The Stack

Well, I’ve managed to squeeze my way through the last level of Defense Grid‘s campaign mode. The game content never acknowledges my speculations from the previous post, treating victory as simply victory and the salvation of the planet. But I suppose that if every game had a subversive twist, it would stop being subversive. Something has to play it straight, and Defense Grid is a straighter game than most I play.

So let’s just comment on a few more points of mechanical interest. In fact, let’s start with interest itself, one of the game’s major experiments that I neglected to describe in my previous post. Any money you keep in reserve compounds interest at a rate determined by the number of power cores at your base. Obviously this provides an extra motivation to protect the cores, but it also gives you a reason to spend as little as you can get away with in the early part of the level, and these motivations are in tension. The net effect is a desire for maximal efficiency in situations where other tower defense games would have you just trying to overpower stuff by building as fast as possible.

My own experience is that the need to build stuff generally wins out. This is especially true in the final few levels, which are large enough that the space always seems underutilized. The temptation to build everywhere is strong, but should be resisted, if my narrowly-successful strategies are at all indicative of the correct approach. Building level-1 towers everywhere just takes away money you could be spending on upgrading existing towers, which seems to usually be the more efficient approach; that is, in most cases, upgrading a tower increases its damage potential more than an equal cost of additional towers would, especially if it’s wisely placed. Note that there’s still often good reason to refrain from upgrading, because upgrading isn’t instant. In fact, it’s agonizingly slow, and while it’s in progress, the tower undergoing the upgrade is inactive. So you always need enough active towers to pick up the slack when you get the cash to upgrade something, and this guarantees that you won’t be able to get the cash to upgrade it as soon as you want to.

Mind you, if you’re frequently spending all your money, and therefore not earning interest on it, you don’t really need the power cores to be at your base all the time. And, in fact, having them scattered along the path can at times be a boon in disguise. You see, there are occasional enemies that fly. They don’t follow the same path as the ground units, but follow their own swooping flight plan, unobstructed by your towers. If they manage to fly their entire path without getting shot down, they’ll grab a power core from your base and immediately take off into the heavens, without giving you a chance to get it back. But if all your power cores are elsewhere at that point, they just leave. There were points when I was sure I was doomed, because I had only one or two power cores left and couldn’t get my missile bases operational fast enough to wipe out oncoming fliers, only to be saved when a ground unit grabbed the power cores first.

Incedentally, there are only three levels of tower — that is, you can only upgrade towers twice. It turns out this is enough to be satisfying. There are really only three conceptual slots for an upgrade system anyway: you’ve got things that you haven’t upgraded at all, things you’ve upgraded some but can still upgrade further, and things that are at max level and can’t be upgraded further. Furthermore, three is few enough that the game can assign a vividly distinct color to each upgrade level — specifically, green, yellow, and red — thereby making the level of everything on the screen immediately apparent. Aliens use the same color scheme, and therefore presumably also come in only three strengths per type. I just wish that the types of tower had something like the same level of clarity! Most of them just look like towers in slightly different shapes. I sometimes lose track of which tower is which type and wind up hurriedly upgrading the wrong things.

The environments in Defense Grid: The Awakening are all brown and barren, rocky deserts with old and crumbling structures on them. There isn’t a lot of in-game information about the world and its history, but you know that the “aliens” have attacked in the past — long enough ago for the defense grid to need awakening. The sole speaking role belongs to an AI, the uploaded mind of a human who was involved in the defense the last time. He talks like a stereotypical British colonial officer and enthuses about how beautiful things used to be. At times he gets confused and addresses you by the name of his dead son, who he failed to protect during the first invasion. In other words, there’s every sign that the human presence is long wiped out, and this is a dead planet, with nothing worth defending.

This would be a familiar twist from a certain other tower defense game, but I don’t think that’s the only reason I think of it. I can believe I’m unduly influenced by the degree to which this game feels like a RTS game, though. All tower defense games are of course descendants of the RTS, but I’m talking here mostly about superficial matters like the mere presence of voice acting and the measured pace at which it introduces new elements. (For a while, it felt like the entire game, like the single-player campaign in many a RTS, was an extended tutorial.) In a typical Warcraft/Command & Conquer-influenced RTS, your base is a hive of activity, with autonomous worker units harvesting resources and repairing buildings, but here, it’s just a repository for power cores. So even the things you’re defending aren’t alive (even if they are the only things on your side capable of moving under their own power).

