Hunt: Showdown 1896 review in 2026. How the game everyone wrote off plays today, after setting its record six years late and still growing

Some games explode in their first month and slowly fade. Hunt: Showdown did the exact opposite: it launched in a state that was painful to watch, spent a year and a half at rock bottom, and then set its all time player record six and a half years after release. That shape of graph almost never happens in this industry.

This is the story of how a cancelled co-op monster hunting game turned into the best shooter ever built around a single shot. About an engine that is ten years old and still out-looks multiplayer games half its age. About rifles that each have a real nineteenth century original behind them. And about a studio that could not pay its own staff while the game was being made.

It started as a completely different game

In June 2014, at E3, Crytek showed Hunt: Horrors of the Gilded Age. If you know today's Hunt, you will struggle to recognise it in that one.

It was a third person action RPG. Camera over the shoulder, four player co-op against the computer, late nineteenth century America, a squad of hunters clearing out monster lairs. Procedurally generated levels, bosses, loot. The plan was to give it away free and monetise later.

It was made by Crytek USA, a studio in Austin. Crytek assembled it in 2013 from people left jobless by the collapse of THQ: the team from Vigil Games, the people behind Darksiders. David Adams, director of the first Darksiders, led the project, and he said plainly what he was after.

The plan was laid out: a closed beta late in 2014, then a release on PlayStation 4 and Xbox One. Neither the beta nor that release ever happened.

If you want to see for yourself how different it was, watch IGN's coverage from that show. Camera over the shoulder, four in a squad, none of the tension that comes from other people being on the map, because there are no other people yet.

And then the trouble started, not with the game, but with the company.

Crytek hits bottom

By 2016 Crytek was in a situation you can only call a disaster. Salary delays began in May. By December staff still had not been paid for October and November. People talked about it publicly because there was no other way left to change anything.

The company owned seven studios around the world. In autumn 2016 it announced the closure of five of them: Austin, Budapest, Sofia, Seoul, Istanbul. Only Frankfurt and Kyiv survived. Around two hundred and fifty people lost their jobs.

Austin closed along with the project. Hunt: Horrors of the Gilded Age vanished: the work moved to Frankfurt, and the team that had come up with all of it scattered. Formally the game was never cancelled. In practice what remained was a folder of assets and a general idea.

Second birth: May 2017

In May 2017 Crytek revealed Hunt: Showdown. What survived from the old version was the setting, the Louisiana swamps and the monsters. Everything else was rewritten, including where the player looks at the world from.

Co-op against the computer turned into a format that barely had a name back then and is now called an extraction shooter. Several teams of two or three enter the map. Everyone is looking for the same target. Finding it is not enough: you have to kill the boss, run the banishing ritual and reach an extraction point. The other hunters know where you are the whole time, because the ritual lights up for the entire map.

Formally it is PvPvE. In practice it is a game where people, zombies and your own greed are all trying to kill you at once.

22 February 2018: the launch that nearly finished it

Hunt entered early access on 22 February 2018. It came, honestly, straight out of alpha. There was no open beta, the closed tests drew a small audience, and the first mass wave of players met a build that was not ready for a mass wave of players.

That last point deserves its own explanation, because it turns a technical fault into a personal tragedy. There is no respawn in Hunt. A dead hunter is struck off your roster along with everything they were carrying and every skill they had earned. When the game loses one to its own bug, there is no worse feeling available.

But even that was not the main problem. The main problem was that the game ran terribly.

Thirty frames on the best hardware money could buy

Crytek itself acknowledged that Hunt was CPU bound on most configurations, and that the graphics options were there for comfort more than for frame rate. In practice that looked like this: people with the best components of 2018 were getting numbers that would earn a refund today.

Within days of the early access launch, forum reports looked like this: Core i7-6700K and GeForce GTX 1080 Ti, 1920 by 1080, medium settings, 35 to 40 frames. The flagship card of its generation could not hold sixty at medium in Full HD. The wording in those posts was not flattering.

Streamers had it twice as bad, for an obvious reason. A broadcast is encoded on the same machine, and Hunt was strangling exactly the CPU the stream also needed. The person showing the game to an audience got the same picture as everyone else, only slower, and live.

