Inside the Category: A Designer's Study of Shooter Subgenres

I have sat in the meeting where somebody asks what genre the game is. It is never a design question. It is a question about which shelf the store puts you on, which competitor set the publisher benchmarks you against, and which two words go at the top of the deck so that a person who will never play your game can nod. I have watched a good vertical slice die because the answer to that question was three words long instead of one.

So when I read a taxonomy of shooters, the neat kind, shoot 'em ups over here, first-person over there with arena and hero and tactical and looter hanging off it, I do not read a family tree. I read a procurement document. Somebody assembled it after the fact from games that were already sitting on shelves. Nobody in a room like the ones I have worked in ever said "let us make an extraction shooter" before there were extraction shooters, because the phrase did not exist and the pitch would have been meaningless.

What I want to write down here is what these categories look like from the floor, where you are making the decisions that later get called genre.

---

The founding decision was a production decision

Everyone tells the id Software story as though the first-person shooter was a creative vision. It reads to me like a scheduling problem.

Romero wrote something like seventy-four games before he was twenty-two. That is not a portfolio, it is a symptom. Softdisk needed a game a month for a magazine business, and one person cannot do that forever. Carmack's actual breakthrough was smooth scrolling on PC hardware, and the demo he built to prove it was a port of Super Mario Bros. 3. Nintendo passed. The engine became Commander Keen. Two men who bonded over Miyamoto ended up defining the genre that would spend thirty years being blamed for violence.

Then they did the thing that mattered, which was not graphical at all. They separated level data from the program so the data could be swapped without a rebuild. I have worked on codebases where this was not true, and I can tell you what it costs: every content change is an engineering change, every artist is blocked behind a programmer, and your throughput becomes the size of your engine team. Fixing that is why id could ship. Selling the fix to other studios is why the first-person shooter propagated instead of remaining a curiosity that one company in Texas happened to be good at.

That is the whole mechanism. The FPS is universal because the code was for sale. Not because it was the best idea in the room.

Tom Hall wanted story in Doom and was pushed out for it. Carmack's line about the story in a game being like the story in a porn movie gets quoted as a philosophy. It is not a philosophy, it is a budget. Story does not license. An engine licensee buys your renderer, not your narrative designer's outline. When your revenue comes from selling the tech, every hour spent on the thing that does not transfer is an hour you are eating yourself. I have made the same call with worse justification.

And then Quake, which is the part that should be taught in production courses rather than design ones. True 3D meant modelling instead of spriting, and modelling did not scale the way sprites did. They brought in Michael Abrash to make the renderer survive. They finished at four in the morning on 22 June 1996. Over half the staff left afterwards, Romero included. The arena shooter, as a category, is essentially what was left lying around after that death march: one game's multiplayer, canonised later by people who loved it and formalised by people who were not there.

The founding trauma of the genre is a labour problem. Every subgenre that follows inherits it.

---

Mods are free research and development, and one of them ate the market

Counter-Strike was two people modifying Half-Life. Valve hired them, took it standalone, bought the IP. That is the second mechanism, and I have never seen a cheaper one. Someone else does years of unpaid iteration in public, against real players, and you buy the result once it has already proven retention.

The numbers are still faintly ridiculous. Global Offensive cleared 1.5 million concurrent players around the Counter-Strike 2 announcement, beating its own pandemic-era record. One evening snapshot put it at 648,000 players, roughly fourteen percent of everyone on Steam at that moment. Reported revenue past four hundred million in 2018, with case sales estimated in the tens of millions a month. A remake of a remake of a mod, eleven years old, outdrawing every game built specifically to kill it.

The explanation people give is that the series is focused on gameplay and nothing else. I believe it, and I want to be clear about how brutal that is as a practice. It means saying no to almost everything. It means letting the menus rot, letting the presentation rot, refusing every feature that would look good in a trailer, and defending a twenty-year-old economy against every designer who arrives with a better one. Most studios cannot do this because most studios have to justify headcount. Focus is not a design virtue, it is an organisational one.

The other thing Counter-Strike gave the industry is the part nobody credits properly. Source, released as an engine showcase and a Steam flagship, introduced skins. Every free-to-play shooter I have worked on or pitched against is standing on that.

---

The map that shows you what a genre does to a team

If you want to see a subgenre constrain the people inside it, look at Office. I keep coming back to it because it is the closest thing we have to a controlled experiment.