It all makes me speculate that this scenario is the most natural fit for a tower defense, this defense of the dead from the living (which would make Plants vs Zombies a clever inversion). It’s part of the genre’s definition that the enemy is active and your tools are passive, waiting for something to kill. Perhaps you could make a satisfying tower defense set in an Egyptian tomb, placing curses to foil looters and acheologists.

But then, I seem to be a bit obsessed with finding themes relating to death and mortality in games recently, so take it as you will.

So, I bought a couple new Steam indie bundles recently. (They’re calling them “bundles” now. They used to call them “packs”. I’m guessing there’s a perfectly humble reason for this change in terminology.) And one of them contained Defense Grid, another of Steam’s perennial discount items that I’ve somehow managed to avoid purchasing until now. I’d been curious about it, however, because it seemed to be the first tower defense game with A-list production values — by which, admittedly, I mainly just mean 3D models and voice acting. But that’s a somewhat less rare combination today than three years ago when it was released. Having played it most of the way through now, I have to say that it’s pretty by-the-book, its basic gameplay not much different from Desktop Tower Defense and its myriad online imitators. Things come along a path, you place towers to kill them, and in the process you earn money that you use to build more towers or upgrade existing ones. But it does do a few interesting things that I think are worth pointing out.

For starters, there’s the control scheme. You have a cursor in the center of the screen. Move your mouse and the cursor stays put while the rest of the world moves. This is, of course, basically how first-person shooters work, but you’re not rotating in place here, you’re moving in a plane just like the cursor would if it were moving. And anyway, the fact that this game is so mechanically similar to so many Flash-based games on the web, which generally don’t lock the cursor in place (because that would be really annoying on a web page), means that I’m aware, as I play, of the inversion from how these things usually work. That’s why I describe the mark in the center as a “cursor” rather than as a “reticle”. And the fact that you click on things to open up sub-menus (generally either “choose a type of tower to build here” or “upgrade/sell”) makes it seem even more cursor-like.

The 3D modeling isn’t just window dressing. The curving paths that the invading aliens come in on can cross over and under themselves, like in Zuma, with the result that you’re not just concerned with level geometry, but level topology as well. 2D games are relatively easy to think about, because we’re good at associating information with locations, thinking “This area is secure” or “If the enemy reaches that point it’s time to take desperate measures”. Paths that go underneath the main playfield confound this sort of thinking. Sometimes the paths form a confusing tangle that you need to simplify by blocking most of the pathways off. And that becomes an optimization puzzle: ¹And probably an NP-complete one at that. which pathways do you block off to give the advancing enemies the longest route to your base, the most exposure to your guns?

Probably the most interesting thing is the matter of what happens when an alien survives the gauntlet you’ve set up and reaches your base. The normal thing for a tower defense is to do is for the monster to knock off a fixed number of hit points or civilians (sometimes a higher number for tougher monsters), and then either be absorbed (as in Immortal Defense) or teleport back to the entrance for another run-through (as in Gemcraft). In Defense Grid, your hit points manifest as “power cores” that the aliens are trying to steal. This means that the aliens aren’t just trying to reach your base. They have to actually carry the things offscreen. Sometimes the exit is in the same place as the entrance, so that they have to pass by all the same towers twice. Sometimes it’s at the end of a completely different path. Sometimes the level topology is mutable enough that you get to decide how much they have to backtrack. From the way I’ve just described it, you might think that making them double back most of the way is optimal, but that’s not the case. When you kill an alien before it makes it out, any cores it carried drop on the ground and start inching their way back to the base. While they’re on the way home, any other alien with carrying capacity to spare can pick them up. So a core dropped on the path in is just going to shorten some alien’s path. Furthermore, this means that the genre-typical swarms of individually weak creatures are among the game’s most fearsome adversaries toward the end of a level, because they make the panicked last-ditch attempt to rescue your last core futile. As long as even just one member of the swarm survives, it can pick up the core from where it falls and keep carrying it away.

NightSky is the latest from Nicklas “Nifflas” Nygren, author of Knytt, and as with Knytt, most of the point of playing is to observe the stylish, elaborate-yet-minimalist visuals. This is a quiet game made out of silhouettes in various themes, levels with silhouettes of gears and levels with silhouettes of distant trees and so forth. A glowing glass orb navigates this environment under the player’s control.