In a shooter decided by fractions of a second, thirty frames is not a cosmetic problem. It is the point where you physically cannot keep up with your opponent, and no amount of skill covers for it.

How heavy the game stayed even later shows in the benchmarks from the 1.0 release in August 2019: a locked sixty frames at 2560 by 1440 was held only by an RTX 2080 Ti, the most expensive card on the market. That was after a year and a half of optimisation.

The first impressions all said roughly the same thing: great idea, raw state, come back in six months. And readers listened.

The graph that explains everything

A new game almost always peaks in its first month. After that the curve goes down, and the only question is how gently. Comparing different games by absolute numbers means little, but comparing the shape of the curve means a lot.

Hunt's curve is broken in both directions. The bad one first.

Look at what happens. The launch peak of 10,296 is modest on its own: for a loud game from a famous studio that is not much. Then it gets worse. In two months the peak falls to 2,262, a drop of 78 percent. The average in April 2018 is a little over a thousand people. A thousand. For a game by the company that made Crysis.

And this was not a brief dip. Through all of 2019 up to August the game sits at an average of about 1,300. In April 2019, fourteen months after launch, the average is 1,258. That is the level at which shutting the servers down usually gets discussed.

For a Crytek that had just closed five studios and was dealing with unpaid wages, this was a very bad moment for very bad numbers.

Now the other broken direction. Hunt set its all time record in August 2024: 59,968 players at once. That is almost six times the launch peak, six and a half years later. By the summer of 2026 the game holds an average of around sixteen thousand and monthly peaks above thirty thousand, roughly thirteen times higher than at the bottom.

You can count on your fingers the games whose all time player record landed in their seventh year. Usually it means the game was rescued not by marketing but by itself: people held on until the technology stopped getting in the way.

Why Crytek was making a shooter at all

Crytek was never simply a studio that makes games. It sells an engine. Crysis in 2007 was less a shooter than a demonstration: look what CryEngine can do, come and license it. The old joke about whether your PC can run Crysis was born right there.

Hunt was built on the same logic, only the task was harder. Plenty of studios could show a beautiful single player game with lighting calculated in advance. Crytek wanted to show something else: dynamic global illumination computed in real time, working on a large open map, surviving destruction, and still holding up a twelve player networked shooter.

The technology is called SVOGI: Sparse Voxel Octree Global Illumination. The name sounds frightening, the idea is simple.

About those voxels: how it actually works

Care is needed here, because a claim circulates around CryEngine that it is "the only engine in the world with voxel rather than pixel rendering". It is a nice line, but it is not accurate, and understanding it beats repeating it.

CryEngine draws its picture the same way everyone does: by rasterising triangles into pixels. There is no voxel rendering, meaning no image built out of volume cubes instead of polygons. The voxels live somewhere else, and that somewhere is the interesting part.

The engine builds a volumetric grid of cubes from the scene, an octree in which empty space is not stored at all, hence the word sparse. It traces cones of light through that grid and works out how light bounces from surface to surface. It is not calculating the direct beam from a lamp but the secondary light: why the shade under an awning is not absolute black, why a wall next to a red barn picks up a hint of red, why light from a window lands as a patch on the floor and lifts the ceiling.

Then come the two things this was all for.

So the honest version is duller but stronger: CryEngine was one of the first engines to use voxels for real time global illumination, and in 2018 that was rare. It was not the only one in the world then and it certainly is not now. What it was, is early to the party.

Is the light really being calculated while you play?

A fair question, because "real time lighting" turned into a marketing sticker a long time ago and gets slapped on anything. With Hunt it can be checked, in two independent ways.

First, the graphics settings. Hunt has a separate Global Illumination option, and it is one of the most expensive settings in the game in frame rate terms. Baked light cannot be switched off with a slider, it is painted into the textures and costs nothing. If both the picture and the frame rate change when you toggle it, something is being computed right now.

Second, somebody else's analysis. Digital Foundry took Hunt: Showdown 1896 apart on consoles and confirmed SVOGI runs on all three, Series S included. They call Hunt's lighting one of the best showcases in gaming.

