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


Table of contents [expand all] [collapse all]

First experiences of TASes

I first knew about the existence of TASes through archive.org. The first runs I watched were SMB1+3 and Track and Field. My experiences of watching SMB1 were not good as it showed illogical tricks like passing though walls and backwards jumping as well as alternate pipe glitches all of which made no sense, although I knew about the wall jump as I'd pulled it off accidentally on many occasions myself. Track and Field had little to offer other than watching the triple jump, but SMB3 did because it had obvious inhuman precision and reactions and took unnecessary risks in the autoscroller levels. I also knew these autoscroller levels had a tendency to lag a fair bit which made me appreciate why all these extra lives were obtained. Still I never checked the site during that period.

Noticing the site

The first run that made me finally look into the site was the first SM64 TAS, which I saw on Google videos, after seeing it I thought it was only improvable by about 6 seconds (how wrong I was), the video did have a link to the site which contained FAQS so I was able to get effective understanding in relation to what a TAS was. I lurked on the forum but never registered.


Current Projects

No runs are being currently worked on. I largely consider myself semi-retired from TASing these days.

Abandoned runs

A lot, but the most notable one I did was for The Legend of Zelda: Ocarina of Time: Master Quest (long title, but I felt like doing that). Basically the game is a GameCube port of the original N64 game and has graphical updates and removes a few bugs while having overall less lag. The main difference however is that the dungeon concepts are changed meaning that they can be done faster and has an effect on the overworld route. I abandoned after discovering how to get past the Door of Time meaning all dungeons could be skipped, making the runs are effectively the same.


Runs I've done

I prefer to do games that I'm familiar with and would enjoy TASing, mainly because I'd rather do a run that is fun to do but maybe not regarded as a good game to watch by others; than a popular game that I don't enjoy doing, but would be shoe in for publication. That way I can keep my motivation going and not be disappointed if people don't like it.

Sonic The Hedgehog –

The original and best Sonic game in my humble opinion, the most interesting element is trying to maintain and control Sonic's speed. I decided to do an all emerald run because no one had ever done a full run of any Sonic game at that point. I was very new to TASing so my methods were slow, inefficient and struggled to find any improvements and ended up largely following JXQ's run. The run in the end was rejected through being to similar and boring compared to the current any%. In a way I think its better I did this run because it would have looked bad if I improved the current run by only a couple of seconds and then get beaten by 10+ seconds.

World of Illusion -

The reason why I choose this game was because I wanted to try and make a 2 player game, I watched the cancelled 1 player run and it was largely unimpressive I found a couple of glitches which basically kept my motivation going throughout, although I didn't put a lot of effort into optimization since I knew there was a high probability that the run would get rejected. Still, I learned a lot about the art of recording 2 player movies. Its not to difficult if you break it down into smaller parts and then slowly combine them altogether, It was unique and fun to do. It was rejected because many of the levels were very boring and simple by nature.

Mario 64 -

I'm not a particular big fan of SM64, SMB3 is the only Mario game that I've really liked. But I really enjoyed TASing the game and still do Today. After Nitsuja released the TAS input Plugin I decided to try it out on a few games, SM64 seemed to be the best game as it has very stable emulation and the graphics update each time frame advance it pressed. I asked FODA if I could do a 16 star since I felt he was going to do it himself. So jumped onto it and got started right away. It became obvious how unoptimized Spezzafer's TAS really was when I started working on the run. To anyone who wants to do N64 TASing,I strongly recommend you use the 0.6 Plugin, because otherwise your runs will be greatly unoptimized compared to what potentially be achieved. They were a few new tricks and strats, most which were found by others including FODA, but I did find some myself. Overall even without the new tricks in comparable sections I was 18 seconds faster overall, through much better precision. Overall my run was 47 seconds faster.

Mario 64 v2 –

After being quickly obsoleted, I decided to make another one because I knew specific parts that could be improved. This was never accepted because there was a new trick which saved a small amount of time and wouldn't hex in properly. Just when I was about to update, a new run came up. Generally, I'm not in favour of attempted obseletions as soon as run is submitted. Normally I'd choose to best another run as soon as it becomes published. Since during that time, it allows me to look actively look for new stuff.

Duke Nukem 64 –

I was wanting to do an FPS for a long time, my initial choice was Quake, but the game has significant problems such as large loading times, heavy unnatural lag and a few emulation errors. Duke seemed to emulate correctly and its my favourite old school shooter. Since my knowledge was thin on the speedrunning element, I did all the research I could. I wanted to do this on the easiest setting because it only contains more enemies and your really just dancing through them, but I knew the guidelines were that the hardest level needed to be picked. The run was largely enjoyable to make and has a very abusable engine. Unfortunately the programmers were lazy when it came to porting the game. I owned the Saturn version and it was a brilliant port, but the N64 version is noticeably a little choppy. This is fine when playing the game normally since you're carefully ploughing through the levels, exploring them and finding secrets. But its noticeable that frames are being skipped when you're hurtling through the levels. Ironically the game doesn't actually lag that much its just very inconsistent updates between 1-10 frames. The run was rejected because it was deemed unwatchable to some and there were some known improvements.