I suppose the spherical avatar begs comparison to Within a Deep Forest, Nifflas’ previous game with a ball for a protagonist, but the gameplay reminded me a lot more of Gumboy: it’s a heavily physics-oriented game where your main activity consists of just rolling left and right and trying to get enough speed to go flying off ramps to where you want to go. It doesn’t push this as hard as Gumboy did, though. NightSky is more about variations on the theme. The default controls include a button to speed up and a button to stay put (the latter being useful for clinging to moving objects), but some levels don’t allow them, or replace them with other things: the stay-put button activates a machine, or the go-faster button reverses gravity. Some levels even disable the move-left/move-right buttons and make you navigate entirely by manipulating the environment — say, releasing a hammer to smack a sled with the orb in it, or using a pair of pinball flippers. There are orb-driven vehicles, including ornithopters. It’s just one shadow of a thing after another.

The irrelevant frame tale involves dreams, and that’s a good way to think of this game: as in a dream, you just find yourself in situations without preamble or explanation. These situations last three screens, no more and no less. If there’s a navigation puzzle that only uses two screens, you get a third one that’s pure scenery. I rather like this pointless symmetry, although I’m not sure I can explain why.

The game offers both an easy and a hard mode, the difference affecting the content of the levels but keeping them roughly the same, like the light-world/dark-world thing in Super Meat Boy. It suggests reserving hard mode for your second play-through, and I agree that this is the correct approach for this game, as the hard versions of the levels are sometimes quite frustrating, and there’s a good deal of satisfaction in just breezing through the easy versions. After the epilogue, you can unlock a final set of “slightly nonsense” levels that break the game’s carefully-crafted ambiance entirely, with brightly-colored checkerboard levels and levels made of ASCII art and suchlike. It’s a good way to end it. I can’t really say that the game takes itself seriously, considering the orb-driven ornithopters, but it’s arty enough that the silly bonus content helps to strike a kind of balance. Braid could have used an ending like this one.

A colleague of mine has on his desk a book titled 1001 Videogames You Must Play Before You Die. The main purpose of this book seems to be to provoke disagreement. (Why else would it, for example, list Silent Hill but not Silent Hill 2? I think that most fans of the series would agree that if you’re going to play only one of those before you die, it should be the latter.) In the course of flipping through its pages and arguing, someone noticed a listing for Intelligent Qube (aka Kurushi) and recommended it specifically to me. I had never heard of it before, but was intrigued enough to seek out a copy. Apparently, despite a loyal following, IQ didn’t sell well outside Japan. Consequently, it’s now one of those games described as “rare” and “hard to find”, although in the age of eBay all this really means is that used copies sell for more online than you’d expect of a Playstation game from 1997.

IQ is in the “puzzle game” genre, but not in the sense that I was expecting from the way it was described to me: it’s not a “Think!” game, but a “Think fast!” game. In other words, it’s in the broader school of Tetris, down to the inexorable descent of groups of blocks that you have to deal with before they reach the bottom, except that, this being designed for the Playstation, it’s all in 3D and instead of falling downward, the blocks are coming at you.

Also unlike the usual sort of falling-block game, you play it from the inside, kind of like the Royal Puzzle from Zork III: you control a little man who runs around trying not to get steamrolled by the blocks. Getting run over doesn’t end the game, but it does prevent you from taking any more actions until the current wave of blocks has completed its journey, and that’s often enough to make losing inevitable. As such, getting run over always feels like a cheap shot. The movement of the blocks isn’t constant in this game: they take discrete steps, rolling from face to face like in Edge, and they often pause for a little while to give you a chance to do something. But sometimes they don’t pause as long as you think they’re going to. Presumably there are rules governing this, but I have yet to figure them out.

And what does your little man run around doing? Setting traps! Pressing the X button sets a blue marker on the floor tile you’re currently standing on; pressing it again detonates it, destroying any block currently on that spot. This is a big source of confusion during panicked moments: losing track of whether you’ve pressed that button an odd or even number of times. There are three sorts of blocks. First, there are the normal ones that you want to destroy before they reach the end of the track. Then there are “advantage” blocks that leave green marks behind when destroyed, which you can detonate at the press of another button, either immediately or after letting the blocks advance more, destroying any blocks in a 3×3 area. (But not your avatar, fortunately.) Finally, there are the “forbidden” ones, which you aren’t supposed to destroy. The punishment for making mistakes is always the same: the playfield is shortened by one row, making the game harder, and ending the game if you’re standing on the row that got deleted. But you can miss several normal blocks before this happens (there’s a counter on the screen keeping track of how close you are to this penalty), whereas deleting a forbidden block always incurs the penalty immediately. Thus, avoiding deleting forbidden blocks is more urgent than getting all of the normal ones — although if you can do both, clearing a wave perfectly, the reward is that the playfield lengthens by one row, giving you a little extra breathing room. Thus, this is very much a positive-feedback game: the reward for doing well is that it becomes easier to continue doing well, the punishment for doing badly is that it becomes harder. This too is very Tetris-like, but the dynamic is different: where Tetris starts off feeling easy and turns desperate once you’ve crossed a certain threshold, I felt like the tipping point in each level of IQ was the point where I acquired enough skill to stop failing.

