Superauthenticity: Arcade game aspect ratios

Photo by redditor AKJVermont

Early-to-golden-age arcade games, with bespoke video hardware and no real standards, pose some unique properties concerning aspect ratio and resolution. Among them are:

• Vertically-oriented monitors
• Vector monitors with point-plot resolution
• Non-standard monitor configurations
• Variable resolutions

Most of these games, even back then, were designed for a 15Khz 4:3 monitor. Even vertically-oriented arcade games would simply output a 90-degree rotated display and expect the monitor to be rotated likewise; MAME thankfully automates this for you by default. For the most part, you can achieve a reasonably authentic aspect ratio in MAME by simply letting it do its thing.

However, my superauthenticity theory speculates that some games simply didn't account for the distorting effect that non-square pixels would have on the final display, and if these games could be identified, then authentic aspect ratio is a flaw, and by using square pixels, instead of a corrected 4:3 aspect ratio, their presentation could be enhanced in a non-destructive manner.

I'm not even going to examine pixel aspect ratio here; it's too complicated to delve into that for such a wide variety of video display types, and arcade games generally had more precise video signals than console games of the day did, with much less overscan. I've already explained why it's bad to assume 4:3 DAR is authentic, but if I'm going to analyze arcade games, it's too much trouble not to.

 

Atari 

Breakout

Data Driven Gamer's first ever whale defies resolution analysis. MAME considers the resolution to be 896x252, but a raw screenshot captures at 228x1440. In reality, the video hardware was analog and any resolution capture is just an approximation.

Just let MAME deal with it. It's fine.

 

Centipede

Scaling:	1x 2x
DAR:	Raw 3:4

At 240x256 resolution, raw pixels give Centipede a nearly square screen layout. A vertical layout with authentically thin pixels just looks better. Right down to the font.

Verdict: 3:4

 

Star Wars

DAR:	Raw 4:3

Internally, Star Wars has a 'tall' resolution of 502x562, but this is a horizontal game. But since it plays on a vector monitor, resolution doesn't work in a typical way; 502x562 is effectively a grid of positions from which lines segments may begin or end. The line segments themselves are not bound to this grid (no staircases! no aliasing!), and because of that, the game's effective resolution is limitless.

Incidentally, vector games are a situation where I feel pixel shaders are appropriate for screenshot documentation purposes. I never use them for screenshotting raster games, as they are destructive to the raw image data, but vector games don't have raw image data, at least not in any sense that would produce a pleasing visual.

As for the best aspect ratio, MAME does provide an option to play at a tall aspect ratio, and in some ways, this actually looks better than 4:3. Fonts and 2D screen elements seem more natural, and the iconic Star Wars text crawl looks better this way. 

DAR:	Raw 4:3

But the Death Star is round in 4:3, and narrow in tallscreen, so it's clear that the artists gave some thought to the 4:3 aspect ratio. They just weren't consistent about it.

Verdict: 4:3

 

Marble Madness

Marble Madness runs slightly pillarboxed but has an output resolution of 336x240, which is extremely close to 4:3. I expect this was deliberate.

 

Paperboy

Running on a high-for-its-time-resolution 25Khz monitor at 512x384, Paperboy was designed for square pixels, and there's no other option.

 

Gauntlet

Another 336x240 resolution Atari game.

 

I think we can conclude from this that arcade game manufacturers like Atari considered display aspect ratio pretty much from the beginning. Star Wars was the only time that there was any inconsistency - I guess vector math is hard enough without also having to worry about aspect ratio. But let's look at some more, by other companies.

 

Namco 

Galaxian

DAR:	Raw 3:4 7:8

Galaxian is an odd case. The playfield, which is rendered by tile-mapping hardware, is 224x256, a 7:8 ratio. But the starfield behind it is generated independently of Galaxian's tile mapper.

Stars are not exactly pixels, but analog pulses of color on each scanline (which run vertically, not horizontally). Since each star is generated on a single scanline, a star's width is the same as a playfield pixel, but its height is smaller than one, and its vertical position can be anywhere.

MAME has to work with pixels, though, and simulates this by rendering 224 scanlines, but treats each one as being 768 pixels tall, and renders the 224x256 playfield with triple-tall pixels to fit. The simulated starfield is then 224x768, which provides a reasonable approximation of Galaxian's sub-pixel stars.

So in this case, the "raw" 224x768 resolution is definitely wrong and isn't intended to be right. Bringing the display in alignment with the playfield's 7:8 aspect ratio simply requires tripling the pixels horizontally. But an authentic 3:4 aspect ratio just looks more correct, I think.

