Live Service and the Game That Refuses to End

Every game used to have an ending, even the ones without a final boss. A single- player campaign ran out of level; an arcade board ran out of quarters or loops you back to the start with a harder difficulty flag set; a budget C64 title ran out of screens. The economics of the format guaranteed it — a publisher sold you a finished object, once, and had already been paid by the time you found out whether it was any good. Live service inverts that arrangement completely. The game is sold as a beginning, the revenue depends on your continued attention past that beginning, and the design has to manufacture reasons for the game to keep not-ending indefinitely. That single change in the business model has produced almost everything strange about the last decade of big-budget design.

The service is the product, and the game is the delivery mechanism

The clean way to see the inversion is to notice what a live-service pitch actually promises a publisher’s shareholders: a recurring audience that can be monetised repeatedly through seasons, battle passes and cosmetic stores, with a “content roadmap” replacing the idea of a finished release, valued above the one-time revenue from units sold. Diablo IV shipped in 2023 as a complete, playable Diablo game and then immediately adopted the seasonal-battle-pass structure the genre’s live-service neighbours had normalised — a fresh season every few months, a paywalled cosmetic track, character progress that resets so there’s a reason to keep logging in. None of that is dishonest exactly. It is a structural admission that the loot loop alone wasn’t trusted to hold attention on its own merits, so a calendar got bolted onto it to do the holding instead.

That’s the tell to watch for across the genre: when the roadmap becomes the selling point ahead of the game, the studio has already decided the finished object isn’t the thing being sold.

Concord proves the model can fail before the servers even warm up

The starkest failure of the last few years arrived and left almost too fast to review properly. Concord, Firewalk Studios’ hero shooter published by Sony, launched in August 2024 into a market already crowded with free-to-play alternatives, asked for a full purchase price on top of live-service upkeep, and was pulled from sale and taken offline within two weeks — a shutdown of a newly launched, fully funded, multi-year project on a timescale with almost no precedent at that budget level. Sony refunded every buyer and later confirmed the studio’s closure. The lesson isn’t really about the shooting, which reviews at the time described as perfectly competent. It’s that a live-service game has no fallback position the way a finished single-player release does — a mediocre singleplayer campaign still exists and can be replayed for a decade afterwards on a shelf or in a library, but a live-service game with no audience has, definitionally, no game left to have an opinion about once the servers go dark. The service is the product, so a service nobody uses is a product that has ceased to exist, not a product that underperformed.

Skull and Bones shows what happens when the wait outlasts the pitch

Ubisoft’s Skull and Bones is the opposite failure mode, running in slow motion instead of fast. Announced in 2013 as a spin-off of Assassin’s Creed IV: Black Flag’s celebrated naval combat, it went through repeated internal reboots and delays before finally launching in February 2024 — a public gap of roughly eleven years between announcement and release, an interval long enough that an entire console generation had come and gone underneath it. A live-service game’s central bet is that a persistent, evolving world justifies years of ongoing investment after launch; Skull and Bones spent that investment before launch instead, on a game that shipped into a live-service market that had moved on twice over while it was still in reboot. The finished product a decade of development produced was reviewed, fairly, against contemporaries built in a fraction of the time, because a customer buying a game today has no obligation to grade it on effort spent a console generation ago.

There’s a further wrinkle worth naming: the eleven years weren’t spent building one coherent live-service vision, they were spent rebuilding the premise itself, more than once, as internal leadership decided a singleplayer-leaning pirate adventure wasn’t the live-service product the market wanted and pivoted toward something closer to a persistent multiplayer economy. Each pivot restarted systems work that a finished, non-service game would never have needed to redo, because a finished game only has to satisfy the version of itself that ships. A live-service game has to satisfy a prediction about what live-service audiences will still want to log into years from now, and predicting that correctly turned out to be the harder design problem than the sailing.

