Why PUBG Bet on Unreal Engine 4: The Borrowed Shooter Tech Behind 100-Player Battle Royale

PUBG should not have worked. A Korean MMO studio with 60 people, led by an Irish photographer-turned-modder with no shipped game, tried to put 100 real players with rifles, cars, and red-zone artillery on an 8x8 km island - in 2017, on consoles and toasters.

It worked because they rented the hard problem. PUBG: Battlegrounds bet on Unreal Engine 4, a C++ shooter engine that already knew how to draw guns, replicate movement, and ship on Windows, Xbox One, and PlayStation 4.

That choice explains both the miracle and the mess: record Early Access growth and 75M+ copies sold, plus two years of parachute stutter, lobby crashes, and "optimization" memes. PUBG sold over 75 million copies on PC and consoles and passed $13 billion in total revenue including mobile, went free-to-play in January 2022, and still runs on UE4 today.

My take, upfront: everyone says PUBG won despite UE4 jank. I think it won because UE4 forced it to stay ugly, slow, and readable. That is the opposite of the usual "pick the best graphics" advice, and it is the lesson indie devs should steal.

The bet in 30 seconds

• Game: PUBG: Battlegrounds, originally PlayerUnknown's Battlegrounds, battle royale for up to 100 players, solo/duo/4-man squad, last alive wins.
• Who: PUBG Studios (ex-Bluehole Ginno, 60 people in 2015) under Krafton, directed by Brendan "PlayerUnknown" Greene and producer Kim Chang-han.
• Tech: Unreal Engine 4, released March 19, 2014, written in C++ with Blueprints visual scripting and physically based materials.
• When: Steam Early Access March 2017, full Windows release December 20, 2017, Xbox Game Preview December 2017, PS4 December 7, 2018. Free-to-play since January 2022.
• Scale: 100 players per match, maps from 1 to 64 sq km, Erangel as the iconic 8x8 km launch map, 30-minute rounds, shrinking blue zone + red-zone bombing + flare-gun care packages.
• Business: $19/month + 5% royalty when UE4 launched in 2014, then source-available on GitHub. For a 60-person team that had just laid off a third of staff after a failed MMO, renting was cheaper than building netcode for 100 connections.

If you only remember one sentence: PUBG did not choose UE4 to look next-gen. It chose UE4 so a modder's idea could become a server product in 18 months.

Why UE4 was the only sane pick in 2016

1. Greene was a modder, not an engine programmer

Greene's resume was DayZ: Battle Royale mod for Arma 2 in 2013, PlayerUnknown's Battle Royale mod for Arma 3, then consultant on H1Z1's battle royale mode for Sony Online Entertainment. He joined Krafton (then Bluehole Ginno) in 2016 after H1Z1 caught their eye.

Arma gave him the design - airplane drop, wide terrain, leaderboard - but Arma's Real Virtuality engine could never ship 1M concurrent players on Xbox. UE4 gave him instant C++ replication, character movement, weapons, vehicles, and console export without hiring an engine team.

This is my first contrarian point: UE4 was Greene's co-founder. Solo-concept designers usually fail because they try to learn engine code. Greene skipped that and spent his leverage on rules: when the plane flies, how loot spawns, how the blue wall squeezes.

Indie takeaway - if your design needs 100 networked bodies, do not build the bodies.

2. 100 connections was an unsolved problem - even Epic struggled

UE4 officially supported Windows, macOS, Linux, PS4, Xbox One, Switch, iOS, Android, and later PS5/Xbox Series X/S as of 4.27, but large open worlds with 100 high-bandwidth connections were not turnkey.

Epic admits this itself: to ship Fortnite Battle Royale in September 2017, Epic had to modify UE4 to handle up to 100 connections to the same server while retaining bandwidth and to improve large open-world rendering, then rolled those fixes back into the engine.

PUBG hit the same wall nine months earlier. That is why launch PUBG is famous for late-parachute houses, desync on doors, and cars that kill you when you exit. It was not lazy coding. It was two different licensees discovering the same UE4 ceiling in parallel.

Fortnite's answer was to stay cartoony and add building to hide pop-in. PUBG's answer was to stay military-sim ugly and add fog, grass, and ditches to hide pop-in. Same engine limit, opposite art.

3. Cross-platform was the business plan

PUBG shipped Windows, Android/iOS March 19, 2018, Xbox One September 4, 2018, PS4 December 7, 2018, Stadia April 28, 2020, and Xbox Series X/S + PS5 November 13, 2025. You cannot do that in 2017-2018 with a custom C++ engine on a 60-person team.

UE4's one-codebase export to console + mobile is why PUBG Mobile could be built in just 4 months by LightSpeed Studios with Tencent, evolving UE4 for mobile with smoother performance and enhanced graphics. That mobile version later hit around 1.75 billion downloads and over $15 billion gross by January 2026. No custom engine would have allowed a 4-month mobile port.

UE4 royalties hurt, but missing mobile would have killed the company.

What UE4 gained - and what it cost

Gained:

• Speed to market: mod idea in 2015 to Early Access phenomenon March 2017.
• Readable gunplay: UE4 character + weapon + vehicle replication meant bullets, 8x scopes, UAZs, and boats worked day one, even if badly.
• Talent pool: every Korean shooter contractor already knew UE4 materials, Blueprints, and profiling. Hiring was "who knows UE4?" not "who will learn our engine for 2 years?"
• Live-service tooling: patches, crates, event modes, spectator tools for esports - all easier on a mainstream engine.

