original work until stage 5's boss.
- eien86: Stage 5's boss onwards. Reused some parts of the original work.
- Scumtron: For being the one constant contributor to this movie throughout the years.
- The entire Ninja Gaiden speedrunning community: who give their support and are always ready to help and contribute.
Encoding Info
Encoder Movie
The re-record count should be the sum of this movie's and the previous WIP.
Suggested Thumbnails
<- feos’s favorite
<- eien86's favorite
arkiandruski: Claiming for judging.
arkiandruski: Looks good, The lack of weapons and powerups, and the need to avoid enemies present a lot of challenges that aren't present in the currently published run. Accepting as a new branch.
despoa: Processing...
feos: Fixed rerecord count.