Office shipped in 1999, was remade for Condition Zero in 2003, remade again for Source in 2004. Two and a half real versions in five years. Then it froze. The 2004 layout survived the Global Offensive refresh and arrived in Counter-Strike 2 on a new engine, twenty years on, with folders and mugs and monitors that in places render at lower resolution than the Source assets did.

3kliksphilip's breakdown of the Source 2 version is the best free lesson in what a generational leap actually is. The conference room keeps its chairs, its table, its ceiling lights. What changed is lighting and materials, and the room reads as modern. The projector is the clean case. In Source it lights the player with a generic bright light and casts no shadow. In Global Offensive it projects the image onto the player model correctly and still casts no shadow. In Counter-Strike 2 it does all three, including self-shadowing on items dangling from the body. Twenty-year-old props. Several generations of apparent improvement. Zero additional polygons.

Now the part that interests me more.

Monitors used to split apart when shot and now take decals. Keyboards used to shed keys. Folders used to come apart over several hits and now do not move. Cabinets and chairs are nailed down. The radio loses its antenna. Hostages do not bleed, do not crumple, do not die.

Read that list as a bug tracker and it stops looking like laziness. Every one of those is defensible in isolation. Loose cabinets in a tactical shooter mean players blocking corridors, and Source physics were never stable enough to trust with that. Destructible props are collision churn and network traffic. Hostage gore is a ratings conversation and an animation budget. Each ticket has a reasonable owner, a reasonable argument, and a reasonable outcome. Interactivity does not get killed. It gets closed as Won't Fix, forty times, by forty people who were each individually correct.

Meanwhile the same version added three-dimensional holes in snow whose shape depends on the entry angle, and fabric decals that suggest torn foam underneath. That is not a contradiction. That is what happens when a feature can be delivered by one person in materials without touching simulation, netcode, or the competitive integrity review. Cheap wins land. Expensive wins die. The map gets prettier and less alive on the same sprint.

The one thing in that analysis that would actually keep me up is the shadows. The new indoor lighting can telegraph a player around a corner, and it does not render on low settings. That is not a fidelity gap, that is an information asymmetry tied to a graphics option, which means the game now rewards a hardware decision. If I owned that map, everything above is negotiable and this is not.

And here is the joke. Office is a hostage map that nobody plays competitively. It got balanced anyway. The category eats the map, whether or not the map is playing.

---

Destruction is the expensive part, classes are the cheap part

The Finals is the clearest current example of the third mechanism, which is that forms move inside people's heads.

Embark is DICE. The people who built The Finals built Battlefield from 1942 through to Battlefield V, which was where a lot of them stopped. You can see it in the animation and the art direction and above all the destruction, where geometry breaks into separate objects with independent physics rather than replaying a canned collapse, and it is server-side, which is the sentence that should make any working developer sit up. Server-side destruction means every one of those fragments has to be authoritative and agreed upon by everyone in the match. That is not an art feature. That is a netcode budget most studios would refuse outright, and they spent it because they had already spent it before somewhere else and knew what it bought.

Then look at the classes:

| Class | The design | What it costs to build |

|---|---|---|

| Light | Fastest, thinnest. SMGs, knives, air dash, grapple, cloak | Mobility tools are the hard part, grapple breaks level design assumptions |

| Medium | Average everything. AK, SCAR, healing beam, defib, dematerializer | Cheap, except the dematerializer, which is destruction wearing a gadget |

| Heavy | Slowest, fattest. RPG, minigun, sledgehammer, shields, goo gun | Goo and shields are geometry authoring in real time, so also destruction |

Light, Medium, Heavy is the oldest arrangement in the business. It is not why the game is good. Two of the three interesting gadgets in that table are the destruction system handed to the player under another name. The identity came out of DICE inside somebody's skull, and the class layer is the part you could have specced in an afternoon.

Embark also refuse the phrase hero shooter and prefer hero builder. I understand the instinct completely. One of the loudest videos on the subject is titled "Every Game is a Hero Shooter Now and It Sucks." That label is not a description any more, it is an accusation of homogenisation, and it arrives before you get to explain yourself. Fighting it is futile and I would fight it too.

---

Extraction, or what happens when your placeholders ship

This is the one that should be printed and pinned above every whiteboard.

The ancestry is older than the internet. Bartle's MUD1 at Essex in 1978 already had permadeath, inventory loss on defeat, and player versus player. When EverQuest and Ultima Online could not reconcile PvP with the people who did not want it, Bartle proposed the badlands: an opt-in unsafe zone inside a safe world, holding rewards you cannot get elsewhere, where every entry risks everything. That was 2001. That is The Division's Dark Zone in 2016, contaminated loot that has to be extracted, a cordon, a clock, everyone fair game.

Escape from Tarkov landed the same year and took the credit, and what it really did was translate the vocabulary. Potions and enchantments became tourniquets, medkits, stimulants, surplus ammunition, ballistic plating.

Then the detail that makes me want to lie down. Tarkov's fixed extract points, its timer, its permadeath, its character loss were placeholders. The pitch was a persistent open-world survival game like DayZ or S.T.A.L.K.E.R. The defining features of an entire genre were scaffolding that nobody got round to removing.

I want to be honest about why that lands so hard. Placeholders ship constantly. Half the systems shipped in demos were temporary, and the temporary ones were sometimes the ones players named the game after, because a constraint you did not agonise over is often cleaner than one you did. The lesson is not that Tarkov got lucky. It is that you cannot tell which of your compromises is the product until people play it, and you are the last person who will find out.

For years there was no name. Tarkov read as a harder battle royale. Vigor got the same treatment in 2018. Hunt: Showdown had started life in 2014 as something else entirely and sat awkwardly in the pile then and now. Only after the 2019 Twitch breakout did the term become necessary.

What happened next is the graveyard, and every headstone is a pitch meeting.

Hazard Zone bolted the literal mechanics of extraction onto Battlefield 2042, a game that was already in trouble. DMZ did better by treating extraction as a replacement for the circle rather than an addition to it. The Cycle: Frontier made the clock optional. Marauders shipped a double extraction loop into early access and learned what early access is. Creative Assembly pitched Hyenas to Sega in 2017 as PUBG crossed with Destiny and Tarkov, went free-to-play, and went nowhere. The Division: Heartland started as a battle royale mode for Division 2, became a standalone free-to-play title, and was cancelled with its people moved to XDefiant. Seekers of Skyveil, built by veterans of BioWare, Blizzard and Riot, peaked at 384 concurrent players and shut.

Read those together and the pattern is not that the ideas were bad. It is that they were pitched to the label. You can hear it in the descriptions. Somebody in each of those rooms said the word, and the word did the work that a design should have done.

The other half of the story is the boundary falling apart from inside. Quasimorph went top-down and turn-based and took as much from XCOM and Ultima as from Tarkov, which is to say it walked back to the same 1970s root by a different road. Dark and Darker is the only entry I believe would exist unchanged in a world without Tarkov. The Forever Winter proved PvP is not required. Hull proved hardcore simulation is not either. Once you accept both, Left 4 Dead qualifies, because you do literally extract at the end to win, and so do Deep Rock Galactic, Warframe, Payday, GTFO, Vermintide, Darktide, Helldivers and Lethal Company. Arena Breakout cloned Tarkov to mobile and came back to PC as its first serious free-to-play rival. Delta Force runs extraction through hero classes, which collapses two of the four FPS variants in the taxonomy into each other and nobody blinked. Marathon proposes something similar. Arc Raiders arrived and went vertical.

A category that holds both Escape from Tarkov and Left 4 Dead is not describing a mechanic. It is describing what an evening feels like.

---

What I actually do with the taxonomy

I use it to talk to people who are not designers. That is all it is for, and it is a legitimate use, because the store needs a shelf and the publisher needs a comparison and both of those things pay for the studio.

But the mechanisms do not care about it. Technology spreads a form when somebody sells it, which is how the FPS got everywhere. Communities produce forms when you let them modify your game, which is how a mod became one of the largest businesses in the medium by accident. People carry forms between companies, which is how Battlefield's destruction ended up inside a game show. Not one of those is a design document. Not one of them starts with a category.

So the honest version of what I know goes like this. Tarkov was placeholders. Office is a 2004 layout with better shadows and a shorter leash. The Finals is DICE's destruction wearing a class system it did not really need. The label always arrives late, always arrives from outside, and by the time it is stable enough to pitch against, the thing it describes has already moved. If you build to the name you are building to a photograph of something that has left.

Build the thing. Let them name it.