Now the caveats, without which this would be a flattering picture rather than a true one.

In the 1896 relaunch Crytek added a light cache and specular tracing on top of this, so bounced light now lands on glossy surfaces too, not only matte ones. Plus multiple bounces: the light gets to travel more than once.

The engine is ten years old and still better looking than the new ones

CryEngine V shipped in 2016. By industry standards that is old age. And yet Hunt still looks better than a good share of the multiplayer shooters that came out after it.

The reason is not polygon counts or texture resolution. The reason is light and air. Morning fog over water in which a human silhouette registers a second before you work out who it is. A dusty shaft of light through a gap in a barn roof. A thunderstorm that turns the map white for a fraction of a second and then black again, and in that instant you realise somebody was standing twenty metres away. Beauty in Hunt is functional: it directly decides whether you see the enemy or not.

And here is a detail people usually skip. For all those years Hunt did not run on a current CryEngine but on its own branch of the engine, split off back in the early access days. It got small, targeted updates, because swapping the foundations under a live networked shooter is frightening. A game of the mid twenties ran on late tens technology, and still looked better than many of its contemporaries.

In August 2024 that was fixed in one jump. Hunt: Showdown 1896 is not a sequel but the same project moved onto the internal CryEngine 5.11 with DirectX 12 and modern upscalers. The work took the studio several years, and it is a rare thing: almost nobody changes the engine under a live networked game, because it is expensive and dangerous.

Crytek also cut off the previous console generation. PS4 and Xbox One were left behind, only PS5 and Xbox Series X|S are supported, but now at sixty frames instead of thirty. A debatable call for owners of the older machines and absolutely the right one for the picture. That same month produced the record of 59,968 players.

A game of chess where any piece can take the king

The main difference between Hunt and every other shooter fits in one line: any shot can end your match.

A hunter has 150 health, split into chunks. A headshot from almost any weapon kills instantly. A body hit from a heavy rifle leaves very little of you. No armour, no shields, no hero who leaps behind cover and regenerates in three seconds. There are medkits, but they heal slowly, loudly, and they do not resurrect.

Out of this grows a completely different tempo. In an ordinary networked shooter you run forward because death costs ten seconds of waiting. In Hunt you stand behind a tree for four minutes and listen. Because death costs a hunter, their gear, and every bit of loot you have already collected.

Reaction time and aim matter less in Hunt than in the average shooter, while position, information and patience matter more. Hence the game's reputation: it is frightening to start and hard to leave.

Sound is the main weapon

In Hunt your headphones give you more information than your screen. The map is built so that crossing it quietly is nearly impossible, and all the trap wildlife works as an alarm system for the neighbours.

The computer controlled enemies have their own audio design. They have several states: calm, alert, fighting, dying. Each one sounds different. An experienced player who hears zombies in a distant compound shift from calm to fighting already knows three things: there are people there, they are making noise, and they will be busy for the next thirty seconds.

The result is a second layer on top of an ordinary shooter. You are not so much shooting as reading a sound map, while trying not to appear on somebody else's.

The hunting mechanic: clues and Dark Sight

The name of the game is not about hunting people. It is about the fact that the target has to be found first, and that is a separate mechanic other shooters simply do not have.

Clues are scattered across the map. To see them the hunter switches to Dark Sight: the screen dims and the relevant points start glowing. Each clue you collect strikes part of the map out of the search area. After the third one the circle shrinks to a single compound and you know exactly where the lair is.

The elegance of it is that it works against you exactly as much as for you. While you collect clues you are running around the map making noise. And you are not alone: another team is collecting the same clues from the other side and shrinking the same circle. The closer you get to the target, the higher the chance of meeting people who came for the same thing. By the third clue you both know where you are going, and you both know the boss will not be the only thing waiting.

Who lives on the map

The monsters in Hunt are not scenery and not filler for a kill counter. Each type has a job, and almost always that job is to get you heard.

And this is where the variety genuinely surprises, especially remembering that this is a networked shooter and not a single player campaign. Usually such a game has one enemy type, maybe two: a dummy that runs at you and swings. Hunt has around fifteen, and they are built on fundamentally different principles.

