Conquest of Elysium 5

There is a moment, roughly forty minutes into a first game of Conquest of Elysium 5, when it becomes clear nobody is coming to help. No tooltip explains why Bakemono raiders evaporate against a stack of militia. No quest marker suggests researching that ritual. No difficulty slider apologizes for the fact that a rival Necromancer can march an undead legion out of a hole in reality the player didn’t know existed. The game is not being cruel. It simply assumes you’re an adult who can read a manual.

That assumption is the entire thesis of Illwinter Game Design, the two-person Swedish studio that has spent two decades building the Dominions series — notoriously impenetrable strategy games that nonetheless sustain a devoted audience. Conquest of Elysium 5, released on Steam in August 2021 for $29.99, is their lighter offering, and calling it “lighter” tells you everything about their calibration. Where Dominions asks for a week, CoE5 asks for an evening. It compresses the same alien design philosophy into a roguelike-adjacent, turn-based sprint: a fresh procedural map every session, a class picked from 24 radically asymmetric options, and a scramble for resource-bearing terrain that resolves in an hour or three.

The market positioning is almost aggressively unfashionable. This is not a 4X in the Civilization tradition, with its onboarding funnels and advisor pop-ups and carefully tuned dopamine schedules. It looks like something that escaped from 1998. And its long-tailed Steam audience — the kind that hemorrhages the wrong customers instantly and keeps the right ones for years — is the signature of a game built for a specific kind of player and nobody else.

What Illwinter understands, and what much of the industry has forgotten, is that discovery is a mechanic. Over 1,500 monster types, 500+ rituals, 60+ paths of battle magic, and up to ten planes of existence stacked on top of one another mean the game’s real content isn’t the map. It’s the enormous, uncatalogued space of things you don’t yet know it can do. Every session is an archaeology dig.

• 24 Asymmetric Classes: Not factions with different unit rosters — different games. The Baron collects taxes and hires mercenaries. The Necromancer harvests corpses and has no economy to speak of. The Cultist courts Elder Gods with blood slaves and will eventually summon something that may or may not obey them. The Troll King is a troll king. Balance was clearly never the goal; distinctiveness was.
• Ten Planes of Existence: The surface map is one layer. Beneath and above it sit Hades, the Inferno, elemental realms, and a sky plane whose flying armies can descend on the world below at a time of their choosing. Vertical strategic depth in the literal sense — territory you can’t see is territory that can still kill you.
• Procedural Everything: New map every game, and — critically — randomized ritual and spell availability per match. You cannot memorize a build order. The optimal line changes because the tools changed.
• Terrain-Driven Economy: Mines, farms, villages, herb gardens, blood slaves. Resources are geographic, which makes every expansion decision a commitment and every border a pressure point.
• Auto-Resolved Tactical Battles: Armies meet, the game plays out the fight on a tactical grid while you watch, and you don’t touch it. Composition is your input; the outcome is the output.
• 1–24 Players: Hotseat, LAN, or online, cross-platform across Windows, macOS, and Linux, with Steam Workshop mod support.

The Gameplay Loop

The core loop is plain on paper. Move armies onto terrain that produces things. Spend the things on troops and rituals. Repeat until someone is dead. Written out, it sounds like a mobile idle game.

In practice, the loop’s tension comes from resource specificity. You don’t accumulate a generic currency; you accumulate iron from mines, food from farms, gold from taxed villages, herbs from wilderness, and — depending on how morally flexible your class is — human beings. Each class converts these differently, and most classes are structurally starved of something. That scarcity is what makes the map matter. A mine two provinces away isn’t an optimization. It’s the difference between fielding an army and not.

Layer the planes on top and the strategic geometry gets genuinely strange. A quiet border on the surface means nothing if your opponent has been quietly farming the Inferno. The sky plane, in particular, is a feat of design anxiety: flying armies operate on a layer you may not be contesting at all, and they choose where and when to come down. Most 4X games model threat as adjacency. CoE5 models it as something that could arrive from anywhere, including up.

Then there’s summoning. With 500+ rituals in the pool and a random subset available per game, the ritual screen is less a tech tree than a slot machine you’re allowed to study. Marching an army across the map is the slow, reliable path. Rituals are the fast, volatile one — and because the pool rerolls each session, the strategic question is never “what’s the best opening” but “what does this game let me do, and how quickly can I bend it into a win condition?” That’s a fundamentally different cognitive exercise than the optimization puzzle most strategy games offer, and it’s why runs stay fresh long past the point where a fixed-tech-tree game would calcify.

The Battle Abstraction

Here’s the design decision that splits the audience: you don’t control battles. Armies collide, the tactical grid resolves it, you watch. Your influence is entirely upstream — unit composition, positioning on the strategic map, magical support, whether you brought something that can handle armored infantry.

We think this is correct, and it’s the game’s most under-appreciated choice. Tactical combat in strategy games is a time sink that rewards micromanagement over planning; by amputating it, Illwinter keeps every decision on the strategic layer and keeps sessions to an evening. The pace is fast for the genre.

But the criticism is fair and we won’t wave it away. When a carefully assembled army disintegrates for reasons the auto-resolver doesn’t explain, it stings — and the game does a poor job of teaching you why it happened. The information to understand the loss exists. It’s buried in unit stats and magic path interactions you’re expected to have internalized. Losing without learning is the single most common way new players bounce off this game, and it’s a real failure of communication, not player skill.

Onboarding, or the Absence Thereof

There is essentially no tutorial. You are pointed at a manual and a community wiki. In 2026, this is an extraordinary thing to ship.

There are two readings. The generous one: this is a deliberate filter, and the game’s density resists tutorialization — how do you teach 1,500 monsters? The honest one: a great deal of what’s hidden isn’t deep, it’s just undocumented. Knowing that a specific terrain yields a specific resource is not wisdom. It’s a fact the interface should have told you. Illwinter conflates obscurity with depth more often than it should, and a handful of contextual tooltips would cost the game nothing and save hundreds of players.

The AI Problem

The opposition is weak. This is the most substantive complaint in the game’s community, and it holds up. Experienced players describe an AI that expands adequately but rarely threatens, doesn’t punish overextension, and can’t leverage the multi-plane structure with anything like human cunning. Against a game this asymmetric, that’s a meaningful ceiling — the classes that are overtuned stay overtuned because nothing pushes back.

The answer, and Illwinter clearly knows it, is other humans. Up to 24 players, hotseat or networked. In multiplayer the imbalance becomes texture rather than a flaw; drawing a weak class becomes a story instead of a tuning complaint. Solo, the game is a fascinating toybox with a short competitive half-life. Multiplayer, it’s where the design works best.

The presentation is sprite-based, flat, and unapologetically archaic — hand-drawn icons on a tiled map, a fixed-width interface, and typography that appears to predate the concept of kerning. It is not retro as an aesthetic pose. It’s retro because two people made it and put their hours into systems instead of shaders.

The monster sprites have real character, and the visual density does honest work: you can read a province’s contents at a glance. The planes are differentiated with genuine mood.

The UI, though, is the game’s worst component, and it’s not close. Critical information hides behind clicks that aren’t signposted. Comparative data — the exact thing a strategy game must surface — often requires holding two screens in your head. There’s no clean way to answer “will this army beat that army,” which, given that combat is auto-resolved, is the question the interface exists to help you answer. This isn’t charming minimalism. It’s an information architecture problem, and no amount of goodwill toward small studios makes it not one.

Performance, at least, should be a non-issue. A 2D game of this scope is light on hardware, and turn processing is quick even in large multiplayer games. There’s something quietly refreshing about a strategy game that doesn’t ask for a gaming rig.

On Steam, this is a straightforward mouse-and-keyboard experience and it fits the hardware naturally — dense clickable maps, right-click context menus, lots of screen real estate. Cross-platform parity across Windows, macOS, and Linux is real and rare; the Mac and Linux builds are first-class, not afterthoughts, which matters given how many strategy holdouts run Linux.

Steam Workshop mod support is the platform’s biggest contribution. Given a rules-dense game with an open monster and ritual database, the modding community meaningfully extends the content and, occasionally, patches around balance the developers won’t touch.

The one platform gap: Steam’s networked multiplayer works, but matchmaking and lobby infrastructure are bare-bones. Finding humans to play with is a Discord-and-forums exercise, not an in-client one. For a game whose best mode is multiplayer, that’s a self-inflicted wound.

Conquest of Elysium 5 is a game built entirely without regard for what a game is supposed to be in 2026. There’s no onboarding funnel, no progression meta, no seasonal content, no attempt to be balanced or beautiful or welcoming. Illwinter made the thing they wanted to make, charged $29.99 for it, and left it to find its people.

It found them. The game’s devoted following isn’t consensus — it’s self-selection working as intended, a filter that repels the mass market at the door and rewards everyone who pushes through with a strategy game of uncommon imagination.

The flaws are real and we won’t soften them. The interface is bad. The AI is a pushover. The balance is openly lopsided, and the absence of a tutorial is a choice that costs the game players it should have kept. Any of those would sink a full-price release, and together they keep this from being more than a solid, deeply flawed work.

But what’s underneath is a design no funded studio would greenlight: ten stacked planes of reality, 1,500 monsters, a ritual pool that reshuffles every session, and 24 classes so different from one another that mastering one teaches you almost nothing about the next. That’s not content volume. That’s imagination at a scale the industry has stopped attempting, delivered by two people.

Buy it if you’re the kind of player who’s still here after reading all that. You already know if you are.