Now playing:

Music in the Age of Speed Gaming

2020 June 2119 min read

As a small child I went through a wide range of hobbies and obsessions, but gaming was decidedly not one of them. I have memories of occasionally playing Xbox and PlayStation games at friends' houses, and I definitely sank some time into Webkinz and Tamagotchis back when those were all the rage, but mostly I just had a boatload of other things besides video games that kept me busy in my spare time. Until, that is, I discovered Marble Blast - the game that changed my life.

Released in 2002, Marble Blast is a 3D marble physics platformer often (insultingly) compared to Super Monkey Ball, and its breathtakingly simple and fluid mechanics give rise to an unthinkably vast ocean of puzzles and challenges for which new, beautiful solutions are still constantly being discovered, by a small, tight-knit community of dedicated fans.

It was through immersing myself in this game and its endless intricacies that I was fully introduced to the world of gaming, and speed gaming in particular. To say Marble Blast is a game designed for speedrunning would be an understatement; a thought I often have nowadays when watching speedlore channels like Karl Jobst or Summoning Salt is how dull the more popular speedgames seem to me by comparison; how obscure, finicky, and luck-based all the technique and routing can be for those games. I imagine these people discovering Marble Blast speedruns and being like Plato's cavemen, stepping out of the shadows and seeing the full three-dimensional world of speedrunning for the first time.

I've been playing Marble Blast off-and-on for over a decade now, and I've seen the game continue to develop, spawning thousands of custom levels, fan-made updates and modifications and, eventually, online multiplayer in a variety of gamemodes. At my peak I was the multiplayer champion and the leader in individual-level world records on the Platinum version (right before this guy came in and started wrecking everyone's shit). I also still have the world record for the full-game single-segment speedrun, although this record is much less contested than those for individual levels.

All this is to say: reaching a high level in any speedgame requires dedication, and a kind of "speedrunner's mindset" that I learned to develop. Getting a top time can take countless hours, days, and even weeks, to achieve, and I've become intimately familiar with the often grueling process of refining every last ounce of movement, every jump, every bounce, and every turn, and practicing every advanced technique and every tricky timesaver, in the hopes of pulling them all off perfectly in succession for that one magical run.

Raw skill does, of course, play a huge part, especially in Marble Blast; even at the highest levels, you still see players noticeably improving and crossing what were once their plateaus. Route-finding is also a cornerstone of the game, and most of the top players have become renowned not only for their runs in themselves, but for the sheer ingenuity and creativity of their contributions to the ever-expanding metagame.

Still, when all is said and done, the world record on a level often just comes down to who wants it more; who's willing to put in the hours. All speedgames eventually reach this stage, where runs are so optimized that any slight improvement takes an intense grind. There's that old saying: the definition of insanity is doing the same thing over and over again expecting different results... and to an outsider, these grinds often do look insane. But this experience - endless practice and mind-numbingly repetitive trial-and-error in service of a larger payoff - is not unique to video games; it applies to many things in life. Including, in my case, playing piano.

Recently I published the third in an ongoing series of what I've been calling "Instagram Etudes": bite-sized showpieces written for the social media age. Each etude is crafted to pack as much musical meaning as possible into a one-minute Instagram video, to leave a lasting impression on anyone who happens to scroll past it on their feed. These etudes are an opportunity for me to show off my skills in an easily digestible and shareable format, and I'm using this opportunity to push myself, to test the limits of my own pianistic abilities.

It's never been easier for me to record music, now that I have my own room, tons of free time, and (finally) my own 88-key digital piano. When I recorded the tracks for my 2018 album Fresh Echoes, I had to do it by heading to my local music store with a USB stick, recording a sloppy rendition of each track in one take on an unfamiliar keyboard, and then painstakingly tweaking the MIDIs back home, ironing out the mistakes and fiddling with note velocities and sustain pedal automations to make the songs sound decent with my custom-made GarageBand piano soundfont. Recording the Instagram Etudes is a very different experience: they're much shorter than any of the tracks in Fresh Echoes, but they're also quite a bit trickier, and recording them on video means that I can't be sloppy. If I miss a note, or play it at the wrong time, I can't go back and correct it; I have to either keep it in, or start over and try again. Although each piece is only a minute long, getting the perfect playthrough might take multiple hours - which is time I can only really dedicate to a recording now that I have a piano in my room.