Mario 64 v3 -

I wasn't really interested in doing this because I didn't think I'd be able to show anything new or unexpected. Mukki posted a potential trick that I knew would be useful and with all the number crunching I figured that the trick would be able to save a couple of seconds. The island star in BoB was the slowest, but I knew that the chain chomp star was now seriously under question now. I was working through the game at leisurely pace not making much in the way of noticeable improvements, until I randomly stumbled across Swordless_Link who was asking me questions about hex editing to an unrelated game. He said at some point he was looking to improve the current run and had made some progress. I was fairly interested by this. He showed me what he had done, and found that there were parts in which he was faster than my current progress. Although they were still a few bits where I was faster. I thought about this overnight and decided that we should both work on it together. It took quite a while to do since many of the current areas strategies were largely talked over. Both of us ended up finding new time saving strats. For example I found a wall kick in the second Boswer level which was surprisingly overlooked in the past and he found how to grab Bowser's tail more in the final battle. After many weeks we ended up with a run which was 16 seconds faster. This is surprising given the fact that the current run only improved the previous one by 2 seconds.

Duke Nukem v2 -

I decided to do this because certain people were requesting that I make another version on the easiest setting with the most up to date tricks. Since I had a proper grasp of the engine I was able to find a few things myself that even Duke experts didn't know. The hardest part of the run was the planning and how to make effective use of the steroids.

Chuck Rock -

I played this game quite a lot when I was younger. Its very easy to pick up, play and complete in one gaming session. This of course made me curious to watch the current run. When I watched the run I thought it was unoptimized in a few areas. Plus the run hadn't used many rerecords. It was one of those classic Gens runs that urgently needed improving. The game was relatively easy, but there were 4-5 parts that were very difficult to do, but I thought I handled them well. The game however only rewards players with a few frames if any parts are improved. I even played large parts of the game in real time using save states. Most of the improvements did come through stuff that specifically requires frame advance like upwards damage boosting, which sadly I'm unable to determine to this day if its a programming error. I managed to improve the run by 4.5 seconds, this in my eyes is respectable since the current run had improved the previous one by 9 seconds overall. The one thing that annoyed me was that I didn't break 9 minutes. I would be delighted if someone could save 1.5 seconds off my run, but this isn't necessarily straightforward given how prehistoric the game and the nature of the physics.

Karate Kid -

I got inspired to do this after watching one of AVGN videos. He criticised the game for being short, overly frustrating, unrewarding, awful controls and some near impossible situations. Since I liked the movies, I downloaded the game. I couldn't stand it, but I had to keep on playing, because I had to beat it.. Although that only took me 20 mins to do so. After this I decided to watch the run. I questioned it and had a go at the first level and managed to improve it by 10 frames. Since it was only a short game I decided to play right through to the end. I ended up improving the run by just under 3 seconds. Decent given that the current run had improved the previous one by 1 and half seconds. However, while it may have seemed that I was Daniel San, Mr Miyagi was really Nitsuja in disguise. He showed off a couple of new fighting strategies which looked cooler and saved a lot of time, he was also able to gain a good number of frames through general optimization.

Pocket Monsters Green -

This is the Japanese version of Pokémon which has a very famous glitch called the “dokokashira” glitch, which allows the game to be beaten within four minutes. I wanted to try this out for a few reasons. One being that it was an rpg, two that it was a Game Boy game and three mainly to see if I would enjoy TASing any rpg based games in the future. It wasn't too difficult to make, and it only took a couple of hours to do. Unfortunately I've found that luck manipulation is the most frustrating element in any game, but the key to it in any game is to make it look as hidden as possible. That way people won't think that the run could be improved easily. While I did do a little research, it was quickly pointed out by knowledgeable experts that the run was lacking strategic planning. I quickly conceded defeat and cancelled it.

Mario 64 v4 -

It was brought up on gameFAQs that someone had found a potential blj location that might allow for a 1 or 0 star run. I looked into the thread and read carefully through it. I came to the conclusion that was being stated wasn't credible, but it was something original that hadn't been thought and tried before. I then started playing the game again on my N64 and messed around in the basement looking for potential locations, but completely dismissing anything that had been thought of in the past. There was a crucial post made some time ago that I always kept a note of but never took it to seriously until up to that point. I noticed there was a small ledge in the water just outside the HMC room. I was able after a few attempts to get 3-4 bljs, it soon became clear afterwards that there could be potentially a few other possibld locations. I then told the community that I'd found something. Although there was vocal demands for me to explain it, I didn't want to reveal it right away since I wanted to test it out fully myself. Unfortunately the initial location I found was useless. I then tried the step that was right next to the key door. Unfortunately I couldn't quite manage it but I got so close to getting it to work. I then uploaded an avi show what I'd found. A week later after testing it again, I realised that there was a slight oversight in what I had done, I was able to effectively get a unlimited amount of bljs. At the time I was talking to Swordless Link, since he'd just come back from his holiday. I uploaded an avi so he could see how the trick was done. Although the community was deeply suspicious as to whether or not the trick had been pulled off, we decided to keep this one quiet. Mainly because we we're going to implement it in our own run anyway, plus it was a very nice surprise. Surprisingly, I was a little disappointed that I'd found this and that it killed off a 16 star run, because there were still a few things that could be implemented and perfected. With hindsight I think a 1-star has done more good than harm. Since many people were getting tired of watching 16 star runs which were largely using the same strategies each time. Plus it gives great justification to allow an elusive 3rd category run, this would please both parties who are against both the 1 and 120 star runs for the same extreme reasons.

Quackshot -

I worked on this over many months, although not with any great intensity since there was a few projects which I did inbetween. After watching the previous run I thought it was very improvable and I even found a few time savers before starting, there were a few other potential ones as well, but I either couldn't make them useful or the trick itself backfired. After completing it I spotted a few improvements so I decided to go back and hex them all in. Some of the areas to my surprise were a little random so I had to manually TAS them back in again, usually finding a way to do it better. I feel the movie I did is pretty tight, but could probably still shave a few seconds via memory watch and frame shaving.

Mario 64 v5 -

After being obsoleted I decided to mess around and look for some more stuff, although I knew I would be very limited in what I would be able to find. However, I managed to find a couple things which would save the equivalent of about 8 seconds overall. I also had a good theory on bypassing the DDD barrier, one of the main barriers to a theoretically best route. There was also a few other minor strategies as well, but they would only be expected to save about 2 seconds over the course of the game. The new stuff was considerably more difficult to implement that I thought initially. Especially when we were truly pushing for frame perfection this time round. Swordless_Link was the one who finally cracked the DDD barrier. Funnily enough I wasn't there at the time when it was found, but I was in for a very shocking surprise when I got back. The discovery saved 45 seconds and most importantly allowed the game to be beaten in 0 stars. The only theoretical improvement left is to open the door in the moat outside the castle, but I really can't see that realistically happening, so I claim that the run has a perfect route through the game. In the end it was worth it, and is probably the most optimal run I've produced to date for any game.

Streets of Rage -

I've been wanting to do this ever since I joined the site, but it wasn't clear what the best strategies were, so it always made me hesitant to TAS the game. I tried doing a mock test on it a while ago, but realised it was unoptimized and that there was many likely ways to improve it. There was a whole variety of strategies which could improve the current run. Chemist then registered on the site and showed a glitch which could cut the last level short, which edged me into doing the run. I wanted to get other projects out of the way first before starting on it. Mukki was also very interested in TASing the so we decided to work together on it. I wasn't hell bent on squeezing every possible frame like the previous run I just finished. So there is various trade off's throughout the movie, but we keep it largely focused on speed just to make sure its way beyond human limits.

Twinkle Tale -

My OoT run was on hold, because the emulator was getting updated, so I decided to look for a game that wouldn't take to long to do and no one else would think of doing. After dismissing most games I quickly came across an interesting game, Twinkle Tale. It was only ever released in Japan. It's a top down shmup which involves walking through levels crammed with enemies and killing some very challenging bosses. Much of the game is made a mockery of because if you're very close to enemies you can dispose them very quickly with turbo fire Mega Man style.

Super Mario Bros 3 -

This is my favourite Mario game of the series. It's easily the hardest game to find improvements in. The project from beginning to end took nearly a year, but it was only intensely worked on in the last five months. I teamed up with Lord_Tom whose very good at memory searching and being able to figure out the reasoning behind glitches and certain restrictions. If any improvement is possible, then I can't imagine it exceeding 10 frames. This is likely to be my last TAS on an older console.

Combined RSS Feed
Mitjitsu last edited by Mitjitsu on 2018-05-20 18:19:41
Page info and history | Latest diff | List referrers | View Source
This page appears to be the personal page of Mitjitsu.
The following information is available of this person:

Note: Opinions expressed on homepages of our users are the personal opinions of those people, and don't necessarily represent the opinion of the whole TASVideos community.