The quest for a perfect score: Part 1

by ersatz_cats

There’s a phenomenon I’ve noticed, which happens when a video gamer gets busted for cheating (or at least whenever there’s a cheating controversy). While the wasted time and heightened distrust are a net negative for the community, sometimes the added attention does result in an influx of honest competitors taking up the game in the immediate aftermath of the scandal. You saw it with Minecraft, and Trackmania – and that’s not even counting the surge of fresh Atari 2600 players in 2017 vying for the newly-discovered-to-be-vacant record on Dragster. This could be the result of an innate human gravitation toward conflict and controversy, or perhaps more optimistically, the coverage of an important community matter results in people incidentally learning enough about the game in question to find themselves both interested in playing it and up to speed on the basics for competition.

Whatever the cause, I found myself in a similar situation a while back, having spent the previous year before that mired in research on Pac-Man strategies and early competition on the game. Reading about cornering mechanics and ghost speeds and forward motion patterns and safe spots, I found myself absorbing many of the lessons needed for top play. While some have self-servingly chosen to portray a perfect score on original Pac-Man as an impossible “holy grail”, attainable by only a select few separated from the masses by some manner of divine provenance, my research helped dispel the illusion that the maximum score was beyond my grasp. While I am by no means any sort of gaming god, I came to believe I could execute each of a perfect score’s component parts in isolation, and with practice, could likely do so in sequence. I also, by quirk of fate, ended up with ownership of a website domain called perfectpacman.com, and it only seemed appropriate that the owner of such a domain actually be a perfect score player.

As you might have surmised by the “Part 1” in the title, I have not yet finished ascending that metaphorical mountain. But I had some time off in October, and a golden opportunity to take my first steps toward this goal. I’ve decided to take a moment to document my journey here, for the entertainment of readers of this site, for future aspiring perfect score players looking for assistance, and perhaps even to inspire others to begin this quest for themselves.

NUTS AND BOLTS

As a quick brush-up on the basics of scoring, there are four sources of points in Pac-Man: Regular dots, power pellets (also called “energizers”), eaten ghosts, and fruit. Eating the dots and power pellets is a requisite for advancing each board, and so they’re taken for granted. Eating each ghost is far and away the most difficult aspect of maximizing your points, at least on the early boards. Each time you eat a power pellet, the four ghosts turn dark blue and become vulnerable, with the first being worth 200 points, the next 400, the next 800, and the last 1600. That means you have to eat all four ghosts off each power pellet each time. On six boards (highlighted above), the “blue time” only lasts a single second, requiring you to group all the ghosts tightly together near the energizer for prompt devouring. Thankfully, the ghosts don’t turn blue at all from board 19 on (and also on board 17), which means the fruit, which appears twice each board, is then the only non-requisite collectible remaining. (Yes, bells and keys are “fruit” now. Don’t @ me.)

On board 256, a programming bug causes about half the board to be overwritten by computer garbage. (Interestingly, you can blame the algorithm that draws fruits in the bottom right corner for that.) There is no way to pass this board, since you can never eat enough dots to register a full board completion. For this reason, rather than being an endurance test of how long you can marathon a single playthrough, the game of Pac-Man has a maximum and finite score. There is some historical debate as to what exactly constitutes a “perfect score”, given that dots collected among the computer garbage on this screen clearly constitute glitch abuse (something which was typically disallowed in gaming’s early days). Also, as with all arcade machines of its era, Pac-Man has a series of dip switches to manipulate game functions, including two that grant the player a variable number of lives, which becomes relevant due to the aforementioned glitch abuse. How to approach the split screen will be a topic for “Part 2” (or maybe even “Part 3”), however the maximum score that can be achieved, on settings which grant the maximum possible number of lives, is 3,333,360 points.

But we have a long way to go to get there.

For such a simple game in appearance, using only a single repetitive background, I was surprised at the mechanical details baked into the game’s design. Subtle quirks that might mean nothing to the novice player (such as ghost reversals) are actually important cues, which expert players can even anticipate and exploit. Before one can make serious attempts at a maximum score on original Pac-Man, one should understand what these mechanics are, and how they can be worked to one’s advantage. It’s not necessary to fully recall all of these finer details to get started, but at least an overview of what’s going on will ensure your practice is productive. If you want more in-depth descriptions of the game and its underlying mechanics, I highly recommend Jamey Pittman’s excellent and highly-detailed resource “The Pac-Man Dossier”, from which the rest of the information in this section is primarily sourced:

https://pacman.holenet.info/

Of course, there are the types of mechanical expressions one can expect from an early 1980s video game trying to present itself as a straightforward challenge of ghost evasion by way of crude programming shortcuts. Rather than the board being treated as some kind of fully open world, the screen is divided neatly into about 900 small “tiles”, such that (among other things) each little dot has a game tile to itself:

As an aside, the tile layout helps explain why as Pac-Man you will occasionally pass right through an oncoming ghost. The game registers collision between Pac-Man and a ghost when their positions are both on the same tile. However, with Pac-Man going in one direction and a ghost heading toward Pac-Man in the opposite direction, it’s possible for both sprites to cross over to each other’s previous tile on the same frame, resulting in the game never recognizing either as occupying the same tile at the same time.

Beyond the foundational tile structure underpinning the game’s mechanics, the next most important thing to understand is your antagonists. While the ghosts were given varying bright colors to make the game more enticing, the fact that each has a different color and name is not just a matter of aesthetics. Each ghost has a distinct AI script, giving it a “personality” of sorts. This was done to avoid the unsatisfying challenge of having four identical ghosts always trailing Pac-Man in a straight line. Instead, the threat can come from behind you, or beside you, or in front of you. While each ghost does have three different “modes”, most of the time you’ll be interacting with them in what is called “chase” mode, where they are more-or-less actively trying to pursue you.

In chase mode, each ghost is given a target tile. Blinky, the red one, is the only ghost whose target literally is Pac-Man (or rather, the tile Pac-Man is currently on.) Pinky’s target is four tiles in front of whichever direction Pac-Man is facing. Since Blinky tends to chase you from behind, while Pinky tends to ambush you in the front, they represent your two biggest threats. Meanwhile, the orange ghost, Clyde (sometimes called “Pokey”), alternates between two targets, making him less consistent. When he’s eight or more tiles away from Pac-Man, he behaves like Blinky, moving right in your direction. However, when he’s within seven tiles of your location, Clyde’s target changes to the SW corner of the board. As a result, Clyde is typically the ghost wandering off in some far location, unable to decide whether he’s coming or going.