One just shuffles and makes noise. Another breaks into a sprint and runs you down in the open, and you cannot outrun it, only turn and fight. A third releases a swarm of insects that follows you, gets in your eyes and does not leave until you dive into water. A fourth is blind and hunts by sound, so you can slip past it if you have the patience to crawl. A fifth burns and explodes, and its death is heard across half the map. A sixth stops bullets with its body, and shooting it is pointless.

Bosses and banishing: a reward you pay for

A boss sits in the lair. Several have accumulated over the years: the Butcher with his hook and brazier, the Spider running across walls and ceilings, the Assassin who dissolves into a swarm of insects, and Scrapbeak with claws made of scrap metal. Each demands its own tactics, and each fights in an enclosed space with nowhere to scatter.

And now the interesting part. Killing the boss is half the job. Then you have to run the banishing, and here the game plays its best card.

The ritual takes minutes, you have to stay near it the whole time, and everyone who was hunting the same target converges on you from all sides. Taking the reward is not enough, you also have to carry it to an extraction point, and more hunters die on that stretch than on everything before it.

The economy is simple and cruel. You do not have to take the bounty at all: you may leave the map at any moment, and the experience for clues found and enemies killed stays with you. But only the boss token really pays. So every match is a bet: leave with a little now, or stay and try to take it all.

If a hunter dies, they die for good, together with their gear and their progression. That is exactly why the price of one wrong step in Hunt feels nothing like it does in a game with respawns.

The real hunt starts after the bounty

This deserves a longer stop, because the title of the game earns itself precisely at this moment, and plenty of people who have watched a couple of videos never realise it.

While you are looking for clues and fighting the boss, you are the hunter. The moment you pick up the bounty token, the roles swap. Your position is marked on the map for every living player. In Dark Sight, lightning strikes the ground where the tokens are, visible from far away. The game literally puts a marker on you and shows it to everyone still alive.

Now you have to reach an extraction point. And everyone else has to intercept you, and they know exactly where to go. More than that, they know which way you are heading, because there are only a few extractions and they sit at the edges of the map. An experienced team does not chase you, it gets ahead of you and sits where you have to appear.

Because of that the last minutes of a match are built completely differently from everything before them. A moment ago you were creeping and listening. Now you are running with a marker on your back, knowing there is an ambush ahead, deciding whether to go straight through or take a detour across half the map through the swamp.

You are not left defenceless. The bounty carrier gets tools, but those are built as a choice rather than a gift.

Five seconds for the whole trip to the exit. Spend them now and learn where the ambush is, or carry them to the end and trigger them at the moment the ambush seems to have started. It is one of the best decisions in the game: the resource is so small that every activation is a bet.

There were two bosses, now there are six

It is worth showing separately how the game grew, because at launch it was far more modest. In February 2018 there were exactly two bosses: the Butcher and the Spider. That was it. You followed the clues knowing one of two things waited in the lair, and after a couple of hundred hours that started to wear thin.

Look at the gaps. Fourteen months passed between the second boss and the third, and that is exactly the period when a little over a thousand people were playing. Crytek kept adding content to a project that by every external sign was dying. In hindsight it looks like stubbornness that paid off.

Traits: how a player reshapes the game around themselves

Another layer that is invisible from the outside. A hunter has traits, bought with upgrade points as the hunter grows. There are 85 of them right now, and a single hunter can carry up to fifteen at once.

These are not passive plus five percent damage bonuses. Almost every trait changes a specific rule of the game, and together they assemble into different styles.

Then the same brutal arithmetic applies as everywhere else. Traits live on the hunter, and the hunter dies permanently. Build the perfect character across fifteen traits, pour hours into them, walk through the wrong door, and that is that, start again. Which is why the better your hunter, the more carefully you play, and the game tightens its own tension without a single line of interface.

How many weapons Hunt actually has

Counting the arsenal in a shooter is usually dull, because half the entries are the same rifle with a different scope. In Hunt the count is worth doing.

Then comes the part that makes this worth mentioning. Every base weapon has its own model, its own reload animation, its own gunshot and its own iron sight. Not a shared template with a repainted barrel, but a separately made object.

