How to prototype before you commit

1 The bottom line

Build a small, ugly prototype that tests your biggest risk before you commit to production. Imagination is a poor simulator, so only a playable test tells you whether an idea is fun, viable and worth years of work.

2 Key takeaways

• Play the idea before you trust it. Almost every idea sounds fun in your head.
  • Nintendo designers follow the mantra "make before we talk" and build something playable, such as Splatoon's ink, before pitching it Mark Brown and Luke Muscat, Game Maker's Toolkit0:26.
  • Mark assumed his earlier games would be fun, and only later learned that successful games start as prototypes Mark Brown, Game Maker's Toolkit9:05.
• Keep the prototype ugly, small and disposable. Polish is wasted work on something you may throw away.
  • Klei's first Don't Starve prototype used Link from Zelda as the hero Mark Brown, Game Maker's Toolkit9:49.
  • Dinosaur Polo Club made almost 20 small versions of Mini Motorways before production Mark Brown and Luke Muscat, Game Maker's Toolkit12:07.
• Aim each prototype at your biggest risk and leave the rest for production. A prototype gives confidence, not every answer.
  • Jetpack Joyride's prototype was a machine gun jetpack with a roof and some obstacles, and the missions, challenges and vehicles came later Mark Brown and Luke Muscat, Game Maker's Toolkit13:32.
• Let the prototype change your game. Building shows ideas that planning cannot.
  • Ryan Clark's one-second-turn roguelike felt rhythmic, which made it obvious that Crypt of the Necrodancer should be set to music Mark Brown, Game Maker's Toolkit2:52.
  • Subnautica became a horror game without that being the plan Mark Brown, Game Maker's Toolkit6:49.
• Once the design is proven, rebuild properly and stop prototyping. Prototype code is built to be replaced, and endless prototyping ships nothing.
  • Word Play's game jam version used plain if statements, and it needed a new architecture once it grew past 100 modifiers Mark Brown, Game Maker's Toolkit3:32.
  • "If you never stop making prototypes, you won't make a video game" Mark Brown and Luke Muscat, Game Maker's Toolkit14:45.

the best game designers know that their brains are horrible smelly liars.

Mark Brown, Game Maker's Toolkit

2.1 Pick the risk your prototype must answer

A prototype is a small test of your scariest question, not a crude version of the whole game.

• Prototype only the biggest risks. Mark Brown and Luke Muscat, Game Maker's Toolkit14:45
  • Ask the big questions first: is it fun, is it viable, will people like it.
  • Name the question before you build, because a fuzzy premise will not find the fun for you.
  • Tom Francis hoped that going inside spaceships would pay off as it had in Gunpoint. He built ship generation, AI, combat and a galaxy map with an economy before finding that on-ship combat was the best part of Heat Signature Mark Brown, Game Maker's Toolkit7:37.
• Test the core mechanic before art, story, music or code polish. Mark Brown, Game Maker's Toolkit6:55
  • You can fix bugs and redraw graphics, but you cannot fix a game that is not fun.
  • Carter's Curse had sprites, cutscenes and juicy menus, but its Picross-fighting gameplay was shallow and dull. Mark said he had built a house on shaky foundations.
• Read the build effort as a feasibility check.
  • If the core gameplay prototype takes one or two days, a full game is plausible in one or two years. Check this against your own time, staff, budget and expertise Mark Brown and Jonas Tyroller, Game Maker's Toolkit20:17.
  • In a puzzle game, easy puzzle-making means fertile ground, and hard puzzle-making means an uphill battle Mark Brown and Luke Muscat, Game Maker's Toolkit4:44.
  • Mark's video rental store prototype took weeks, which showed the game was too big for him. Tiny Bookshop later launched as a better version of the same idea Mark Brown and Luke Muscat, Game Maker's Toolkit4:44.
• Prototype art, camera and story when they are the risk. Mark Brown and Luke Muscat, Game Maker's Toolkit3:56
  • The Thronefall team made separate prototypes just to test art styles and camera angles.
  • Pixar often makes a whole film as a rough comic book animatic with stand-in voices to check the story.
  • Mark made his Mind Over Magnet sprites long before the final camera. Large sprites in a 1280 by 720 room scale cleanly to 720p, 1440p and 4K, but not to 1080p, the most common Steam resolution Mark Brown, Game Maker's Toolkit5:45.