As the levels advance and the difficulty increases (mainly by increasing the number of blocks in each wave), the game becomes all about planning out when to use the advantage blocks. You want to use them when they won’t catch any forbidden blocks, and that takes some planning. Just about the worst thing you can do is have two green-marked tiles in inopportune positions relative to each other, because you can’t detonate them individually. I’ve generally tried to avoid this by destroying the advantage blocks one by one, but this means taking more time and letting the blocks get uncomfortably close to the end of the track, and possibly even trapping your little man behind a fence of forbidden blocks. For as the waves grow larger, the game takes on aspects of a maze game, with the forbidden blocks defining where you can go, unless you accept the penalty and blast one.

In short, there is a substantial amount of gameplay here, and I’m more satisfied with the game now than I was when I first realized that there was such a strong time element, although I’m not convinced that the experience is worth what I paid for it. I have by now completed the game by dint of copious use of the continue feature, which starts you over from the start of the current level. This took two sessions, although I probably would have continued playing in my first if I had realized that the game doesn’t save your progress at all. I was fooled by a “save” option in the “options” menu, which I think just saves the high score list. It’s the arcade sensibility, really. The whole thing is meant to be played in a single session, using multiple quarters.

So, why is this a game that You Must Play Before You Die? If I recall correctly, the writeup in that book was mostly impressed with the feel of the thing, that unlike most Tetris-influenced puzzle games, it felt like it “mattered”. And, having played it, I now think this mostly has to do with the music. The game has a stirring, epic soundtrack, like John Williams movie score.

As expected, I finished Edge last night, even to the point of picking up all the crumbs in the game. (They’re actually called “prisms” by the game, which I suppose is accurate: geometrically speaking, cubes are prisms. But I’ll continue to call them crumbs.)

If this game has one really memorable gimmick, it’s the way it exploits the ambiguity of isometric perspective. Without the depth cues provided by perspective, two objects that look like they’re next to each other could be separated by any distance in the direction perpendicular to the screen. If those two objects are floor tiles, it can look like a continuous navigable surface when it’s not. There’s one fairly advanced level that appears identical to the simple tutorial-like level 1, but is actually completely different. Fortunately, the game provides a mini-map in one corner that shows a schematic of the actual level geometry around you. Without this, some bits would be absolutely hopeless, particularly the occasional level containing secret crumbs in places that are completely occluded from view. But it’s not something you’re looking at all the time. I had to keep reminding myself to look at the mini-map occasionally, in case there was some revelation to be found there.

Also significant: unintuitive limitations on movement. Your avatar is a cube that moves by rolling without slipping from one face to another. The usual way to climb is simply to roll up onto an adjacent block; this is why you can only climb walls that are one cube-unit in height. The thing is, it actually enforces the physical consequences: if you’re in a trench, for example, with walls directly to your left and right, it’s impossible to climb out, and you have no option to roll along the trench lengthwise. I found it was easy to overestimate what was impossible, and thereby miss ways to get places. There’s generally an obvious approach, but you can miss out on crumbs or simpler approaches this way, possibly things that would shave seconds off your level-completion time if you care about that.