Verdict: 3:4

Pac-Man

Scaling:	1x 2x
DAR:	Raw 3:4

At 224x288 resolution, Pac-Man is pretty close to a natural 3:4 aspect ratio, but different enough that some stretching occurs. But not so different, it seems, that Iwatani felt it necessary to account for this. Square pixels make Pac-Man perfectly round, the dots perfectly square, and the maze perfectly geometric.

Verdict: Square pixels

 

 

Dig Dug

Scaling:	1x 2x
DAR:	Raw 3:4

Another 224x288 game. Similarly, I think this looks just a bit more correct without the aspect ratio correction.

Verdict: Square pixels

Xevious

Scaling:	1x 2x
DAR:	Raw 3:4

Same resolution. Same thoughts.

Verdict: Square pixels

Pole Position

Scaling:	1x 2x
DAR:	Raw 4:3

Pole Position runs at 256x224 with fat pixels... but it just looks unnaturally stretched out that way.

Verdict: Square pixels

 

Overall, Namco seems less concerned with precise display aspect ratio than Atari did, but again, it's not consistent. Galaxian, one of their earliest games, looks better at an authentic 3:4 aspect ratio, and Pole Position, one of their most advanced of the era, looks better with square pixels, but the majority of their games ran at 288x224 where it doesn't make that much of a difference. I suspect that was a deliberate choice.

 

Midway 

Wizard of Wor

Scaling:	1x 2x
DAR:	Raw 4:3

Wizard of Wor runs at 352x240 and does a similar thing as Galaxian where the starfield signal is independent of the playfield pixels, but MAME doesn't use any resolution tricks to simulate this.

An authentic 4:3 requires thin pixels, and I think this looks better. 

Verdict: 4:3

 

Tapper

DAR:	Raw 4:3

Let's look the logo that Budweiser used in the early 80's:

With fat pixels, the bartender is a bit pudgy, but that banner is a pretty close match!

Verdict: 4:3

 

Spy Hunter

DAR:	Raw 3:4

Spy Hunter's got an interesting resolution of 480x480, though this is one of those situations where the sprites are drawn at a higher resolution than the background layer and MAME just uses that as the overall canvas. Long story short, Spy Hunter only looks right with thin pixels.

Verdict: 3:4

Midway, it seems, was not afraid of experimenting with strange (and often high) resolutions, and understood how to use non-square pixels.

Konami

Super Cobra

DAR:	Raw 3:4 7:8

Super Cobra and Scramble run on the Galaxian board and have the same resolution oddities due to the starfield generator.

As with Galaxian, 3:4 looks better than triple-wide pixels. 

Verdict: 3:4

 

Frogger

DAR:	Raw 3:4 7:8

We're still on Galaxian hardware, but this time the starfield generator doesn't do anything, and we're just looking at a 224x256 playfield where every pixel is tripled in height for no reason.

An authentic 3:4 once again looks the most natural.

Verdict: 3:4

 

Gyruss

Scaling:	1x 2x
DAR:	Raw 3:4

We're off Galaxian hardware, finally, and running at 224x256. This title screen, with the round earth, is the strongest evidence that Konami was taking non-square pixels into account.

Verdict: 3:4

Rush'n Attack

Scaling:	1x 2x
DAR:	Raw 4:3

So much for consistency! Rush'n Attack looks terrible at 4:3.

Verdict: Square pixels

 

Gradius

Scaling:	1x 2x
DAR:	Raw 4:3

Gradius has a bit more horizontal resolution than Rush'n Attack, and because of it, the 4:3 stretch isn't as pronounced, but I think this still looks better without it.

Verdict: Square pixels 

 

Nintendo

Donkey Kong

Scaling:	1x 2x
DAR:	Raw 3:4

We've looked at this one before when analyzing NES games' aspect ratios, and concluded that more often than not, Nintendo didn't really account for pixel aspect ratio, but sometimes they did. Donkey Kong on NES was an example where they did not; the arcade game has thin pixels and a vertical orientation, the NES has fat pixels and a horizontal orientation, and while the NES version's layout was reworked to account for this, the sprites and tiles were not.

But with the arcade original, I think they did account for the thin pixels. The rolling barrel looks rounder that way, and DK himself looks more proportionate, as do Mario and Pauline.

Verdict: 3:4

 

Donkey Kong Junior

Scaling:	1x 2x
DAR:	Raw 3:4

DKJr likewise looks better with thin pixels.

Verdict: 3:4

 

Popeye

DAR:	Raw 4:3

Popeye is another one of those weird cases where the sprites are drawn with much smaller pixels than the playfield pixels are. MAME gives it a canvas of 512x448.

Honestly, I think this looks better without aspect ratio correction. 

Verdict: Square pixels

 

Mario Bros.