Cost:

• Performance shame: Xbox One at 20 fps, Erangel houses made of Play-Doh at 50 meters, infamous "pan kills level-3 helmet" physics bugs. All UE4 streaming + replication tradeoffs on 8x8 km.
• Royalty + size: 5% on gross plus tens of GB client. For a free-to-play game after 2022, every skin funds Epic as well as Krafton.
• Sameness risk: because Fortnite Battle Royale was also built on Unreal (now Unreal Engine 5) after starting in mid-2017 inspired directly by PUBG, launching September 26, 2017 after two months of work from Save the World assets, PUBG could never claim "our tech is unique." It had to win on rules, not renderer.

That last point is my second contrarian idea: PUBG vs Fortnite is the cleanest A/B test in engine history. Same base tech, same 100-player circle, same 2017 window. One added building and cartoon hit-reacts. The other kept prone, bullet drop, and no respawn. Tech did not decide the winner. Verb choice did.

If you are an indie choosing UE4/UE5/Godot today, stop asking which is faster. Ask which verbs you will forbid.

Three things I think differently from most "PUBG engine" posts

Most retrospectives say: modder has idea, picks popular engine, gets rich. That is true and useless. Here is what I actually take away after lining up the timelines:

1. The airplane is a loading screen that players love.

PUBG's C-130 flight path looks like flavor. Technically it is genius resource management: 100 clients stream different parts of Erangel while everyone is stuck in the same cabin for 60 seconds. No lobby timer feels dead, no one Alt-Tabs, and the server staggers spawn load. Fortnite copied it with a Battle Bus for the same reason.

If you build a big-map multiplayer game, do not build a better loading bar. Build a plane.

2. The blue zone is server-cost control disguised as drama.

Everyone praises the shrinking circle as tension. It is also a performance contract: every minute, the relevant replication set shrinks. Fewer grid cells, fewer audible gunshots, fewer cars to simulate. On UE4 in 2017, you could not simulate 8x8 km for 30 minutes for 100 players. But you could simulate 1x1 km for the last 5 minutes.

The design makes the engine's weakness feel like suspense. When you design for a weak engine, make the choke point the fun.

3. Ugly was a moat, not a bug.

PUBG looked worse than Battlefield 1 and ran worse than Overwatch at launch - and outsold both on PC that year. Why? Because muddy brown + fog + grass means a player in a ghillie suit at 200m is information, not eye candy. Fortnite needs bright skins to sell. PUBG needs you to not see the enemy.

UE4's physically based materials were wasted on PUBG on purpose. Low readability ceiling = high skill ceiling for eyes and ears. My hot take: if PUBG had looked like a UE4 ArchViz demo, it would have died. Streamers could not have sold "did you see him?" moments.

None of this is in the infobox. The infobox just says Engine: Unreal Engine 4. The interesting part is how a team used a generic shooter engine to ship a game that feels nothing like a shooter for the first 10 minutes - loot, run, hide - then everything like a shooter for the last 2.

What indie devs can steal this week

• Rent replication. If your prototype needs >16 players, use UE4/UE5, Mirror, or Nakama. Do not write netcode. PUBG's 60-person team could not have debugged packet loss and gun balance.
• Pick one map size and defend it. Erangel 8x8 km is not realistic. It is the largest size UE4 + 2017 servers could handle with 100 players and still force a fight every 3 minutes. Calculate your encounter rate first, then paint terrain to fit.
• Make performance part of fiction. Red zones, fog, night Erangel - all hide LOD pops. If your engine streams slowly, add weather, not a progress bar.
• Plan the 4-month port on day one. PUBG Mobile worked because UE4 scaled down. If you start in Godot/Unity/Unreal, test your lowest target phone in month one, not year two.
• Let the mod lead. Greene proved balance patches beat engine patches. Ship Early Access ugly, fix circle timing weekly, and let YouTube sell the fantasy.

Train the PUBG muscles in your browser

You cannot practice 100-player drops in a 5-minute break, but you can train the three skills PUBG actually tests: angle discipline, burst control, and drop awareness.

If you want close-quarters PvP where positioning beats aim, try Hazmob FPS: Online Shooter - its Team Deathmatch and Search and Destroy modes punish solo pushes the same way Erangel apartments punish loot greed. Play two rounds of Free for All just to warm up crosshair placement before you queue PUBG.

For tactical holding and clutch retakes, Block Contra: Clutch Strike is the closest browser vibe to PUBG's final circles: hold a lane, stop panic-spraying, learn one safe angle and one aggressive angle per area. Its rank-chase loop scratches the same "one more match" itch without a 30-minute commitment.

If your problem is spraying entire mags and dying on reload, run a session of BodyCamera Shooter. The chest-cam fisheye forces short 3-round bursts, cover-to-cover movement, and reloading only behind concrete - exactly the habits that win PUBG mid-game fights in Pochinki.

And if you secretly love only the first 60 seconds - the plane, the "where do we go?" call - Air Battle Royale: Sky Blitz distills that drop fantasy into one-finger dogfights. No looting, just pick a line, hunt rivals, and practice committing to a landing spot instead of dithering over the map.

None of these replace PUBG's sound design or bullet drop, but they keep your decision loop warm when you cannot install 40 GB.

Bottom line

PUBG did not prove UE4 is the best battle-royale engine. It proved a borrowed engine plus a modder's rules can beat a perfect engine with no rules. Krafton started as Bluehole, founded March 2007, reorganized as Krafton November 5, 2018 after PUBG's success and Tencent investment - a mid-size MMO house that accidentally defined a genre because it refused to build tech.

If you are choosing tech today, copy the refusal, not the logo. Rent the engine, own the circle.