Note-for-note, the Instagram Etudes are by far the hardest pieces of music I've ever written, and I'd put Etude #3 in particular up there with some of the hardest piano pieces ever, period. But it's not really a fair comparison, because most classical music is written to be performed live. These etudes are not. They're written to be recorded in long, repetitive grinds, after potentially hundreds of attempts.

During these grinds I've seen that "speedrunner's mindset" manifesting in myself again. Recording the Instagram Etudes has shown me just how many parallels can be drawn between the recording experience and the years of Marble Blast that were preparing me for it. The most basic comparison, of course: that both things involve practice and repetitive trial-error in service of a larger payoff. But as I kept thinking about it, I realized that there's a lot more to this analogy than first meets the eye...

Loosely speaking, doing a speedrun involves three things: the planning, the practice, and the execution. In the planning stage, a speedrunner will map out the route they take, what techniques and tricks they'll use at which sections, and usually, for top players, ironing out every single jump pattern, camera movement, etc. to optimize the run down to the granular level. Often this takes some lateral thinking and deep knowledge of the game, and this stage is where a player's expertise and ingenuity can really shine.

Most piano performance is done with the same level of extensive planning beforehand, and for the Instagram Etudes, it is a necessity. More than ever, I find myself obsessing over not only things like which notes to play with which fingers, or which hands, but also how long I might hold a note before jumping to the next one, and - sometimes - even where my eyes will be, so that I can be looking at the important part of the keyboard when a tough jump comes. The same feeling that a speedrunner gets when they find a clever new tweak to a tiny section of a run, I get whenever I figure out a new fingering or other subtle technique to make a passage easier to play.

Another part of the planning stage in speedrunning, one which many people kind of take for granted, is estimating the difficulty of a run, usually by zooming in on the hardest sections. Often, what you'll do is estimate the probability of pulling off each trick, and then multiply those probabilities over the whole run to get a sense of how long the series of tricks will take to pull off in succession. If you're going for a world record, what you're looking for is not necessarily the fastest theoretically possible time, which might require so many 1-in-100 tricks that it would take years to get, but the specific combination of such tricks that maximizes probability of success given a certain "goal time".

I find myself going through a very similar process when planning out the Instagram Etudes. The third etude, for example, I thought of as in three sections, each of which I had a certain chance of playing cleanly. During practice, I settled on giving myself about 1-in-3 odds for the first section, 1-in-5 for the second, and 1-in-20 for the third... meaning that on average, it would take me 3 · 5 · 20 = 300 attempts to do all three sections in a row.

The nice thing about being both the composer and the performer is that if I ever thought this number was too high, I always have the option to just make the piece easier. This is a privilege not directly afforded to speedrunners, although many games and levels are created with speedrunning in mind, design choices being made to ensure that the experience is as enjoyable as possible. In Marble Blast, an infamous technique known as a "traplaunch" involves squeezing your marble partially between two solids moving relative to each other, eventually launching outwards at ridiculous speeds. It's the most finicky and unpredictable trick in the game, often requiring hundreds of attempts to pull off correctly, but it has the power to completely "break" certain levels, especially those from the original game, because the developers didn't know about the technique. Nowadays, levels are explicitly designed with failsafes in them so that traplaunches are either impossible or a waste of time, and some have argued that the physics should be updated to remove them entirely. This phenomenon, of "changing the rules" of a speedrun to make the experience more fun, is not unique to Marble Blast; many games have certain techniques or glitches that are outright banned from speedrunning, and others that are permitted except in special "glitchless" categories. As a composer, I can exert a similar kind of control over what techniques I "allow" and "disallow" in my pieces.

After the planning stage, of course, comes the practice. Often these two stages come in tandem: sometimes ideas for new strategies pop up during practice, and sometimes practice is required to see how viable a strategy is - this is true of both speedrunning and piano. But the practice stage is also your opportunity to do something that you can't do during an actual attempt: skip around and isolate certain parts of the run, doing each one over and over again until it becomes second nature.