Scaling:	1x 2x
DAR:	Raw 4:3

I noted before that both the arcade and NES versions of Mario Bros. seem better to me with square pixels. I still feel that way.

Verdict: Square pixels

 

Williams

Defender

Scaling:	1x 2x
DAR:	Raw 4:3

Defender runs at 292x240 resolution, which is pretty close to 4:3 but does stretch a bit to fit. The graphics are a bit abstract, but I don't really feel that the stretch is an improvement.

Verdict: Square pixels

 

Robotron: 2084

Scaling:	1x 2x
DAR:	Raw 4:3

Robotron has the same resolution as Defender, and this time I feel pretty certain the graphics are better left unstretched.

Verdict: Square pixels

Joust

Scaling:	1x 2x
DAR:	Raw 4:3

Joust also runs at 292x240, and looks a bit better without aspect correction.

Verdict: Square pixels

Sega

Zaxxon

Scaling:	1x 2x
DAR:	Raw 3:4

Sega's Zaxxon is, unambiguously, intended for thin pixels. It features a true isometric projection, but this needs the 3:4 aspect to be seen correctly.

Verdict: 3:4

 

Buck Rogers: Planet of Zoom

DAR:	Raw Double scan 4:3

Buck Rogers involves some analog video processing and MAME simulates this by setting the canvas at 224 scanlines of 512 pixels each. A double scan results in a resolution of 512x448. To me, this looks better than an authentic 4:3, which looks a bit too stretched horizontally. Either is better than raw output, obviously.

Verdict: Double scan

Space Harrier

Scaling:	1x 2x
DAR:	Raw 4:3

Space Harrier runs at 320x224, which is the same resolution that a lot of Genesis games would eventually run at, and has thin pixels on a 4:3 monitor. It's slight, but it's better this way.

Verdict: 4:3

Capcom

1942

Scaling:	1x 2x
DAR:	Raw 3:4

This looks a touch better at the thinner 3:4, I think.

Verdict: 3:4 

 

Commando

Scaling:	1x 2x
DAR:	Raw 3:4

Commando likewise looks better at 3:4.

Verdict: 3:4

Ghosts 'n Goblins

Scaling:	1x 2x
DAR:	Raw 4:3

But this one, which I did not and will not cover, looks better raw! 4:3 Arthur just seems a bit out of shape for this adventure. 

Verdict: Square pixels

 

Before this dive, I had a theory that western developers were more inclined to take pixel aspect ratio into account. Based on this sample, I can't say my theory is substantiated. While there does seem to be some correlation between region and aspect ratio awareness, it's not a strong one. Every Japanese developer here had at least some games that looked better with an authentic aspect ratio, and all of Williams' games look better with square pixels.

But there is another divide which seems much stronger - vertical vs. horizontal.

Almost every vertically-oriented game I compared looked better with a correct 3:4 aspect ratio. The sole exceptions were three by Namco, which ran very close to 3:4 anyway.

The horizontal games, on the other hand, for the most part looked better with square pixels. All of the exceptions had some unusual pixel properties; Midway's games run at strange resolutions with pixels that are either thin or very fat. Space Harrier uses thin pixels. Star Wars doesn't use pixels at all. The rest of the horizontal games here all have pixels with a PAR between 1.09 and 1.25, and I find they all look nicer with a PAR of 1.

Maybe I just have a bias towards thin pixels? I think it will be interesting to revisit this topic in the future, as vertically-oriented arcade games start to become less common, outside of niche genres.

Game 464: Gun.Smoke

Get Capcom Arcade Stadium 2, which offers Gan Sumoku and 30+ other Capcom arcade games as DLC:

https://store.steampowered.com/app/1755910/Capcom_Arcade_2nd_Stadium/

 

Ok, guys. Just because you can make your shmup with unlimited continues unreasonably punishing doesn't mean you should. I know this won't be the last arcade game to give up all pretense of fair play and be a player burner intended to be beaten with perseverance and deep pockets rather than skill, but at least later games commit to that by letting you continue immediately on the spot and power through that ridiculous boss fight, rather than make you repeat it from the start each time, sans power-ups, as if you have a chance.

Gun.Smoke combines elements of Capcom's prior shmups 1942 and Commando, though it owes more to the former despite the on-foot setting of revolvers and cowboy boots; it's a continually forward-scroller where your P-38 Lightning is replaced by a cowboy who can't stop walking northward, and while you can shoot in six different directions based on whichever combination of three action buttons you press, all of them are some degree of "in front of you."