The sights deserve their own note, because there is almost no optics in this game and you aim with the standard iron sights. The Winfield has one arrangement, the Mosin-Nagant another, the Martini-Henry a third. Each blocks a different part of the screen and behaves differently at distance. Switching from one gun to another is not changing numbers on a stat sheet, it is retraining your eye.

Variants follow the same logic. A Shorty is a cut down barrel with a different animation and a different sound, a Silencer changes both the sound and how far it carries, Aperture and Deadeye are a different front sight or sighting rail. So 92 variants are not 92 rows in a table, they are 92 different feels in the hand.

For a game that in 2019 was held up by fifteen hundred concurrent players, that is an indecent amount of handmade work.

Ballistics: the numbers that decide it

In Hunt the bullet travels. Not an instant hitscan ray as in most networked shooters, but a physical object with its own speed and its own arc. At fifty metres you do not notice. At two hundred you aim above the target, and against a moving opponent you lead the shot.

Every gun has its own set of numbers, and they are not invented: rate of fire and damage follow fairly closely what the real weapons were.

Look at the rifle rows. The Sparks LRR deals 149 damage against a health pool of 150. One point short. This is not an accident or a developer typo, it is a design decision that borders on cruelty: the most powerful rifle in the game kills anyone with a body shot except a hunter at full health. That hunter survives on one point, crawls behind cover and heals. And you, at that moment, are reloading a gun whose cycle takes almost five seconds.

The second interesting point is bullet speed. The fastest are the Lebel 1886 and the Mosin-Nagant M1891, at 630 and 615 metres per second. The slowest is the Sparks, at 533. And this is where the game unexpectedly turns into a history lesson.

The real originals: why the numbers look like that

Almost every weapon in Hunt is a genuine nineteenth century design under a slightly altered name. Sometimes the name was not even changed.

Now put the table of originals next to the table of ballistics, and you can see that Crytek did not invent the numbers, it copied the logic.

Which produces a rare thing: to understand the weapons in Hunt it helps to read about real weapons. And the other way round, after playing Hunt you unexpectedly understand why the move to smokeless powder in 1886 was a revolution rather than a small improvement.

Cheaters: a war that does not end

Now for the thing enthusiastic articles usually stay quiet about. Hunt has a cheating problem, and it has had one for a long time.

The game is built so that a cheater costs more here than in an ordinary shooter. In Call of Duty you get killed unfairly and lose ten seconds. In Hunt you lose a hunter with all their gear, forty minutes of loot and accumulated skills, and none of it comes back. On top of that the clue mechanic works against the honest player: Dark Sight already highlights points on the map, and someone with third party software only has to take one step further and see the people too.

Crytek fights this, and unlike many it does so in the open, publishing reports.

And the newest part: in April 2026 Crytek brought in Anybrain, an external system that collects in game input from mouse, keyboard and controller and looks for abnormal behaviour in it. Put simply, it does not look into process memory, it looks at how a person moves their hand, and tells a living movement from a software assisted one.

It was pointed at consoles first, where it catches mouse and keyboard play disguised as a controller. That is now a bannable offence, and 296 bans had been issued as of the report. On PC the system is being trained to catch aimbots and scripts, but Crytek is deliberately holding back mass sanctions: first the criteria get tuned, so that legitimate players are not swept up.

The studio also noted separately that the system is GDPR compliant and collects no social models, no mood tracking and no personal data. Given how news about AI powered anti-cheat usually sounds, the clarification is not superfluous.

The problem is not solved, though. It has got smaller and better tracked, but writing it off would be premature.

Hit registration: killed after you took cover

The second long standing sore point is familiar to anyone who has played Hunt. You duck around a corner, you can already see the wall in front of you on your screen, and at that moment you die. It feels appalling, because it looks like cheating or like a broken game.

It is not broken. Crytek explained the mechanism itself, and the explanation is worth knowing, because it changes how you feel about it.

Because of network latency, nobody in a match sees the same thing. You see the world as it was forty milliseconds ago, the server sees it differently, your opponent sees a third version. When somebody shoots, the dispute is settled by the hit validation system on the server, and it is forced to take a side. Hunt, like most networked shooters, takes the shooter's side.