Doing this on a piano is simple, and is such an integral part of practice that it seems silly to even mention it. But doing it in video games is often trickier; if you could just skip to any part of a game, there wouldn't really be much of a game. Speedrunners sometimes get around this problem by using an emulator, where pressing a hotkey at a certain moment in the game gives you a "savestate" that you can start from over and over again. Obviously, this is only allowed in practice, but in extreme cases it's such a vital practice tool that people make special versions of the game with specific savestates pre-programmed in. In Marble Blast, speedrunners sometimes use the level editor to create a spawn point at a specific part of the level... but honestly, more often than not, the levels are either short enough or have enough actual checkpoints in them that you can get by without any cheats at all, and most players just do this.

An amusing consequence of this is that speedrunners are often better at the earlier parts of a run than they are at the later parts - and this difference can be even more striking in the execution stage, where failing at an earlier part means you restart and don't get to play the later parts. Over time, if you're not careful, all that practice on the later parts will fade, and by the time you make it there you'll be much less prepared, so speedrunners will often take time to finish "dead runs" just to get the extra practice in. Likewise, when I'm doing my long recording sessions for my Instagram Etudes, even though each one is only a minute long, I have to make sure that I don't focus too much on the first 30 seconds of the song, lest I be caught off guard by the last 30 seconds.

This is partially why, when speedrunners are given the option, they choose to do the hardest parts of the run closest to the beginning. Not only do you become more practiced at those parts, but failing earlier rather than later in a run means less time wasted overall. Unfortunately, due to the structure of video games, this often isn't possible. Game designers, sadistically enough, like to start you off easy and increase the difficulty towards the end, rather than the other way around. As it is with video games, so it is with music as well; you'll note above that the easiest part of Instagram Etude #3 is at the beginning, and the hardest part is at the end, where the climax is.

In speedrunning, it is the execution stage that is usually both the most tedious and the most intense. You might be stuck for several minutes making the same exact mistake, seemingly getting nowhere or even feeling like you're getting worse. And on the flip side, when you finally get a run on a good pace, there's an immense pressure not to mess up, because you don't know when you'll get another chance. If you're not used to it, this pressure itself can be what clouds your mind and prevents you from pulling through.

As a long-time speedrunner, I've learned to numb myself to the "emotional whiplash" you get when long stretches of utter boredom are peppered with brief moments of extreme tension, usually followed by bitter disappointment and frustration. For me, it really helps to think back to the planning stage, and the probability calculations I talked about there. Focusing on the numbers is what grounds my emotions: if I need to pull off a trick mid-run that I've deemed to be, say, 1-in-20, I just relax and imagine I'm rolling a D20. I will have already figured out how often I'll get the chance to try the trick, and how often I'll be able to pull it off - and thinking in terms of those broader numbers makes each individual attempt seem almost meaningless.

It's this same mindset that got me through those long, arduous recording sessions for my Instagram Etudes. While recording my third etude, for example, I had not only an estimate of how many attempts it would take, but also a breakdown of what those attempts would look like. I knew that about 100 of those 300 attempts would make it past the first section, since I gave myself 1 in 3 odds of doing that section properly. Of those 100, about 20 would make it past the second section, and one of those 20 would be the magical playthrough that made it to the end. Rather than placing emotional stake in each individual attempt, I soothed myself with the knowledge that (1) each individual attempt is most likely going to fail, and (2) it's only a matter of time before the successful attempt does come. Every time I finally made it to the third section and then failed, I thought to myself: "I knew I was gonna have to do that about 20 times, and now I'm one step closer".

It turned out, at least in this case, that my estimate was eerily accurate. If you do the math, taking each section to be about 20 seconds, then having 200 attempts fail on section 1, 80 on section 2, and 20 on section 3 would take 200 · 20 + 80 · 40 + 20 · 60 = 8400 seconds, or 2 hours 20 minutes. And indeed, it took me just over two hours to get a satisfactory recording of Instagram Etude #3.

The recording wasn't perfect; I do miss a couple left-hand notes, one of which is quite audible near the end, and I screwed up the rhythm as well at one point, but I deemed these to be relatively harmless mistakes that don't sound too out of place. Speedrunners will be familiar with this experience; certain mistakes in a speedrun are not only more costly than others in terms of time lost, but also more obvious and visually jarring, and that in particular will affect how satisfied you are with the run, even if maybe it shouldn't. (Certainly, I've had the experience many times of being pissed off after getting a world record because the previous one "looked" better.)