Like Commando, enemies aren't bound to 1942-like attack formations and can behave unpredictably - you are locked into your steadfast march forward, but enemies can and will slip past you, turn around, and shoot you in the back. Or just pop in from the lower-end of the screen side edges and shoot you in the back. It's a worst-of-both-worlds situation for your own survival, and when you also have Commando-level enemy aggression, bullet spam, and obstructive terrain with endless area denial attacks from dynamite-tossing Indians, you're going to be using those unlimited continues. Lots of them.

And that's before you have to fight the bosses, artfully introduced at the start of the game as a flyby of wanted posters.

I had beaten Commando, which doesn't allow continues, with the use of a single save state right in the middle of the game. I needed two save states to reach the midpoint of Gun.Smoke and I still couldn't beat the boss there, a firebreathing carnival freak named Pig Joe. And I don't think I want to.

As with 1942, I've made a video of the attempt, but edited out several minutes of repeated failures where I made no progress. These are mostly against the bosses.

 

Level 1 starts off gently enough, with no more than three or four gunslingers, snipers, or riflemen on screen at once as you take back the besieged town. At least not until you reach the gang leader, Master.

Destructible barrels grant powerups - boots enhance your speed, bullets enhance your firepower, and rifles extend your range. Trying to take time to shoot them while dealing with bad guys can be tricky, but you'll want at least one of each before reaching the boss, or indeed any other boss. You might also find a horse, but holding onto it before reaching the stage boss is always a challenge, and if you die fighting a boss, you're not getting another one.

Master is the only boss I've ever reached with the horse, and he's not too bad if you're fully powered up. Just focus on the underlings when he's prone, hang back far enough to give you room to dodge his bullets, and gun him down when he stands up.

Is that a cowboy or a hopping vampire?

 

Level 2 suggests a train robbery.

Stay away from the edges. It seems safe, but baddies love to spawn there.

The boss, Roy, throws knives faster than his men can shoot bullets. And here, the real nightmare begins - bosses are downright cheap. Starting with Roy, you face overwhelming enemy spawns during boss fights, but on top of that you have to deal with the boss himself who takes multiple hits and has absurd invulnerability frames as he hops around the screen, pummeling your personal space with barely telegraphed projectiles that you couldn't possibly react to.

 

Your only way to survive is to never, ever stop moving, but also make sure you don't move into any of the bullets that rain on you from all directions and tend to get lost in the background. Sometimes you can shoot his throwing knives down, but counting on it will get you killed.

It gets worse. It gets much worse.

Narrow banks and bridges around this river hideout mean little room to dance with these bandits. Just getting to the end feels like a matter of lucky spawn patterns - if too many appear in places you can't hit, you get swarmed and die. If one appears on the same edge of the river as you, you get hit and die.

And the ninja boss at the end is a real bastard too.

He's like Roy but worse. He spends most of his time in the air, invulnerable, and even when he lands, he's still invulnerable for a few frames. You can't take your sights off him or you'll miss your window to land damage. Except you have to, because the screen is being flooded with bullets from his minions. But focus on them and the ninja will hit you with a shuriken. It's insane.

I had to use a save point at the start of this fight to get through. Even reaching it from the checkpoint with a powerup set (boots, bullets, and rifle) felt like a matter of luck.

Stage 4 in comparison isn't so bad - it's a wide open valley with a fight against a boomerang-throwing bushwhacker. Things quickly get horrible again in the next one, a burned-out town, which is where I finally gave up.

Sigh. Why's your range got to be so short? Forget hitting the broad side of a barn; these bullets couldn't pass the broad side of a barn, and I've got to move up the screen to get close to Pig Joe, which is usually fatal. And why do my bullets just go through him nine of of ten times? Is this some i-frame nonsense, or just run-of-the-mill sponginess?

I had another save point here, but after a good hour and a half of dying and reloading, I tried another tactic - leave the window snipers alone so that more mobile enemies don't spawn in their place. It worked, but it just meant I had to keep myself on the far right edge of the screen with no way to dodge Pig Joe's fire breath.

 GAB rating:

I hated this game. 1942 was more or less balanced but overlong and repetitive, but Gun.Smoke just feels cheap and predatory when it isn't also overlong and repetitive. Even on a cheatless "tutorial" playthrough, put together by a player who obviously knows the game inside out and can tell you exactly where to go, what barrels to shoot, and the tricks you need to use to beat each boss, there are still many cheap deaths and edited-out game overs - so many that he claims the Wolf Chief fight on stage 6 took almost nine hours of failures before having a minute-long success, and I'm not sure that's an exaggeration. It's worse than Ghosts 'n Goblins, which I have similar feelings about (and didn't/won't cover).

Unlike Ghosts 'n Goblins, the NES port of this one makes significant gameplay changes, such as adding the ability to buy powerups from townsfolk. By multiple accounts, it's much less punishing, and therefore more fun, than the arcade original. But I'm not interested in playing it myself.