Just not the skill you’d expect.
Many fans of video gaming are familiar with the concept of speedruns, be it through the semiannual streams of Games Done Quick, the old competitions among DOOM players, or just sitting down with a friend to see who can beat Super Mario Land the fastest, speedruns have long been considered a display of skill, a demonstration of both extensive knowledge of the game in question and the reflexes and speed to put that knowledge into action. And then there’s the tool-assisted speedrun, or TAS. Similar to a regular speedrun, the aim is (usually) to play through the game as quickly as possible, but with tools that assist (surprising, I know). Such tools may slow the game down, reveal hidden data, or allow the TAS creator to back up and try again when a mistake is made. Because the tools make the speedrunning “easier,” many don’t consider them legitimate. And it’s true that they shouldn’t be compared to conventional speedruns—tools create an unfair advantage that often leaves regular speedruns far behind. That said, labeling them as easy, with no skill involved, is a reduction as oversimplified as “platformers just move the player character to the right.”
While the in-game recording function of Doom boasts the first examples of using tools to complete a speedrun without mistakes, it wasn’t until 2003 that the internet at large picked up on the idea, starting with an 11-minute run of Super Mario Bros by a Japanese gamer named Morimoto. (Since then, the records for both the regular speedrun and the tool-assisted speedrun are under five minutes, separated by 0.1 seconds.) Despite the short length of the video itself, the actual creation of it, using resets from saved states and slowing down the game massively, took two years of work. Even with the advancement of tools, which are able to do things like peek at the variables and values the game uses in real time, or slow the game down to a frame-by-frame play, the time to create a TAS is often exponentially longer than the TAS itself. Many TASers track this work by the number of re-records made in the process, which can sometimes stretch into the millions. Why put two years of effort into an 11-minute video? Morimoto’s answer was, “I want to see what an ideal Mario run would look like.”
Unlike a regular speedrun, the point of a TAS is not to display skill in running the game. Rather, they are used for entertainment, often to answer Morimoto’s question: “What would a perfect speedrun look like?” Depending on the game and the tricks discovered, that answer can vary greatly. Some games use massive sequence breaks: Super Mario 64, a game that normally requires 70 stars to reach the final boss, can be beaten without collecting a single one through precise ramming of game boundaries. Some can break games by abusing the rules of in-game physics: for example, TASers get Sonic going so fast in Sonic Advance that the “camera cannot keep up with him.” Alternately, Super Mario Kart breaks the game by abusing the in-game definition of a lap. Some, such as the runs of Brain Age and Family Feud, take advantage of data input methods to have a little fun. And some games, like Pokemon Yellow and Super Mario World, can be broken so thoroughly TASers can insert their own code in. Yes, this is done solely with the input of a game controller!
So how does one go about making these videos? Time and patience are a must—sometimes, a TASer discovers a trick to improve the beginning of a run when they’re close to having it completed, prompting a rerun of the whole game. Having an eye for entertainment and knowledge of the game, of course, helps. Programming skills are also beneficial: many TASers create scripts or macros to assist in creating or even improving the emulators they work on.
And the programming aspect is one area in which TASers excel. More often than not, a TAS has a TASer behind it that can give you the important memory addresses, such as showing speed and position of the player character. Some discover how enemies are generated, what causes the game to lag, or how luck-based events are determined. The scripts they then put together can display those memory addresses, test which input combinations reduce lag, or generate pseudorandom or comical input. Outside of perhaps the Pokemon metagame community, there are few games that have been torn apart and understood so thoroughly as ones targeted by TASers. And while many such mechanics and tricks are found by disassembling game code, some are found merely by going “what if I press this weird combo of buttons” or “I’m going to try literally every sequence of input to see which I like best” or “I’m just going to turn the game off now, what’s the worst that could happen?”
For me, reading the creator’s notes is often as interesting as watching the end result. Digression: who here remembers the “Mew glitch” of the first generation of Pokemon games? I remember thinking all the steps seemed so very random. But by poking at the coding and trying different things, the community discovered not only that there was logic behind it—that it wasn’t some random happenstance—but what that logic was and how it could be used. Nowadays, that same glitch used to generate a wild Mew can also be used to generate a level 100 Nidoqueen, or bypass Snorlax without using the PokeFlute. Like solving a complicated math problem or successfully tinkering with a recipe, there’s a beauty in seeing the logic fall into place.
Over time, the “speedrun” of tool-assisted speedrun has become somewhat of a disservice. While there are still many TASes that aim for a theoretical fastest time, TASing as a whole is more about putting the game mechanics on display, and then hoisting those mechanics by their own petard. It’s less about a competition and more about creating art—art where the canvas is a game with its own idiosyncrasies and quirks, the only paintbrush allowed is a controller, and the finished picture is a video of the unexpected. Sometimes the art is even literal, with a controller used to draw pictures or create music. TASes are case studies of programming exploits done for the benefits of the audience who just really want to see Link hauling around a Bow-Wow, Marin, a ghost, and a rooster at the same time. They’re memories of old, sometimes obscure games brought back to life with new ways of looking at them, and shared with a fresh audience. They’re for all the weirdos who look at a game and see a playground, and the weirdos who want to see what sandcastles they’re making.
Comparing TASing and speedrunning is like comparing apples and oranges, or Mario and Zelda. But the one thing they have in common is a lot of hard work behind them. Let’s appreciate both for the separate things they are.