Of course, sometimes things happen that make me wonder if any of this is paying off, if the efforts I'm putting in to learn and record these pieces are even perceptible at all. I got a rather puzzling comment recently on my second Instagram Etude - to that point, the hardest piece of piano music I'd ever written - complimenting me on my "improv skills". It reminds me of when they would put on hip hop cyphers in my high school cafeteria, and I would go in and spit an obviously pre-written verse, only to have people inevitably come up afterwards and say how amazed they were that I could "come up with all that stuff off the top". But that's another story.

Let's expand this analogy beyond the Instagram Etudes. If recording a one-minute etude is like grinding for an individual-level world record, in Marble Blast for example, then live performance is very much like doing the single-segment full-game run. I mentioned that I still have the world record there, which sits at 37 minutes 40 seconds, and needless to say, doing a run that long is quite different from doing a single level, and requires a different mentality. Extensive practice is all the more essential, since you might go hours without getting a run to the end at all. The execution process is a lot more emotionally draining; both the periods of boredom and the spikes of excitement last a lot longer, and are that much tougher to endure for it.

But most importantly, in the full-game run, you need to be a hell of a lot more conservative with the types of tricks you try to pull off. Every time you introduce another coin-flip into the run, you're doubling the amount of time it'll take to get it, which for a 30-minute run is a huge deal. You'll notice in the video for my world record that I have a helpful window on my left sidebar that says this:

Chances of getting:
Platform Party - 41%
Upward Spiral - 76%
Twisting - 48%

Those are the levels with the three hardest tricks in the run, chosen very specifically because they save enough time to be worth the risk. I got these numbers by playing each level 100 times and counting how many attempts were successful. With this knowledge, I knew that my combined chances of getting all three tricks were about 15%... and while this might not seem like a low number, keep in mind that this is only three of the 100 levels in the full-game run. Granted, most of the other levels are much easier - but doing full-game runs is uniquely frustrating because even if the chance of messing up on any particular level is just 1%, when you're doing 100 levels, those tiny percentages add up, and you're actually more likely than not to mess up at some point during the run (which I did in a couple places).

I'm all too painfully aware of how true this is for piano as well. Even more so, in fact: in a live performance, you can't have sections of a song that you're only 41% competent at. Usually you're not allowed to settle for anything less than 100%, and this is the most frustrating part of piano practice. Sometimes it seems like I can practice every single section of a piece until I'm blue in the face, but at the end of it I'll never be able to get from 99% competent to 100% competent - or if I do, I'll never be able to tell unless I play the whole song at least 100 times. And if I'm only 99% competent in every little part of the piece, then all it takes is a long enough piece and I'm bound to make some screw-up somewhere that I have no way of predicting ahead of time.

When producing Fresh Echoes, I did what I suppose is the musical equivalent of a tool-assisted speedrun, where cheats such as savestates and slowdowns are used to produce a theoretically "perfect" run of a game, one that a human could hypothetically pull off if they had the superhuman ability to press the exact right buttons at the exact right times down to the millisecond. The ability to eliminate all human error, and incorporate all optimal strategies no matter how difficult, is what makes tool-assisted speedruns fun to watch - and likewise, my power to take out mistakes from, and make minute adjustments to, my piano recordings is what makes them fun to listen to. And honestly, I'm fine with that being my goal as far as music goes; I'm not a world-class pianist, I'm just trying to create a fun listening experience however I can. Without "tool assistance", producing a high quality recording of a full-length piano piece is just like pulling off the full-game Marble Blast run: a ton more work for me, and a lot less payoff for the audience.

As weird as it may sound, going from Fresh Echoes to the Instagram Etudes has been a breath of fresh air for me. They're tougher, and they're tedious to record, and every little bit needs to be just right for the video, and my housemate downstairs probably has perfect pitch now from how much she's heard that repeated D ostinato. But I'm used to that kind of grind by now, and I know how to get past it. Persistence is a skill, and I learned it from Marble Blast.

Contact Me

Follow me on: