the unremade
Entry two of three in “the dimension problem.” Part one builds the vocabulary this post spends; part three tests it to destruction on a single case.
The remake business is the cleanest natural experiment this medium has ever run. Same design, same audience, twenty or thirty years apart, public unit counts at the end. Everybody is running the experiment. Almost nobody reads the results.
Part one gave us the instrument. Two variables, not one: presentation dimension (how it’s drawn) and ludic dimension (how many axes the player reasons in). A core verb that is either spatial or compositional. A scale floor, the minimum budget an entry in a category costs once you’ve paid for the axis.
Run the record through it.
the four results
Held the category, upgraded the presentation. This is the whole game.
Super Mario RPG remake: 3.31 million against the SNES original’s 2.14 million lifetime. Paper Mario: The Thousand-Year Door: 1.94 million inside a year, past the GameCube original’s entire run. Link’s Awakening on Switch: 6.46 million. Dragon Quest III HD-2D plus I & II: past 4 million combined, and Dragon Quest III was Japan’s only million-selling physical game of its release year.
Every one of them is the same move. Keep the ludic dimension exactly where it was. Spend the whole budget on the presentation dimension. Ship for a fraction of AAA and beat the original.
Held the category in 3D. The axis is not the enemy. Leaving your category is.
Resident Evil 2 remake: 19.75 million as of June 2026, the best-selling game in a franchise that has been running since 1996 and Capcom’s second-biggest title ever. The PlayStation original did around five. Resident Evil 4 remake: 13.6 million on Capcom’s official books this spring. Both of them were 3D before and 3D after. Both had spatial core verbs — survive a space you can’t fully see — and the remake gave the verb more of what it already wanted.
Crossed the axis on a spatial verb. Survivable, and the industry’s favorite gamble.
Super Mario 64, Ocarina, Metroid Prime, GTA III. All of them work, all for the reason part one gives: the verb was already spatial, so the axis fed it.
But note what happens when the verb is nominally spatial and actually isn’t: Final Fantasy VII Remake turned a menu-driven, arrangement-reading combat system into a real-time action game, which is a ludic category change wearing a graphics upgrade’s clothes. That case is sharp enough to carry a post of its own — part three spends itself on it.
Crossed the axis on a compositional verb. The reliable way to kill a franchise.
Golden Axe: Beast Rider took a game whose verb was control a lane of enemies, dissolved the lane into a volume, and handed the player a camera to fight alongside the orcs. Bomberman: Act Zero took reason about a grid and put the grid inside a corridor. Syndicate in 2012 took isometric squad tactics and shipped a first-person shooter. In every case the publisher believed it was modernizing a presentation. It was changing genres and keeping the box art.
the fifth failure
There’s one the dimension model doesn’t cover on its own, and it’s the ugliest.
Silent Hill HD Collection increased the resolution and deleted the fog. The fog was a PlayStation draw-distance limitation that the developers turned into the entire aesthetic — the thing the game was actually selling was not being able to see. More pixels, less product. The re-release also re-recorded the voice acting, replacing an awkwardness that had become the texture of the thing.
Warcraft III: Reforged shipped a visual overhaul that fans read as a downgrade, dropped most of what was promised, and — the unforgivable part — overwrote the original so nobody could go back. GTA: The Trilogy – Definitive Edition ran twenty-year-old art through modern lighting and made it look like a fever.
Call this presentation regression: the remake upgrades the rendering and destroys a medium-specific effect the original was built on. It’s Lessing’s problem exactly. Fog, sprite readability, grid legibility, draw distance, the particular flatness of a hand-drawn backdrop — these are not limitations the original suffered under. They are the material the work was made of. Remove them and you have restored something nobody was looking at.
the rubric
Five questions. Answer them before anybody opens an engine.
- What is the core verb? Spatial or compositional. One word if you can.
- Does this hold the ludic dimension? If it crosses, is the verb spatial enough to survive the crossing?
- What medium-specific effect is the original actually selling? Name it. Then check whether the rendering upgrade deletes it.
- Where is the original’s audience now, in their earning life?
- Does the target budget clear the scale floor for the category you’re actually shipping in — not the category the box art implies?
Question five is where studios die. You can build an HD-2D remake of a 1994 RPG for money a two-million-seller can repay. You cannot build a $200 million one, and if you try, the correct decision at every subsequent milestone will be to broaden the audience, which means leaving the category, which means question two now fails.
the demographic case
Here’s the part the spreadsheets get wrong.
Somebody who bought a game in 1995 at fourteen is forty-five now. The cohort that bought the 16-bit and early-32-bit libraries is somewhere between thirty-four and fifty-six, which is the peak-earning decade, with the highest nostalgia elasticity of any cohort alive and a documented willingness to pay full price twice for the same design.
Which makes the original’s unit count a floor, not a ceiling. Discount for mortality and taste drift, then multiply by an install base an order of magnitude larger than the original platform’s, plus the people who weren’t born yet. Super Mario RPG and Thousand-Year Door both cleared their originals — that’s the pattern, not the exception, as long as the category holds.
And when there are no unit counts — most of the interesting candidates are PC games from an era that didn’t publish numbers — use the better metric.
A volunteer reimplementation that has survived a decade is a demand signal with a cost attached. Surveys are cheap. Somebody spending eleven years of unpaid evenings rebuilding an engine is not. OpenXcom. The Ur-Quan Masters. OpenRA. OpenMW. DevilutionX. CorsixTH. FreeSpace Open. KeeperFX. OpenRCT2. ScummVM holding up an entire genre by itself. Every one of those is a market signal that the rights holder is refusing to read.
the unremade
Candidates, with the category call attached. Sorted roughly by how obviously correct they are.
Chrono Trigger. Five million lifetime across every re-release, and the format that should have shipped it already exists inside the same publisher. Compositional verb, 2D ludic dimension, HD-2D is sitting right there — Square Enix proved the pipeline on Dragon Quest twice. This is the most obviously correct remake in the industry and it is not being made. The trap is reverence: the team will want to add an axis to justify the budget, and the axis will eat the game.
Star Control II. Conversation and a physics duel — one compositional verb, one spatial, cleanly separated, which is why it still works. The Ur-Quan Masters has kept it alive for two decades. The trap is entirely legal, not creative.
Sid Meier’s Alpha Centauri. The writing is the product. Compositional, flat, tiny floor. Nobody has matched it in twenty-five years of trying, including the people who own the adjacent franchise.
Thief II: The Metal Age. Spatial verb — occlusion is the mechanic — and still unmatched in its category. Crossing dimensions is not the risk here. Presentation regression is: modern lighting will delete the darkness the same way HD deleted the fog.
Herzog Zwei. The origin of the real-time strategy genre, 1989, and a scope small enough for four people. Compositional. Nothing about it needs a volume.
Crusader: No Remorse. Isometric, destructible, mean. Compositional verb with a spatial skin — the exact profile that dies if somebody “modernizes” it into third person, which is exactly what will be proposed.
Skies of Arcadia. The most-requested remake in the Dreamcast catalog for twenty years. Already 3D, spatial verb, holds its category cleanly. Sega has no excuse that survives the Suikoden I & II HD Remaster posting sales Konami called “extremely strong.”
Master of Orion II. Compositional, turn-based, near-zero floor. The market has repeatedly demonstrated it will buy this and repeatedly been sold something with more axes and fewer decisions.
Terranigma / Illusion of Gaia. Enix’s other 16-bit line, never fully re-released in the West, sitting in the same vault the HD-2D pipeline already has keys to.
Wave Race 64 and 1080° Snowboarding. Purely spatial verbs — the water and the carve are the game. These are the rare candidates where crossing further into the axis is correct, and Nintendo has left them on a shelf for twenty-five years.
the part that matters: rebuild the category, not the IP
You will never get the rights to Syndicate. Nobody at EA is going to return that email.
But nobody owns “isometric squad tactics where you control four cyborgs with persistent equipment and no undo.” The IP is locked. The category is public domain — and part one’s whole argument is that the category, not the IP, is what the audience was buying. The franchises are the packaging. The verb is the product.
So the homebrew move isn’t a clone with the serial numbers filed off. It’s a new game, with new art and new fiction, built native to a category that got evicted by the scale floor and has never been un-evicted. That is a legitimately open market with proven demand and zero legal exposure, and it is reachable by one person with an agent fleet in a way it simply was not five years ago.
Five briefs. Each one specifies the verb, pins the ludic dimension, names the medium-specific effect to protect, and sets the floor. Hand them to whatever you’re running.
ultracode Build a vertical slice of an isometric real-time squad tactics game.
Core verb: COMPOSITIONAL — read an arrangement of enemies and terrain, commit,
live with it. Ludic dimension: strictly 2D reasoning on an isometric grid; the
camera never rotates and never occludes. Four agents, persistent equipment and
injuries that carry between missions, no undo, no save scumming inside a mission.
Medium-specific effect to protect: total legibility of the arrangement — the
player can always see the entire problem. Scope floor: one 20-minute mission, a
loadout screen, three enemy archetypes. Original art and fiction only — this is a
category rebuild, not a clone. Deliver: playable build, a design doc naming every
rule, and a written argument for why each rule serves the core verb.
ultracode Build a turn-based 4X strategy prototype in the pre-3D tradition.
Core verb: COMPOSITIONAL — the player reasons about numbers, adjacency, and a
tech graph. Ludic dimension: 2D map, no camera control beyond pan and zoom.
Priorities in order: (1) tech tree legibility, (2) a ship/unit designer where
every component choice is visible in combat, (3) an AI that plays the same rules
the player does and can explain its last three decisions. Medium-specific effect
to protect: a player can hold the entire strategic state in their head. Explicitly
out of scope: 3D planets, cinematics, voice acting, real-time anything. Deliver:
playable prototype, plus a written post-mortem on which systems earned their
complexity and which should be cut.
ultracode Build a single-screen arena puzzle-action game in the grid tradition.
Core verb: COMPOSITIONAL — reason about a grid under time pressure. Ludic
dimension: 2D, one screen, no scrolling, no camera. The entire game state is
visible at all times and this constraint is non-negotiable — reject any feature
that violates it. Presentation dimension: spend freely here. Modern lighting,
particles, and animation over a flat grid; treat HD-2D as the reference. Include
four-player local multiplayer from the first commit, not as a later addition.
Deliver: playable build, and a one-page note on every feature you rejected for
violating the visibility constraint.
ultracode Build a first-person stealth prototype where darkness is the mechanic,
not a filter. Core verb: SPATIAL — occlusion, light, sound, and patience. Ludic
dimension: full 3D; the axis is correct here and should be fed. Hard constraint:
resist presentation regression. Modern rendering must not raise the floor-level
visibility of dark spaces; if a lighting upgrade makes the shadows readable, the
upgrade is wrong and gets reverted. Build a light-and-sound propagation model
first and a level second. Guards must be deceivable by a player who understands
the model, and never by a player who doesn't. Deliver: one building, three
approaches, and a technical note on how darkness was kept expensive.
ultracode Build a 2D exploration-and-combat RPG in the 16-bit action tradition,
targeting the HD-2D presentation tier. Core verb: COMPOSITIONAL — read a screen,
solve a screen. Ludic dimension: 2D throughout; no third axis anywhere, including
in the menus. Budget model: assume a total production budget a 500,000-unit seller
could repay, and record every scope decision against that constraint in a running
log. Deliver: two connected regions, one boss, a save system, and a
BUDGET-DECISIONS.md that shows what was cut and what the scale floor would have
been if any part of this had been rendered in a volume.
Each brief encodes the theory. Verb first. Dimension pinned. The medium-specific effect written down as a constraint the build is not allowed to violate. And a budget floor stated up front, because the floor is what quietly changes every other answer once somebody starts arguing for one more axis.
Thirty years of this industry’s decisions came down to reading a bigger number as a better product. The remake business hands us the counterfactual for free: the same design, twice, with receipts. The receipts say the same thing every time. Keep the category. Spend on the surface. Let the floor stay low.
The games worth rebuilding are not the famous ones. They’re the ones that got evicted from a platform by an accounting threshold and never came back — and the rent on that category is now approximately zero.