Now, I mentioned that there was a particularly difficult move, apparently called “edging”, where you start to roll up a vertical surface, but arrest your motion by supplying just enough force to keep from falling. (Usually this is done as a way to hitch a ride on a moving object, but it’s also a way to stay put on a precipice while you wait for a moving floor tile to position itself under you.) I’ve discovered the secret to executing this, by the way: switch the controls to virtual-gamepad mode, and all it takes is a particular rhythmic tapping on the virtual button. But the reason I bring it up is because of one particular level that illustrates everything I’ve said so far: level 18, “edge time”. (The game only uses lower-case letters.) This is a level based mostly around increasingly-difficult acts of edging — or so it seems at first. But if you explore, and look at the mini-map, and think about how the grooved areas can really be navigated, you can get through it without any edging at all. When you first encounter this level, it functions as a tutorial on a difficult technique that you’ll need later. But to a sharper inspection, it’s an exercise in doing things the non-obvious way. Now, there are definitely crumbs in the game that can only be reached by feats of edging, but this level makes me wonder if there are other places where there are alternative routes that I missed. I’ve mastered the game enough to eat every crumb, but not enough to get high speed ratings, and this sort of inquiry might be crucial to that sort of achievement.

Which, however, doesn’t make speed play more appealing to me. If this game has more secrets, they’ll probably remain unfound by me.

More iOS gaming on the bus today while I contemplate what, if anything, to do about my repeated inability to run PC games without crashing. Today, I try out Edge, a game that I probably wouldn’t have heard of without Tim Langdell‘s attempt to suppress it. Langdell is so loathed in the games industry that I’d like to say that I relished giving money to his competitors, but as Mobigame (the makers of Edge) is an actual game company, they can’t really be said to be in competition with Edge Games, just as First National Bank wasn’t a competitor of John Dillinger.

But about the game! Edge is essentially a simple isometric platformer in a retroesque style: it’s all monochrome cubes, except for the player avatar and the crumbs you’re supposed to eat along the way, both of which are cubes that cycle through pastel hues in a pulsing, Atari 2600 way. Challenge is created mainly by moving elements, either cycling or triggered: cubes that threaten to knock you off platforms, cubes that you have to ride on top of, and, trickiest of all, cubes that you have to cling to by an edge in a diagonal posture without letting your angle decay or resisting the decay too hard and pivoting to the top and slamming into a wall and falling down. The last is something I still find very difficult, regardless of what control scheme I use.

Now, about those control schemes. There are three, and they’re all awkward, but they’re awkward in different ways. By default, you have a touch-and-drag interface where the relative movement of your finger is turned into a directional force. This is awkward mainly because of the mismatch between analog, any-direction finger movements and four-direction, discrete-steps cube-rolling, but also partly due to the limited space available to drag your finger around. Even on an iPad screen, I find myself sometimes running off the edge. Alternately, you can switch on a four-button directional virtual gamepad, which at least links the discrete directions to discrete inputs, but has the problem that it’s easy to lose track of where it is while your eyes are fixed elsewhere. Finally, there’s an accelerometer-based tilt-to-move system, about which the less said the better. Edge has recently been ported to PC, and it seems like pretty much any PC-based control scheme would be easier to use than what we’ve got here on its native platform. (Sort of like Machinarium in the opposite direction.) But then, making it easy may not be the point.

Also possibly missing the point: playing it for an hour at a time, like I’ve been doing. It’s a phone game, hence it is made for quick bursts, not obsessive play. I’ll probably finish it tonight.

Last night, I was tired, and not in the mood for anything stressful or taxing. Going back to Super Meat Boy, or even to the lesser challenge of Heroes Chronicles, was out of the question. So I turned to my largish sub-stack of things bought in recent Steam sales that I haven’t even tried yet.

The first thing I tried was Safecracker: The Ultimate Puzzle Adventure. I suppose it says something about me that a puzzle game — yea, a self-proclaimed ultimate puzzle game — is my idea of something neither stressful nor taxing. But I had every reason to believe that this would be essentially just a gallery of soup cans, where the scope of every puzzle is well-defined and there’s no possibility of negative consequences. After all, that’s what the original Safecracker was. I could be wrong; I realize that it’s not the same game. But I didn’t at first. It was many months after S:TUPA was added to Steam that a discussion in a completely different context (roughly “This is just like that puzzle in Safecracker!” “What? I’ve played Safecracker and I don’t remember any puzzle like this.”) made me aware that it was a sequel. I think understand why the makers decided to obscure this: if it were called Safecracker II, there would be potential customers who would decide not to play it because they hadn’t played the original, or who decided to play the original first and found it so off-putting that they never bought the second. But the title they chose almost kept me from buying it, and I’m their target audience. There must be some better compromise.

At any rate, I couldn’t get S:TUPA going at all on my system. Starting it just locked my machine up with no video output. Possibly it was defaulting to a resolution that my monitor doesn’t support, but even then, you’d think I’d get some background music or something. I have seen this game running on a modern system, though, so it’s probably a solvable problem. But it wasn’t the sort of puzzle I was in the mood for, so I switched games.

Next up, I tried The Ball, a first-person puzzler, which is to say, a game that owes a great deal to Portal, even though the theme here is Aztec ruins (with hints of Ancient Astronaut) rather than sterile white corridors. The main conceit is obstacles that can only be overcome by using a large, unwieldy metal ball, a unique item doesn’t necessarily easily go where it’s needed. Your main control over it is a handheld device that’s something like a ball-specific version of the gravity gun from Half-Life 2: you can use it to attract the ball when it’s in range, and also to smack it like a pinball and send it careening forward. Maybe I was doing things suboptimally, but I found that I used the attract mode to move the ball around most of the time, which means that the ball spent a lot of time right in my face, which is always awkward in a first-person game. The designers understand the problem, and compensate for it by making the ball go transparent when it blocks your view significantly, leaving only some bands solid. I felt that even this cluttered the view uncomfortably.

When I started the game, I noticed that Steam listed some “Last played” data, which struck me as strange, because I had never actually played it before. But then I remembered that I had attempted to play it back when I first bought it, only to have it crash immediately. This time, I fared better: it lasted about a half an hour before crashing, long enough for me to get not quite all the way through the first level. Since this level is pretty tutorial-like, I still don’t think I really have a good idea of what the gameplay is like or how hard the puzzles are.

With that, I gave up on puzzle games and tried out Lego Batman, something that had struck me as a good idea back in 1997 when I played Lego Star Wars. After an overlong intro sequence involving some rather forced slapstick — perhaps my tastes have changed in the last four years? — I made Lego Batman run around and hit people for a few minutes, just long enough to decide that this is a game best controlled with a gamepad rather than mouse and keboard. But my system wouldn’t recognize my trusty DualShock + USB Adapter until I rebooted, and after that, it wouldn’t start the game again. It kept throwing up Windows “illegal operation” dialogs.

It’s likely that all these problems, and probably other recent problems as well (like my difficulties with Arthur’s Knights), have a common root in my hardware, probably that the fan on the video card is clogged with dust again or something similarly foolish. But I didn’t feel like doing anything as stressful and taxing as troubleshooting hardware, so I spent the rest of the evening watching a movie instead. At least I can scratch two of the three games off the list of things I’ve purchased but not actually played.

Popping back one stack frame, I got some more Super Meat Boy in yesterday. Can I just mention how catchy the music is in this game? Half the reason I came back to it just now is that it kept going through my head even as I was playing Machinarium. The game has three versions of most tunes: a clear-toned and bouncy one for the light world, a variation with more crunchy distortion for the dark world, and a chiptune version for the retro warp zones.

But I haven’t been hearing much of those in my latest sessions. I’ve been spending nearly all of my time in the world called The End, where the light-world music is less bouncy and more mock-epic, the dark world is extremely hard to access, and there don’t seem to be any warp zones. Possibly there are some lurking somewhere, but I suspect not, simply because this is the section that strips away all the distractions and just leaves pure challenge. There are no bandages to collect. You can’t even use any of your unlocked characters; apparently confronting Doctor Fetus is something Meat Boy has to do for himself. And, perhaps unintuitively, that’s what’s driven me to play only in The End. From the standpoint of making progress, the chief reason to go back to earlier levels (including dark world levels) is to unlock additional characters who can help you along. If they can’t help me any more, I might as well keep banging my head on level 6-5.

That’s where I am now, level 6-5. It’s the last level before the boss fight, and the only one I need to complete to unlock it. Its name is Omega, and I have just spent a great deal of time on it. It consists of five loosely-defined floating rectangular structures, bristling with buzzsaws, mostly navigable only by long-distance wall-jumps executed in specific places with split-second precision. Just getting into the first enclosed structure seemed impossibly hard when I started. By now, I’ve actually got to where I can see the level’s end a couple of times, but it’s clear that I’ll have to play for hours more before I can actually finish.

The thing is, I’m reluctant to stop playing again now. I’ve made a lot of progress on completing Omega, but it’s not permanent, tangible progress. It’s progress in the form of knowledge and muscle memory — “controller kata”, as an acquaintance of mine described it — and if I spend a week playing something else, there’s a good chance I’ll lose it. This is not stuff you can write down, for the most part. It’s about getting the right rhythm, and applying it without visual confirmation, like Tommy playing pinball. Dustin Hoffman’s character in Little Big Man, during his gunslinger phase, spoke of “firing a gun without touching it”, by which he seemed to mean performing the action so automatically that you aren’t aware of the weight of the gun in your hand until afterward. That’s more or less how I now feel about the earlier actions in this level. The repetition becomes a kind of meditation.

Except that, even in this state, I’m not executing perfectly — in fact, I’m executing so imperfectly that I only occasionally reach the point near the end that I don’t actually know how to execute. Oh, I manage each particular bit on most attempts at it, but the probabilities multiply out to majority failure. It makes me wonder to what degree meditation exercises of the purely mental sort are subject to error and variation that the meditator doesn’t notice because there’s no machine judging correctness and making you start over.

I said that I’d finish Machinarium on the PC rather than the iPad, but it turns out I was wrong. I can blame my lengthy bus commute, but that’s only part of it. Despite what I said before, it turns out that the touchscreen version of the interface is easier to use in some situations, particularly when you’re pressing on-screen buttons repeatedly. With a touchscreen, you can hold one finger over each button, essentially treating the screen like a keyboard. With a mouse, and without the hotkeys that usually accompany button-based interfaces in PC apps, you have to keep looking back at the buttons to reposition the cursor over the one you want, and that means briefly looking away from whatever the button affects. This is particularly bad in action sequences.

Action sequences? Yes, there are a few, adaptations of old videogames. I’ve already described one: the shooter that grants access to the hints. In addition, there’s a simplified Space Invaders at one point, and, towards the end, a maze-based shooter in a style that reminded me a lot of Atari 2600 Adventure (even if the gameplay was more like Berserk). The context for the Space Invaders is simply an arcade, but the maze game seems to be about cleaning up the software corruption left behind by the bad guys in the mind of the big-headed robot in the city’s central tower.

The bad guys in question are a small band of criminals in black hats who are seen stealing things and even planting a bomb throughout and before the game. Josef recognizes them: they’re responsible, it turns out, for his condition at the beginning of the game, in pieces in a scrapyard, and also show up in a few little flashbacks where Josef remembers when they were mere schoolyard bullies, shaking him down for pocket change and knocking him off the jungle gym, thereby justifying any horrible thing Josef might do to him in return.

Which, of course, provokes the question of whether, and why, robots need to go to school, but the rule of this game is that robots can engage in any sort of human behavior if it’s funny or makes for a decent puzzle. A couple of scenes have toilets in them. An early scene in a jail cell has a cellmate who wants a cigarette, although I suppose that in a sense it makes more sense for robots to smoke than for people to do so. One of the bad guys’ nefarious deeds was to kidnap Josef’s girlfriend and force her to work in the kitchen of a sleazy bar, raising the question of why a robot needs a kitchen, although somehow the “girlfriend” part, with the implication that robots are gendered, doesn’t seem so strange. Well, our concept of gender is at least as much social as biological. Presumably it’s entirely social for robots.

There’s a brief bit where the player even gets to control the girlfriend, which yields one of the game’s better jokes, especially considering that it’s a repeat of a joke you’ve seen over and over by that point. One of the first things we learn about Josef is that his inventory is in his abdomen, and whenever he picks something up or takes it out to use it, he hinges his head open at the mouth like the lid of a trash can. It’s a great sight gag because it combines so many incongruities: he’s turning himself into an inanimate object, but he’s also eating something or vomiting it out, while at the same time being completely unconcerned about what it is or what it’s made of unless its physical properties affect the process of insertion or retrieval (as when he sucks in a length of hose like a child eating spaghetti). Now, one of the things that genders the girlfriend is that her face is more delicate and less machine-like than Josef’s. It would still look monstrous on a human, but relative to the other robots, she’s downright pretty. So when she handles inventory the same way — something she doesn’t even look physically capable of doing — it comes as an extra shock. But it’s also touching in a way, because it also reinforces the sense of a bond between the two of them. They’re two little robot geeks who approach the world the same way.

Also, it helps that there are flashbacks, in the form of line art in thought balloons, showing Josef and girlfriend in happier times. I’ve only seen a couple such — apparently more appear when you stand still in certain locations, which means you’re bound to see one or two over the course of the game, but not more than that unless you’re looking for them. Now that I’m done, I may do that. This strikes me as something that’s missing from most games with kidnapped-girlfriend plots: some indication of what the hero is trying to recover.