2.2 Build it fast, ugly and disposable

Use the cheapest tools and materials that answer your question, and ban everything else.

• Decide up front what is off limits. Mark Brown, Game Maker's Toolkit10:33
  • Ban app icons, names, story and music until the gameplay is proven.
  • Use grey boxes, programmer art, borrowed sprites and untidy code, because art and story also distract playtesters from judging the mechanic Mark Brown and Luke Muscat, Game Maker's Toolkit8:57.
  • Mark's magnet prototype used images from Google Images and buggy Unity code, with no music, story or name Mark Brown, Game Maker's Toolkit9:45.
  • Mark still lost a week on particle effects, shaders and UI before scrapping that version.
  • Breath of the Wild began as a top-down 2D prototype using Zelda 1 sprites Mark Brown and Luke Muscat, Game Maker's Toolkit8:57.
• Borrow engine tools and scripts for solved problems. Mark Brown, Game Maker's Toolkit16:28
  • Mark found Unity's point effector, which works as a magnetic field, and downloaded a character controller. Understand what you borrow.
  • Pick a small, readable controller you can adapt. Mark chose Shinjingi's over tarodev's and Mix and Jam's, because less code is easier to bend Mark Brown, Game Maker's Toolkit2:04.
  • Fujibayashi used the Breath of the Wild engine to attach spinning dungeon cogs to a slab for a car, then built a cannon from a tube, a bomb and a ball Mark Brown, Game Maker's Toolkit1:16.
  • When prototyping a UI, one script spawned about 100 echo items from wiki images in one click, so five menu ideas could be tried quickly Mark Brown, Game Maker's Toolkit6:10.
• Prototype in whatever medium answers the question fastest. Mark Brown and Luke Muscat, Game Maker's Toolkit10:23
  • Thatgamecompany prototyped Journey in Macromedia Flash before making anything for PlayStation 3. Storyteller's developers used cardboard mock-ups, and Mark used Scrabble tiles and handmade cards for Word Play.
  • For a simulation with many interlocking systems, test it in a spreadsheet or a tabletop game before coding Mark Brown and Luke Muscat, Game Maker's Toolkit11:06.
  • Locator's developer ran a detective prototype over Discord by sharing sketches and asking players what they would do next. Kojima planned Metal Gear Solid levels in Lego viewed through a tiny camera.
  • A prototype does not have to be playable. Ori and the Blind Forest started as simple animations Mark Brown and Luke Muscat, Game Maker's Toolkit11:42.
• Use placeholder shapes and tiles until the design is locked.
  • Splatoon's prototypes had you controlling a big white block. The squid kids came later, as a character that could swim in ink and separate inking from swimming Mark Brown, Game Maker's Toolkit6:55.
  • Mind Over Magnet used one plain blue tile set with no decoration, so Mark could move elements and restart levels while focusing only on puzzle design. He painted over the layouts once they were finished Mark Brown, Game Maker's Toolkit5:28.

2.3 Keep it small, sketch first, block out for real

Many small, specific tests beat one big crude copy of the game.

• Make each prototype a tiny sample of one feature, and make many. Mark Brown and Luke Muscat, Game Maker's Toolkit12:07
  • Keep gameplay, art, technical and audio prototypes in separate project files.
  • Fiddling in the engine can spark ideas that brainstorming would not.
• Sketch many scenarios with one mechanic before building any. Chris McEntee, Moon Studios1:58
  • Take the elements you have and create as many unique situations as you can on paper.
  • Chris sketched collapsing platforms used to force a drop and a double jump, and used to guide a projectile across a room by breaking every platform.
• Give a new ability a clear job by sketching what is impossible without it. Chris McEntee, Moon Studios4:16
  • Recreate those sketches as blockouts and hand them to programmers as the target for how it should play.
  • In Ori, spikes block a wall jump, but digging through the wall launches you out. A dive into sand and a dash clears spikes that a normal jump would hit.
• Block out spaces with the real camera and controller. Mark Brown and Evan Hill, Naughty Dog2:00
  • You only know a space works once you can play it in 3D, before any art exists.
  • Evan Hill built the museum level in blockmesh with Ellie's real controller and camera.

2.4 Play it, and read what it tells you

Put the prototype in front of players early, and judge the feeling, not the concept.