The targeting system for Inky, the blue ghost, is a whole ‘nother story, involving a calculation harkening back to high school geometry class. While it isn’t necessary to remember it exactly, I’ll attempt a written description for the record. First, you take Blinky’s position, then you take the position two tiles in front of Pac-Man, which I’ll call Inky’s “subtarget”. Then, you draw a line (or, geometrically speaking, a “ray”) starting at Blinky, going through Inky’s subtarget, extending indefinetly in that direction. Finally, as if using a compass on paper, you find the point along that line that is exactly as far away from the subtarget as that subtarget was from Blinky. Since I’m sure a verbal description is likely to produce many glazed eyes, here are a few visual examples of what I mean, with a green “S” tile as the subtarget and a blue “T” tile as Inky’s actual target:

As you can imagine, while the other ghosts’ targets are in motion, Inky’s target can really be in motion, as the two source points in this geometric calculation (Blinky and Pac-Man) are almost always changing position, sometimes in conflicting directions. However, if Blinky’s on your tail, and you’re facing away from him, Inky will typically follow right along with both of you.

Note that this ghost targeting system is bugged when Pac-Man is facing up. In that case, the target (or subtarget) is additionally offset to the left. This doesn’t affect either Blinky or Clyde, whose targets do not involve Pac-Man’s direction, but it does affect Pinky’s target (which is now four tiles up and four tiles left of Pac-Man) and Inky’s subtarget (which is similarly two tiles left of where it would have been).

The important takeaway is knowing which threat is approaching you at any given moment. If you’re freehanding a board, and need to shake Blinky off your tail, doing so will require different movement than shaking Pinky or Clyde off your tail. It’s also worth remembering that the SW corner of the board is the most dangerous, by virtue of the fact that as long as you’re in that corner Clyde will be targeting you no matter his position.

I mentioned a moment ago that each ghost operates in three “modes”, with the aforementioned descriptions applying to their default “chase” mode. If you’ve played more than one game of Pac-Man, then you’re already familiar with “frightened” mode, which begins any time you eat one of the big dots, called an “energizer” or “power pellet”. For a set duration (the length of which depends on which board you’ve advanced to), the game’s music changes as the ghosts turn dark blue, reverse direction, and slow down. As mentioned before, it is during this interval that they each become vulnerable to Pac-Man’s chomping, with all four ghosts worth a total of 3000 points. While in frightened mode, rather than using the target system, the ghosts’ turns become pseudo-random, making their path less predictable if you fail to eat them quickly. Note of course that later levels have no “blue time” whatsoever (meaning, the ghosts never enter frightened mode even when you eat an energizer), but we’ll get more into that later. Even on boards with no “blue time”, the ghosts still reverse direction when you eat an energizer, giving you a useful tool to keep those approaching apparitions off your tail.

The third ghost mode is called “scatter” mode, and happens at regular timed intervals. This is visually indicated by all four ghosts spontaneously reversing direction all at the same time (and not because you ate a power pellet). From a game design standpoint, the reason scatter mode exists is because the game’s designer, Toru Iwatani, wanted the game to have a recurring sense that the ghosts were all closing in on you. However, this is only possible if the ghosts occasionally let off the pressure momentarily, so they can begin closing in on the player again. (This is similar to a song or a movie that can’t always be upping the intensity, and must occasionally relax the tension so it can begin building up again.) While in scatter mode, each ghost targets a different corner of the board, briefly sending them off to the four winds before they switch back to chase mode and resume pursuit. Scatter mode operates on a global timer, usually occurring every 20 or so seconds, with scatter mode itself typically lasting for either 5 or 7 seconds, making it something expert players can anticipate and use to their advantage. (Refer to Pittman’s guide for the exact timetable.) Note that if you stay on a single board long enough, the ghosts never enter scatter mode on that screen again, staying in chase mode indefinitely.

Sprite speed is another mechanical detail which aspiring perfect score players should have at least a surface understanding of. It doesn’t take a player long to notice that the first board (the Cherry board) runs slower than boards 2-4, and that subsequent boards run even faster. This is true of both the Pac-Man sprite and the ghost sprites. Normally Pac-Man moves faster than the ghosts, preventing them from simply running him down, but he does move slower when eating dots and energizers, allowing them to overtake you during long stretches of uninterrupted dot consumption if you’re not careful. However, starting on board 21 (also known as the “9th Key”), and until the end of the game, Pac-Man slows down overall while the ghosts maintain their top speed, making it much easier for the ghosts to catch you in a race.

As if this wasn’t bad enough, Blinky (the red ghost) is programmed to get even faster as you clear dots on the screen. At two set intervals on each board (such as when there are 60 dots remaining, then again when there are 30 dots remaining), Blinky gets a speed boost, turning into a heat-seeking missile players colloquially call “Cruise Elroy”. Recall that Blinky’s modus operandi is to attack from behind. Thus, this acceleration keeps him from perpetually eating your dust if you aren’t careful. In fact, the final Cruise Elroy speed is faster than Pac-Man on every board, and especially faster on the 9th Key boards where Pac-Man moves slower, making him a particular threat to watch out for.

But even when those encroaching house-haunters are closing in, there are still a few techniques you can use to escape. The most obvious one is that, even on those later boards, the ghosts slow down to almost half their speed when entering the side tunnel, allowing you to give yourself some breathing room. But the coolest way to shake off those stalking specters is to take advantage of these forbidden turns hidden in the game. These are the four turns (seen above) that you the player can make, which the ghosts cannot. (Try it if you don’t believe me!) The ghosts can still travel those paths coming from the other direction, so those spots are not totally safe, but they can’t enter from the direction specified. Even on the advanced boards, the ghosts can be right on your tail, approaching for the kill, but if you’re near any of these spots, a quick exit can have them wandering the board again – provided that you don’t forget to watch for another stray ghost coming at you from the other end of that turn. (Technically, this restriction only applies to chase and scatter modes. The pseudo-random movement of frightened mode can choose these turns.) This also makes the top portion of the board a bit safer in general, as the ghosts can only enter that area through the two side paths – although that can also be a downside if you’re attempting to gather the ghosts to be eaten near the top two energizers.

These forbidden turns, in conjunction with the ghost AI and some astute timing, actually do provide safe spots where the player can park their Pac-Man, in some cases indefinitely. The most common safe spot is seen in the above example, located below the bottom right corner of the ghost pen. Blinky chooses whichever direction will take him closest to you, excluding the forbidden turn, resulting in him looping the bottom “T” forever. Clyde approaches you from the left, then decides he’s too close and switches to targeting the SW corner, then gets far away and switches back to targeting you, stuck in his own endless loop. Since you’re facing up, Pinky targets a spot just below the ghost pen, and thus circles that pen indefinitely. Inky is a bit more unpredictable, but typically alternates between a few different paths that keep him away from that critical area to your direct right. These ghosts will break out of this routine every time they enter scatter mode, however they will typically resume these routines when they re-enter chase mode. Eventually, the pre-determined scatter modes will cease, and you can stay safely in this spot for as long as your game remains powered on. Note that, due to this confluence of ghost AI and other factors, your safety in spots like this depends on the positioning of the ghosts when you enter that spot. This is not a skill I have particularly mastered, making it perhaps a topic to revisit in “Part 2”. However, in my attempts I was able to safely park in this spot numerous times during breaks without much trouble.

The last game mechanic to understand before even beginning discussing perfect score strategy is the cornering mechanic. While each path down the game’s aisles is represented by the appearance of a single straight line, the way Pac-Man may enter that path can vary significantly. To put it simply, you can begin each turn slightly before you actually arrive at that turn. It’s not something apparent to the naked eye, however turns using this fast cornering are obvious when examined frame-by-frame. (While I don’t know if this is intentional, the rounded edges seen on each corner could be thought of as a visual representation of this mechanic.) This fast cornering is important both in how it is applied in continuous motion patterns (which we’ll get to next), and in how it can give Pac-Man a slight distance boost when attempting to outrun those wily wights who can’t use quick cornering, especially on the later boards where the ghost speed is innately faster than yours.

PATTERNS WITHIN PATTERNS

With the mechanics out of the way, it’s time to talk strategies. The most common distinction you’ll hear is between “pattern” play and “freehand” play – so much so that the site Pac-Man Forum notes the playstyles of the perfect score players they recognize:

https://www.pacman-forum.co.uk/perfect-scores-arcade/

When playing “freehand”, the player uses repetitive grouping (sometimes called “clumping”) techniques to patiently gather the ghosts as closely together as possible, before directing them to the vicinity of an energizer so they can be eaten within the brief “blue time” window. Elaboration on what these “freehand” techniques are will be a subject for a future installment, once I myself learn more about them.

Patterns, however, are an entirely different beast. Through extensive trial-and-error (either on MAME or on a Pac-Man cabinet set to free play), experts developed set routines through which they can clear each board methodically while collecting each fruit and each ghost along the way. The patterns work because there is no true randomness in the game. (Even the pseudo-randomness of the ghosts’ blue time behavior is dependent on your play on each board, allowing pattern designers to account for it.) It wasn’t long in Pac-Man’s life cycle that books and magazines provided players all sorts of patterns, such as this pattern for the 9th Key found in 1982’s The Video Master’s Guide to Pac-Man by Jim Sykora and John Birkner:

Before we continue, I should note that the distinction between pattern and freehand play (such as seen on Pac-Man Forum) applies only to the first 20 boards. From boards 21 through 255, a single repeating 9th Key pattern will get every dot and both fruits on each board over the course of about three hours. Every perfect score player uses patterns for that stretch, known as “the Desert”, both to make play times manageable, and to mitigate the aforementioned dangers of the Pac-Man sprite’s slower speed.

When I began reading about the finer points of Pac-Man, one of my early questions was why there had to be so many different patterns on so many different boards. Okay, the patterns for boards where you don’t eat the ghosts have to be different, since you aren’t sending devoured ghosts back to the pen. But surely, for the “blue time” boards, there could be just one single pattern that works on every board, could there not? However, the mere existence and length of blue time is not the only factor that has to be taken into account. Obviously sprite speed is another, since the ghosts do not have a consistent speed relative to yours as the early boards get faster. (In other words, on board 2 you get an 8% speed increase, while the ghosts get a 7.5% speed increase.) Even minute differences in timing, location, and even the direction Pac-Man is facing have cascading effects that can throw a pre-arranged pattern totally out of alignment. Another more subtle factor is that the timing of Blinky’s switch to “Cruise Elroy” changes after every three or so boards. On board 12, Blinky enters his first “Cruise Elroy” gear with 80 dots remaining, but on board 15, he does so with 100 dots remaining. Thus, even with all other factors being identical between those two boards, a pattern for board 12 will not necessarily work on board 15. As we’ll see in a moment, many of the available patterns have identical beginnings, but deviate as the pattern progresses, often specifically because “Cruise Elroy” otherwise goes out of alignment.

In the olden days, some players considered their perfect score patterns to be proprietary, and thus treated them as closely guarded secrets. “After all,” so the logic goes, “I spent so much time and put in so much work developing these patterns. Why should someone else come along and claim the fruits of my labor for their own?” Of course, we’re in a much different gaming era now, and not just because of the ubiquity of information on the Internet. While the idea of preserving “secret strategies” might have some allure in theory, in practice it served to protect habitual liars and cheaters, preventing the community from discovering that their private submissions were either illegitimate or nonexistent. There is also a newer appreciation that, while the developers of advanced strategies should certainly be recognized for their work and contributions in that field, gaming competition is only truly fair if the participants all have access to the same information, and thus can compete on an even playing field. Or to put it another way, if a new strategy doesn’t simply make the game too easy, the “world record” still comes down to execution. Modern speedrunning is also seen as more of a group effort, where the community collaborates to push the limits of a given game, even if an individual player is ultimately credited as holding the world record.

As noted before, perfect score patterns of varying quality had been printed in books and magazines in Pac-Man’s early days. However, many of these included pauses and reversals, which can lead to a pattern being unreliable, or unnecessarily difficult to execute. The more such pauses and joystick flicks, the more predictably susceptible a pattern is to human error. What were really sought were “continuous motion” patterns, where the player is always moving Pac-Man in a desired direction, eliminating these possible points of failure. As late as 2005, players on the Internet were still seeking reliable patterns for Pac-Man’s most difficult boards. However, that changed later that year when legendary Pac-Man master Neil Chapman published his perfect score patterns for all to see:

https://nrchapman.com/pacman/

I had decided my first approach to a perfect score would be to simply learn the patterns and make attempts at executing them. While a true Pac-Man master would ultimately conquer both pattern and freehand play, giving them the tools to recover even when a pattern fails, I figured that as a novice I could rely on unlimited restarts and simply not put any faith in my ability to recover from a broken pattern (aside from whatever sloppy grouping and evasion techniques I might be able to improvise).

While Chapman’s patterns were obviously outstanding and expertly designed, strictly for my own purposes, I still wasn’t quite satisfied with two of them. Neil’s Strawberry pattern included a “small pause”, leading me to seek another pattern on YouTube eliminating that potential point of error. (In hindsight, Neil’s pause is really insignificant, as the board is almost entirely clear at that point.) Neil’s 2nd Apple pattern also included a small joystick flick toward the end, which I suspected could be eliminated. (Indeed, Neil has since uploaded a modification of that pattern missing that flick.) I also went with a common 9th Key pattern called “Stacked”, since I had already watched enough of it to be familiar with it.

My first mission was to commit each of the selected patterns to memory, relying primarily on Neil’s own outlines. However, I noticed I had difficulty following each sequence in my mind, with the way Neil had broken the diagrams of these patterns into several little sections. What I wanted was a single board showing the entirety of the pattern, on which I could trace my finger (either physically or mentally). Thus, instead of learning seven short segments, I could approach the pattern as the single continuous thread that it would be in live game play. Neil’s displays also didn’t always acknowledge the similarities between many of these patterns. All of the “two second” boards (where the blue time is two seconds) were helpfully acknowledged as a single page, but patterns which shared the same opening while deviating late were posted separately, effectively treating them as two distinct patterns to be fully memorized. I wanted the process of pattern memorization to be as accessible as possible.

To that end, I spent some time redrawing Neil’s patterns in my own style. While all but the 9th Key pattern did end up requiring two screens (and the complicated “one second” patterns required four), this still satisfied my desire to have boards I could trace along continuously. In a way, splitting these patterns into two halves also became a feature, in that it allowed me to focus on consistent beginnings of patterns and treat the later deviations as separate sequences. Although redrawing these patterns in this way was a tedious task, I decided it was time well-spent, as it helped me familiarize myself with the subtle details I could expect on each and every turn along the way. Later on, I also added little marks to indicate when ghosts and fruit are eaten, since the ghosts especially affect the timing of your play.

I’ve compiled these redrawn patterns into a zip file, containing 22 images in all, in the hope that other aspiring perfect score players may find them of benefit. Let me reiterate that most of these patterns represent Neil Chapman’s fantastic and meticulous work. I have omitted all of Neil’s original clarification notes, which you should review directly on Chapman’s site. This includes important notes on which direction Pac-Man should be facing at specific moments such as ghost capture points, which as we’ll see become crucial to the correct execution of the pattern. My redrawings of the patterns attempt to reflect these details, albeit in a way that does not explicitly call them out. Again, all I’ve done is provide a different visual medium for Neil’s outstanding work.

https://perfectpacman.com/wp-content/uploads/2022/11/Pac-Man-patterns.zip

While redrawing the patterns helped, it was only a first step. When attempting to memorize large amounts of data, it’s good to understand how the human memory works. In the 1950s, psychologist George A. Miller observed that most people’s short term memory is limited to seven objects (such as a seven-digit phone number), but that a process called “chunking” can extend that power. For instance, remembering nine letters can tax the human brain, but remembering a single nine-letter word is effortless, as it can be memorized as a single object. Association also assists with memory. Remembering the eight planets is made easier by memorizing the phrase “My very excellent mother just served us nachos”.

To this end, I developed personal scripts to go along with the patterns. While remembering “right, up, left, down, left, up, right, down, straight, straight, right, down, right, up, right, up, right, left, up, left” might be taxing, a lengthy sequence could more easily be remembered as a single maneuver identified with, say, a line from the Simpsons (such as Principal Skinner reciting “Over, under, in and out, that’s what shoe-tying’s all about”).

In writing these scripts, I developed my own terminology for maneuvers and locations on the board, to be used as shorthand in these scripts. The two spots at the top became the “sun” and the “moon”. The s-channels on the bottom became either “chutes” or “ladders”, depending on which way you were going. (I guess in other countries “Chutes and Ladders” is known as “Snakes and Ladders”.) The main path around the board became the “highway”, and the area inside was the “car”, which led to the lower portions being labeled “wheels” and “bumpers” appropriately. (I suppose those inside “corners”, which got named well after the “posts”, could’ve been the “trunk”.) Names like the “mosh pit”, or Muhammad Ali’s famous “float like a butterfly” line, let me easily identify maneuvers common among several different patterns.

As one example, I filled my script for the Strawberry board with lines from Beatles songs, including “here comes the sun” (meaning, head up toward the sun), “yes I’m gonna be a star” (head past the sun and swing over the moon), “take you down because I’m going to strawberry fields” (head down, then swing around for the strawberry), and “there will be an answer, let it be” (meaning, don’t turn at the intersection, keep going all the way to the wall). The most common pattern, first seen on 1st Apple, has four variants, one of which is seen four times. I fashioned this pattern after a fictional nightclub called the “Grapeland”, since the default variant is used on both Grape boards (although admittedly Pac-Man’s “Grape” icon, as it is sometimes called, may actually be a melon). The four boards with the basic “Grapeland” pattern would be normal nights at the nightclub, but occasionally they do 90’s night (2nd Apple, where my script includes lines from 90’s dance music), 80’s night (2nd Galaxian, with a script full of 80’s song lyrics), and 70’s night (2nd Key, 70’s lyrics). Movie lines, Star Trek quotes, sports references, you name it. All of it personalized to what I could remember. Anything to turn this mess of maneuvers into a halfway cogent narrative.

Scripting out the patterns also helped make my work more portable, as it was now text instead of graphics. After months of memorization time, I could trace out any given board in the air at any moment, or alternatively, I could recite my whole script for that board, which had the same effect. While the script wasn’t strictly necessary, memorizing one assisted in memorizing the other.

For a little in-game assistance, I produced a small “cheat sheet” I could have open next to my game screen as I played. The thick lines represent the game’s five intermissions, which divide the early boards into six sections. I also later added a reminder column for which “post” (bottom corner of the board) I should head toward first, just in case it takes me a moment to recall the specific pattern I should run for that board. (The Strawberry pattern I used was the only one with a special opening, going down the right chute.)

Late in the process, I had another idea on how to assist applying all this memorization to actual game play. It’s one thing to know each of the patterns in isolation, but it’s another to recall which one to execute in what order, especially when dealing with different patterns which share common openings. As a comparison to other games, I pondered on what allows me to immediately know what lies ahead of me any time I begin a level on, say, Super Mario World. A big part of that is the visual look. Due to technical limitations, games around the time of Pac-Man’s release tended to have few backgrounds which are heavily repeated. However, on later games, different levels tend to look distinct, using different sprites and color palettes, which in addition to giving the game an engaging look also cue players into what practiced routine to run. Obviously, aside from the fruit counter in the corner, every board on Pac-Man looks identical, making such visual cues difficult. However, on-screen graphics are not the only media at a Pac-Man player’s disposal. I considered the idea of putting together an external music compilation, which could run for the first 20 boards and which I could reset every time I restart a game attempt. Since each pattern has a timed duration, you could set it up so Van McCoy’s “The Hustle” always starts when you begin the “70’s night” board, or Smashing Pumpkins plays while you do your board scripted by grunge rock lyrics. Yes, if you go off-pattern, then the music becomes awkwardly desynced from your game progress, but assuming you stay on track, the music would help remind you which variation you should be playing on each board. It wasn’t long after I had this idea that I realized some disgruntled nitpickers may object to such a timed soundtrack as an unfair advantage. But to be perfectly honest, imagining their protests only made me want to use this method even more. “That’s not fair! He had outside musical assistance!” I ended up not compiling such a soundtrack for my October attempts, but it is something I could consider revisiting in the future.

The last thing I needed in my preparation was to decide what sort of controls I would use. I didn’t own an actual Pac-Man cabinet, and wished to do these attempts in my home and not at some barcade or something way up north, so MAME was the way to go. (While I obviously can’t share the ROM, if you want to use the previously linked patterns, you should be looking for the ROM set wherein the attract mode ends with Inky catching Pac-Man above the SW energizer, as seen in the first example from this video.) Admittedly, I’m a complete noob to everything arcade, having grown up on home consoles. Original Pac-Man uses what’s called a “leaf” joystick, which I assume is named as such because the underside sort of resembles a tree with four branches. At the tips of these branches are metal pads which make contact with surfaces external to the joystick mechanism, resulting in easy and silent movement. This is in contrast to the “microswitch” style joysticks common to later Pac-Man releases, where the connectors are inside the joystick mechanism. Not really knowing about this distinction, I looked for a well-rated PC joystick, the options for which appeared to universally be of the microswitch variety. The most important factor for me was direction restriction, since I wanted to make sure that when I played games like Pac-Man I could move strictly up and not in a vaguely up-like direction (which could be far enough to either side as to register in MAME as left or right). I ended up settling on a Mayflash F500, which among other things came with an 8-way restrictor gate that I could easily install.

At any rate, between various real life considerations, I had a good idea I would have a block of free time in October ideal for streaming. I set that as my target, drilling myself on pattern memorization for weeks, even as I’d go about my daily business around town. When that time came, I wished to be ready to go.

IMMEDIATE FEEDBACK

Of course, no amount of reading strategies and studying charts can fully prepare you for when you sit down and fire up the game. Obviously, I had played Pac-Man before at various points in my life, albeit always on re-releases and multicades. But for this, I wanted to see how I would perform from a fresh, otherwise fully prepared start. For over a year, I intentionally starved myself of Pac-Man and Ms. Pac-Man, relying exclusively on external information and research. This included relying on others when I had questions about Pac-Man relevant to my other research projects which could have been readily addressed with a quick bit of game play. Even if I saw a “Reunion” machine at a bar or theater here in Olympia, I limited myself to Galaga, waiting for this streaming window, which had now arrived.

I used a stream layout with a manually tracked “restart” counter in the upper right corner. For a background, I recycled a full-size perfect score chart I had previously made for a YouTube video. At the moment when you begin something you had been looking forward to for over a year, there’s always a bit of anxiety. It can’t be helped. But as with a new home console that momentarily dazzles you with its technological superiority, all of those feelings soon pass, and it becomes just a contest between you and the game.

I struggled for a while to complete the perfect pattern for the Cherry board. Obviously I knew the pattern on paper quite well. But the execution was more difficult than I was expecting. When you eat the SE energizer, the one ghost is supposed to be off to the left, but the other three should be coming right down in your direction. Occasionally this would go according to plan, but often one or two would be wandering off. Other times, they were all a mess. This was literally the first power pellet of the entire game, and I was struggling to get beyond it.

That was where the size of the chasm separating theory and practice really hit home. I knew of Pac-Man’s quick cornering mechanic, but I was not prepared for how it affected game play. For the pattern to hold, you need to be hitting each of those turns as early as possible. The pattern is designed that way to eliminate human error, allowing the player to simply buffer all their turns in advance. Otherwise, the player would have to be hitting each of these thousands of turns each on perfect frames, and obviously no one would be doing that, lmao.

A slight delay on the turn affects Pac-Man’s positioning, which can affect the timing of his direction changes, which affects ghost targets, etc. These seemingly minor fluctuations can snowball out of hand very quickly. All of that I knew on paper, but I didn’t expect the massive impact it has on game play. It’s almost like playing with near-constant input lag, where the key to success (pun always intended) is to make each turn as you’re watching the previous turn. In narrow corridors like the winding “S”-channels, I found I got my best results by tricking my brain into holding up to go left, then left to go up, then up the moment you’re going left again. It feels wrong as you do it, but it produces the desired results.

I also learned the hard way that those pattern notes from Neil Chapman are there for a reason. Again, it’s easy to acknowledge that on paper, but it’s something else to see it play out. At that first energizer, if you turn left after the second ghost, you miss the third one entirely, despite visually appearing to collide with it. Neil’s instruction to hold up until you see “800” is not merely a suggestion.

Later on, near the NW energizer, you really do have to hold right as you approach a frightened Inky, such that you actually see Pac-Man face right for an instant before the eaten ghost animation begins – if you don’t see that right turn, the pattern breaks down later. (Being a child of the ’80s, I began confirming whenever I saw that right turn frame by declaring that “Mikey likes it”.) I soon settled in, and was able to clear the first stage on my sixteenth attempt.

Having finally conquered the Cherry board, I arrived at the Strawberry board, and a boost in game speed. Once again, this took many tries to become accustomed to. The difference here was that one point in particular always seemed to be my downfall. On the earlier Cherry board, I would subtly go off-pattern at some point, causing the ghosts to go out of alignment in a way that would spell my downfall some time much later on the board. However, the Strawberry pattern I was using included a spot where you were intended to cut narrowly in front of Blinky near the SW energizer. Even a pixel delay in the pattern up to that point would cause Blinky to overtake you before you could reach the life-saving power pellet. I began calling this an “immediate feedback” moment. Unlike the spots where you miss a turn and don’t discover you’ve been off-pattern until much later, this turn with Blinky tells you right away whether you’ve been doing perfect turns up until that point. As I remarked on the stream, I was “playing chicken with Blinky”, and at least at first, Blinky was winning more often than I was.

Various friends swung by to spectate and chat, or to comment on my loud microswitch joystick (which admittedly I was giving quite a beating to). Of course, there were plenty of jokes to go around:

Twitch chat was not all jokes, though. David Race, current Pac-Man champion, dropped in regularly to watch my progress. Recent perfect score player Derek Sevcik also showed up to offer encouragement. He noted that the Strawberry pattern I was using may be more difficult than Neil Chapman’s. However, aside from not wanting to spontaneously re-learn a whole board pattern, I knew the pattern was not the problem. I simply had to get better at execution. I had to pay attention to my own play, and notice where I was absent-mindedly taking sloppy turns. The immediate feedback of “Chicken with Blinky” was forcing me to improve. It was a constant, blunt reminder that I couldn’t let up my focus, not on a single turn.

Indeed, Blinky kept me honest, as did another “immediate feedback” moment with Pinky on the 2nd Peach, which I called “beating the Pinky train”. (Thankfully, no matter how close Pac-Man and Pinky appear there, we never occupy the same game tile at the same time.) By the end of day 1, after 156 attempts, I had completed a perfect run through the Cherry board, the Strawberry board, and both Peaches consecutively:

Day two was partially interrupted by baseball playoffs on television. Still, in the morning and evening, I was able to add the first Apple board to my consecutive streak (albeit not always consistently). Also by this time, I had gone into the dip switch menu to change my settings from the default three lives to one life, so I wouldn’t have to constantly kill off remaining lives each time I wanted to restart – a practice that was made even more annoying by Pac-Man’s occasional habit of passing through the ghost I was attempting to run into. (Note that you can’t simply take another shot at the same pattern after a death, since Pac-Man’s dot-eating speed is different than his speed through clear channels.)

Day two was also when I began playing with the “rack advance” feature to skip boards at will to do repeated practice on later levels. Some stream viewers asked why I wouldn’t simply save state for this practice; aside from the fact I was playing on a WolfMAME variant which did not seem to have save state functionality, I wished to play more or less in a fashion that a player back in the day would have done. Normally this might involve no board skipping at all, as a player on a locked cabinet in an arcade wouldn’t have access to the game switches. However, some players even back then did have access to their own personal machines, and so I decided the built-in “rack advance” feature was fair game. Using this feature, I knocked off the 2nd Apple and 1st Grape boards as well, albeit not yet consecutively.

At this time, I added a feature to my stream, which for some reason I hadn’t thought of before: a perfect score tracker. Using the fruits to identify each board, I made the fruits for the first five boards I had achieved consecutively regular-sized and in color, while the boards I had skipped ahead to do were half size and monochrome. (Yes, technically the 1st Apple and 1st Grape pattern is identical to the 2nd Grape, but I did not want to add the 2nd Grape icon until I had actually achieved a perfect on that specific board.)

In addition to the practice necessary to run the patterns successfully, I was also becoming more accustomed to spotting failed patterns in progress, which allowed me to bail on them before getting pinched by oncoming ghosts from both directions. Take the above example from the Cherry board, for instance. When you cut in near the ghosts at the 48 second mark, Clyde shouldn’t be pursuing you right alongside Inky. When he is, it’s an indication that later on, Inky will not join Blinky along the left side of the ghost pen like he should, and will instead travel up the highway and ambush you near the NE energizer. (Yes, it’s interesting that Clyde’s position cues you in to whether Inky was about to go off-pattern.) While there are plenty of examples of successful perfect Pac-Man patterns on YouTube and Twitch, there aren’t really dedicated demonstrations of failed patterns, or common errors, or examples of what to look for in such situations, making this pattern recognition another thing you have to absorb through actual game play. At least in the case of the Cherry board, the slow speeds and long blue time allow even a novice like myself to collect all the ghosts freehand much of the time, but only if you recognize the broken pattern and bail on it before it’s too late.

Even when you do perform the pattern as intended, I found that some patterns defy expectations. Take, for instance, this moment from a successful 2nd Apple run:

More than once, I bailed on this pattern, because it looked like I had pooched it. Clyde’s down here, Blinky and Pinky are off behind me, and Inky’s way the heck out in the tunnel! But it turns out I had it right all along. Unlike most patterns where the ghosts’ locations cue you in to where to go after the power pellet, the chaos on this board was all according to plan. You loop around to catch Blinky and Pinky, then head back down for Clyde, then race off to catch Inky way out by the ghost pen. These are the kinds of moments that merely studying the patterns on paper cannot prepare you for.

Having grown accustomed to the tight cornering, and the increased speed starting on the Apple boards, I was hoping to make large strides with a full day on day three. And of course chat was still going:

As I said before, I was ramming that joystick around quite hard into the grooves of the internal octagonal gate, wishing to make each turn entirely certain. It was at this point I realized this emphasis on these turns was beginning to hurt my hand. Fearing the emergence of a blister, I pulled out a soft kitchen towel to drape over the joystick, which successfully softened the blow as I continued to hammer that sucker in each direction.

Day three indeed saw the gains I wanted. My expressed goal for the day was for a perfect run through each of the first eight boards, all consecutively. And I achieved that on my 399th overall attempt:

I landed at a new personal best of 127,090 following a poor u-turn on the new frontier that was the 1st Galaxian. While that may be about 3,200,000 short of a full perfect score, that was a daunting total for little old me.

This felt like a meteoric rise. Each day I made significant gains. I was outracing Blinky and beating the Pinky train almost every time now. No matter how daunting each board seemed at the start, practice really did make perfect. (You knew I was going to say that eventually.) I had done the work of committing the paper patterns to memory, and now I was committing the physical movements to memory as well. It had only been three days, and I was already almost halfway through all the “blue time” boards. Granted, I still didn’t have the greatest consistency overall, and that could stop me from getting a full perfect score through 256 boards during the current streaming window. And the long stretch of 235 consecutive 9th Key boards is a different challenge entirely. But at least, I felt, I could get a single perfect run through the first twenty boards.

Little did I know that this beginner’s success was quickly coming to a crashing end.

THE GOING GETS TOUGH

My day four of streaming was heavily interrupted by baseball again, this time with my Mariners’ grueling eighteen-inning scoreless marathon that felt like it was literally never going to end. So I didn’t feel bad when I made no headway on the 1st Galaxian that day. The problem came when I made no progress again the following day:

The 1st Galaxian was the first of the “one second” boards, where the ghosts only turn blue for a single second each time. When aiming for a perfect score, these boards represent the highest difficulty in the game. I didn’t expect this “one second” limitation to mean much to me, as I was playing using pre-arranged patterns. However, the patterns are much longer and involve many more tight turns to compensate for these limited windows. After many tries, many more than on the previous boards, I was finding myself unable to execute this sequence as flawlessly as required.

There were five particular points I had my eye on. The first of those was the SW energizer, where the ghost arrangement should look like this:

This was pretty straight-forward. If you see the ghosts out of this alignment, your pattern run is already doomed. To correct this, you just have to get better at all the turns up until that point, which with practice I was able to do with some consistency.

The much bigger concern was at the SE energizer:

After taking several tight turns through the energizer channel, you had to turn left to get the one ghost, then reverse to the right, then sharply hit down to take the fastest turn per the cornering mechanic. This was one of the rare moments in pattern play where you had to switch from sort of pretending you had input lag to, for this moment, hitting an immediate response. If you hit down too early (before you’ve finished eating Pinky), the right turn never registers, and you just keep going to the left. If you don’t hit down early enough, you take your turn late which will mess you up later. And if you try being clever and hitting both down and right, it only registers as one or the other. While not frame perfect, this turn was a very brief window to hit.

As silly as it may seem, I suspected that one factor working against me was the physical distance it took to move the joystick from holding right to holding down. If I moved the joystick to the down-and-right position, I could hear the exact point at which the microswitch clicked between the two, but simply trying to park on the spot to the right of that point, and turn it just down of that point, wasn’t reliable enough for me. My eventual solution was to remap my controls in MAME, turning the Mayflash’s action keys into alternative arrow keys. (I could have done this with the arrow keys on my keyboard as well, but those were further away from my hand in the moment when I was playing. Remember, the move from one control to another has to be made very quickly after executing the previous turns.) And so, when entering this sequence, I would switch from joystick control to big button control, giving me the precision needed to execute this sequence with some consistency.

After that came a return trip through the SW energizer path. I didn’t think much of this at first. Yes, those turns through the lower energizer paths are the tightest on the entire board, but each board runs through this channel, and I had completed the previous boards at this game speed successfully. However, it began to dawn on me that other patterns did not make a return trip through that same path, and that this was relevant to my struggles. Recall in the mechanics overview how Pac-Man slows down slightly while eating dots and power pellets. This “return” trip through that path, at full speed, actually represented the single most difficult sequence I had faced in the entire game thus far.

Coming out of that turn, Pac-Man’s proximity to Inky in front of him seemed to be a good barometer of how well I was doing, as I discussed with the chat:

The fourth point I had my eye on was this turn, where you play “chicken” with the ghosts again, this time with Pinky at the lead:

I guess you could call this Pinky Train, part two? This was the point at which the three variations of the “one second” patterns deviate, meaning I had to clear this six times on a full perfect run. I was struggling to get through this point alive, resulting in another “immediate feedback” moment.

And finally, my big make-or-break moment at the NE energizer. In a successful attempt (seen in the upper left), after I eat the first ghost for 200, the other three all turn down. In most failed attempts (as seen in the upper right), two of the ghosts would head to the left, getting out of my reach before time ran out. Also occasionally (as seen on the bottom), I would go through this sequence only to discover that Clyde was way off on his own somewhere, having snuck away from the pack when I wasn’t looking. Even using rack advance to practice just this board over and over, very few of my runs were getting this far, and fewer still were getting the correct ghost placement to take a stab at the rest of the pattern for a complete board.

Previously, as I had difficulty on one of the first eight boards, I had talked about stopping the stream for a little film study, to see if maybe I could spot where I was going wrong. Each time, more practice corrected the issue, such that I never ended up taking that study break. But this board necessitated that extra study, which I documented in a Twitter thread.

I studied fifteen runs where I at least got to the NE energizer portion of the pattern – meaning, the sample size was defined by my ability to beat out Pinky on that “immediate feedback” turn. Interestingly, literally all fifteen of these were identical at the SE energizer. While my play wasn’t always perfect up to that point (hence my inability to always get to the NE energizer), this suggested that the later Pinky turn was a check for this sequence in the SE corner. In other words, if you beat that Pinky train (which occurs at 74 seconds into the pattern), that meant you were perfect up until this prior point (which came at 57 seconds). Thus, any errors in ghost placement and such were the result of poor turns after the SE energizer – and it just so happened that the difficult return trip through the SW energizer path was within that window.

I also tried to track down the moment Clyde would occasionally veer off course, just to see if that was something I could identify quicker in future runs. I looked through incidental screenshots I had taken following my return trip through the SW energizer, only to notice that in one of them… that son of a gun was looking to the left!! Normally, he’s looking up, which is where he’s supposed to go along with Blinky. Notice also that in the run where Clyde faces up, Pac-Man appears to be a single pixel slower relative to Inky than in the other example. Recall Clyde’s mechanics, how he behaves like Blinky when he’s outside a certain radius from you, but otherwise targets the SW corner of the board when he’s within that radius. It would appear that this single pixel made the difference between Clyde falling within that radius rather than outside of it. As I said, every pixel counts!!

What was most disheartening of all was that, of my four “successful” runs I examined (the ones where I got to the NE energizer, and where the final three ghosts all turned down as intended), each appeared to be inconsistent with the others. This meant that I’d only truly had perfect execution of that pattern either one or zero times up to that point. Of course, some of these differences were minor, but as we’ve seen, those minor differences cascade into major issues as the pattern progresses – and even that far along in the board, I still had one power pellet to go.

While it would be easy to lay the blame entirely on the repeat trip through the SW energizer path, it wasn’t that simple either. Sometimes I could get the first energizer consistently, and sometimes I struggled to get it once through several tries. It was just a difficult pattern to execute all around. That’s not a knock on the pattern’s designer, of course. These “one second” boards necessitated either exceedingly long patterns, or ones with several difficult turns, or a balance between the two. Other players were able to execute these patterns to perfection. I figured, I just needed to get better.

However, in all my October attempts, I never did get a single complete execution of this pattern. Not one. Remapping the joystick buttons got me farther, but not far enough. Switching to playing the whole board using the arrow keys on my keyboard didn’t help either. With other patterns, eventually you’d get into a zone where you just know you’re executing each turn flawlessly. But with the 1st Galaxian, I would feel like I was in that zone, and yet I still wasn’t getting the intended results. After a couple more days of banging my head against this wall, it became clear to me that I wasn’t clearing this hurdle, at least not in the way I was attempting to. I decided to take the loss, for now, and change my approach.

REGROUPING

Starting with overall attempt # 801, I switched to focusing on the other remaining boards. Rack advance was still an option, in more ways than one. Even if I was unable to execute the “one second” patterns I had studied, I was hoping for a consolation prize of clearing the other 14 early boards consecutively, while bringing up the rack advance tab each time a “one second” board arrived. With my true goal having been momentarily dashed, I gave myself until my 1000th overall attempt before I would call it a session and move on to other pursuits (for now).

Of course, Twitch chat kept egging me on:

2nd Galaxian (a.k.a. “80’s night”) wasn’t too difficult. 1st Bell was a formality, given that it was identical to the other basic “Grapeland” patterns.

The high point of these later attempts, the moment which reaffirmed that I hadn’t ruined myself grinding against that 1st Galaxian board, was when I advanced for my first attempt at the 2nd Key pattern (a.k.a. “70’s night”). On overall attempt #809, I nailed that 2nd Key pattern on my very first try.

Okay, so the problem wasn’t me. Well, it was me, in that others were able to execute the “one second” pattern I could not, but it also wasn’t me, in that I could conquer every other pattern I had studied. Surely, there were other variations of patterns for those boards, perhaps which did not include a second trip through the empty energizer tunnel which was likely the primary source of my troubles.

The two other remaining patterns (one for 5th Key and another for 7th and 8th Keys) were simple, as they didn’t involve any “blue time”. As a note I found interesting, I freaked out my first couple times attempting the 5th Key pattern, since (as seen above) it appeared as though Clyde was on a collision course. But the pattern was actually designed such that Clyde enters scatter mode and reverses direction just prior to reaching your position. As with that 2nd Apple pattern, what looked like a doomed attempt was actually according to plan.

My early challenges were long gone. On the Strawberry board, I was beating Blinky at chicken virtually every attempt I made. My overall consistently wasn’t quite perfect, but it was getting closer. I was ultimately able to get the later six boards in sequence (skipping the “one second” boards with rack advance), and I completed the first eight consecutively again. However, I was not able to get all fourteen of those in one go before my 1000th attempt, which ended on the 1st Grape. While I will confess to dreaming of smashing out the first 20 boards, having gotten perfect on fourteen of the first twenty boards, even out of sequence, is still a pretty solid showing for my first eight days of attempts.

The experience of setting aside the drawing board and putting the actual game through its paces shone light on a lot of what appeared before to be incidental details. You can see how the game mechanics discussed earlier are exploited in the pre-arranged patterns, such as where encroaching ghosts are thwarted just in time by a pivot into one of the forbidden turns or the side tunnel. On the 1st Galaxian’s SW energizer (as seen above), Chapman’s pattern results in the grouped ghosts entering scatter mode just before you eat the energizer. Normally, the energizer would reverse them, sending them away, but since scatter mode had already just reversed their direction, this timely double-reversal results in the ghosts heading right in your direction. All of that is baked into the pattern as it was designed. Understanding these mechanics in action allows you to see so much more than what was apparent on the page.

As for my quest for a perfect score, now that I’ve decided the “one second” pattern I was using is not within my grasp, I’m left to decide how to proceed from here. I’ll probably study another pattern, perhaps even one with the occasional pause or reversal, and see if I find it more to my liking. However, freehand grouping would seem to be the obvious next step in my development. While I’ve watched freehand techniques, I made no effort to learn them, given that doing so would require live play and I wanted this first session to be done fresh, relying only on pattern memorization. Indeed, freehand play is a good skill to have anyway, which can add the option of recovering from a failed pattern on a blue time board, or simply evading the ghosts on the 9th Key stretch.

As a casual observer, freehand play appears much more daunting, more scary if you will. I was and still am willing to grant that freehand techniques have a higher barrier to entry. It certainly takes much more active game practice to learn them than it does patterns, which I was able to commit to memory during my down time riding the bus around town. However, with preparation and knowledge being equal, I began to ponder the question of whether freehand play really was the “more difficult” method of Pac-Man play, as I had always heard it described. On its face, freehand grouping seems much more forgiving, not requiring each turn to register at the earliest applicable moment. Is it possible I had been lied to all these years? Could it be that a well-versed Pac-Man master would rely on casual freehanding when they simply wished to get a perfect score without a time limitation? Would such a master only up their game and concentrate their efforts on more difficult pattern play if they wished for a speedier time? Did I have everything backwards all along?

Unfortunately, because of various life things, my next session won’t be soon. This is but one of many balls to juggle in the limited journey that is life. But I do intend to spend time both retaining what I’ve learned and studying more advanced techniques, including on those various “Reunion” cabinets and such staring at me so longingly around town. The hope is, the next time I have a great window for streaming, I’ll be ready to hit the ground running once again.

To be continued…