Suicide Squad shows the tension between IP and structure

Rocksteady’s Suicide Squad: Kill the Justice League is the clearest case of a studio’s reputation running headlong into a structure it hadn’t built for. Rocksteady made its name on the Arkham series — dense, authored, single-player combat systems with a defined beginning and end. Suicide Squad, published by Warner Bros. in 2024, imposed a live-service skeleton onto that lineage: loot tiers, a seasonal roadmap, always-online co-op as the intended mode. The combat mechanics reviewed as recognisably Rocksteady’s; what reviewers and players objected to, consistently and specifically, was the loop wrapped around them — grinding gear stats for a story-driven superhero cast that the studio’s own prior games had proven didn’t need one. A structure isn’t neutral scaffolding. It changes what the studio underneath it is allowed to be.

The tension shows up in specifics that reviewers kept returning to at launch: a fixed roster of eight, then later expanded, playable characters instead of the freeform gadget-and-gliding traversal the Arkham games built their identity on; enemies designed as damage sponges to justify a gear-score system rather than as the tightly choreographed encounters Rocksteady had spent three prior games perfecting; and a story built around killing off beloved Justice League characters, delivered through a structure that asked players to replay the same handful of story missions across a seasonal pass rather than move forward into new authored content. None of that reflects a studio that forgot how to make a combat system feel good. It reflects a structure imposed from above the studio, by a publisher chasing a business model that had worked for other companies’ different games.

The FOMO clock is the actual game design

Underneath the roadmap sits a smaller mechanism that does most of the actual work of keeping a live-service title alive: the rotating store and the time-limited item. Fortnite popularised this at scale — a cosmetic catalogue that refreshes daily, a battle pass with a hard expiry date, collaboration skins available for a fixed window and then gone, in some cases permanently. None of that changes how the shooting or building plays. All of it changes how often a player feels obliged to check in, because the cost of not checking in is the specific, quantifiable loss of an item that will not be offered again. That’s a different kind of hook to the one a finished game relies on. A good finished game keeps you playing because the moment-to-moment mechanics are worth returning to; a live-service store keeps you logging in because the alternative is a fear of missing a thing, and fear of missing a thing works even on a game whose core loop you’ve grown tired of. Destiny 2’s sunsetting of old raid loot and its rotating vendor inventories run the same logic from the other side — Bungie has, at various points, actually removed content from the game specifically to manufacture scarcity and give the returning-content seasons something to feel like an event. A finished game doesn’t need to delete part of itself to stay interesting. A live-service game frequently does, because permanence is the enemy of a calendar that needs something new to announce every quarter.

Why the loop needs the calendar

Set those three next to each other and the actual mechanism becomes visible. A live-service loop needs the season, the pass, and the roadmap because the loop itself — kill things, get loot, kill bigger things — was never going to sustain months of attention on pure mechanical interest alone. Deep Rock Galactic is the counter-case worth naming, because the co-op loop that respects your time proves a live-service structure can work when the seasonal content is additive rather than load-bearing — the base game was already complete and fun before a single season pass existed, so the ongoing service adds to a finished thing instead of propping up an unfinished one. That’s the difference between a service built on top of a game and a service built instead of one.

The calendar the design actually serves

The reason this matters beyond any single launch is that a live-service structure changes what “finished” is allowed to mean at the pitch stage, before a single line of code gets written. A single-player game is greenlit against the question “is this a complete, satisfying experience”: the loot loop with a season pass bolted on is greenlit against a different question entirely — “can this sustain quarterly content for years” — and that second question gets asked by finance long before anyone asks whether the underlying game is good. The earnings call replaces the review as the document that actually judges the release, because the metric that determines whether the game continues to exist is retention and monthly active users, not critical or even commercial reception in the first month.

That’s the quiet cost underneath every live-service failure, spectacular or slow. A finished game that disappoints still exists afterwards, gets reassessed, gets its defenders. A live-service game that disappoints gets switched off, and the switching-off is not a metaphor — the servers actually stop, the purchases stop working, and the thing a customer paid for retroactively becomes a thing that was only ever rented. The industry keeps building games designed never to end, and then ending most of them anyway, just later and more expensively than a finished product would have, with the customer holding a licence to nothing once the lights go out.