The server rewinds the world state back to the moment the shooter saw, and checks whether the hit was legitimate. The higher the shooter's latency, the further back the rewind can go, with a hard cap of 800 milliseconds.

So you are not imagining it and nobody cheated you. On the screen of the person who fired, you were still standing in the open, and the game settled the dispute in their favour.

There have been major fixes: one update revealed that a large number of shots were being invalidated because the server considered the shooter already dead, and things improved noticeably once that was corrected. But it cannot be fully cured while light and cables refuse to travel faster.

And the other half of the problem: server tick rate. Here honesty is needed twice over. On one hand, Crytek does not publish that figure, and every number circulating in the community is somebody's measurement rather than official data. On the other hand, people have been measuring since 2018 and the picture repeats: the rate is not so much low as unstable. Peaks go above a hundred, everything sags under load, and players describe it as feeling like twenty or thirty updates a second.

The tick rate complaint is one of the oldest complaints about Hunt, it is now in its eighth year, and it has not been retired. Formally you cannot say "Hunt has a bad tick rate", because there is no official number. Based on how the game feels, you want to say it fairly often.

Put it all together and it becomes clear why this hurts more in Hunt than in neighbouring shooters. The bullet here is physical and takes its time. The rewind reaches 800 milliseconds. The tick rate floats. And the price of one hit is an entire hunter. In any other game that combination would produce unpleasant moments. Here it produces moments you remember for a long time, and not fondly.

In a game where any hit can end the match, the cost of such a small thing is higher than anywhere else, and complaints about hit registration follow Hunt to this day.

Our verdict

9.5

Tension and the core idea 10

Sound as a mechanic 10

Visuals and lighting 10

Weapons and ballistics 10

Fight against cheaters 9

Networking 8

Good Stuff Tension that exists nowhere else: any match can end in the first minute Sound as a full mechanic rather than decoration Lighting that in 2026 still beats shooters five years younger 147 entries in the arsenal, every base weapon built separately down to the iron sights Ballistics copied from real nineteenth century weapons A target finding mechanic nobody else has A studio that has been fixing the game for eight years and reports on it

Bad Stuff Cheaters have not gone away, and meeting one costs more here than in any other shooter Dying after you reached cover: explainable, still infuriating every time A floating server tick rate, a complaint eight years old and still alive A brutal barrier to entry, in the first hours you die without understanding why Permanent loss of a hunter forgives mistakes less than some nerves allow PS4 and Xbox One owners were left behind by the relaunch

Summary

The best shooter built around a single shot that exists. Half a point comes off for cheaters and for the networking: deaths behind cover and a floating tick rate. Nothing else.

Where that leaves it

Hunt: Showdown is a game that by every law of the genre should have been shut down. The project was rebuilt twice, the first version died along with its studio, the company was not paying wages while it was being made, the launch was a technical failure, and fourteen months after release a little over a thousand people were playing it at once.

Instead of shutting down it set its all time player record in 2024 and lives today with twenty times the population it had at the bottom. The reason seems to be exactly one: underneath all the bugs and failures lay an idea you cannot get anywhere else. A shooter where a shot is an event, where you listen more than you fire, where an old engine keeps the light alive and the rifles behave like their real originals did a century and a half ago.

And here is the thing worth saying plainly: there is nothing else like it. Not in the sense of "the best", but literally: nobody has assembled this combination. The extraction genre is crowded now and has its own big names, but they all run on modern weapons, modern gear and raids about economy. Nobody else has the 1890s, nobody else puts the target in a lair behind a ritual that gives your position to the whole map, nobody else makes sound a separate detection system, and nobody else makes a single shot cost this much. The nearest neighbour in spirit differs from Hunt more than any two ordinary shooters differ from each other.

In eight years nobody has copied it. Usually that means there is nothing to copy. Here it seems to be the opposite: too much of it would have to be built from scratch, including an engine that can compute light the way Hunt does.

If you have never tried it, an honest warning: for the first ten matches you will die without understanding why. On the eleventh you will hear crows over a distant barn, understand what that means, and that is it, you are caught.