• Playtest within days, even with one rough room. Mark Brown, Game Maker's Toolkit9:24
  • Early problems leave time for a real redesign, while late fixes become band-aids. Early tests also stop you investing art and audio in a mechanic that may be cut.
  • Kim Swift's team ran Portal playtests one week in, with one half-finished room and orange textures.
  • Half-Life's first attempt could not be played all the way through, so Valve scrapped it. In the restart they began playtesting about three months in, with random gamers recruited from game shops and registration cards Mark Brown, Game Maker's Toolkit5:54.
  • Rare filmed Phil Spencer and colleagues playing a rough Sea of Thieves, which showed it would produce cooperation and banter Mark Brown and Luke Muscat, Game Maker's Toolkit5:41.
• Test whether the core feeling is right.
  • Check that the prototype is fun to pick up and play, even without polish. Also check that it is original and keeps producing ideas Mark Brown, Game Maker's Toolkit17:56.
  • Luke Muscat's one-button Uzi shooter pushed you backwards with every shot. Always being pushed away from targets felt unassertive, which is wrong for a game about shooting crowds Mark Brown and Luke Muscat, Game Maker's Toolkit3:29.
  • Neon White began with random card deals mid-fight. When Ben Esposito swapped in a breadcrumb trail of hand-placed cards, friends started racing through levels and trading times Mark Brown, Game Maker's Toolkit2:37.
• Cut controls and systems that players cannot easily learn or feel.
  • DON'T NOD prototyped four-limb climbing on four shoulder buttons and a stick-coil jump. Both were hard to teach and execute, so they settled on holding the jump button to charge Mark Brown, Game Maker's Toolkit3:56.
  • Playtest several versions of a resource system and drop the punishing or confusing ones Mark Brown, Game Maker's Toolkit11:59.
  • Jusant's stamina tried per-arm bars, ground-only regeneration, rope-only regeneration and a forced descent. It ended as one stamina that refills any time you click the left stick.

2.5 Follow what the prototype teaches

A prototype is also a way to find better ideas, so let failures and accidents steer you.

• Rebuild the game around the most fun part. Mark Brown, Game Maker's Toolkit0:59
  • Look to the game itself to find where development should lead, not to your plans.
  • Ape Out began as a time-looping stealth game. Gabe Cuzzillo found that holding guards hostage and throwing them at walls was the best part, so he cut stealth and time loops and made the hero a 300-pound gorilla.
  • Rebuilding is a cheap chance to find a better version. Swapping Mark's sprite for a magnet made the character hold a magnet instead of being magnetic, and with two entities "the ideas just started flowing" Mark Brown, Game Maker's Toolkit14:46.
• Turn happy accidents into features.
  • Seb Liese forgot to attach a physics rope to the ceiling. It fell and coiled in an interesting way, so he made it controllable and added a head, which became Snake Pass Mark Brown, Game Maker's Toolkit2:09.
  • Rollerdrome's devs found they could clip through enemies unharmed with the right dodge timing. They kept it as the perfect dodge instead of patching it Mark Brown, Game Maker's Toolkit7:06.
• Try several quick answers when you do not know the solution, and learn from the ones that fail. Mark Brown, Game Maker's Toolkit5:09
  • Diablo 3 tried diluted potions, healing only when unseen, and healing after three seconds without damage.
  • Players chugged four potions, enemy awareness was unclear, and players ran away passively. The last failure pointed to health orbs dropped by killed enemies, which made players aggressive.
• Treat a failed prototype as a cheap success.
  • Cancel a concept when testers show its core does not work. After nearly a year on F-Stop, a camera puzzle game, playtesters made it clear that "Portal without portals just didn't work." Valve restarted Portal 2 Mark Brown, Game Maker's Toolkit16:53.
  • Luke Muscat released Luck of the Draw, a roulette-wheel deck builder, and no one cared. He dropped it after weeks instead of two years Mark Brown and Luke Muscat, Game Maker's Toolkit8:05.

2.6 Move from prototype to production

When the big questions are answered, plan, rebuild and scale up, and keep testing new ideas before they enter the game.

• Treat a jam entry as a seed to rebuild. Mark Brown, Game Maker's Toolkit8:57
  • The 48 hour Roller Dome jam game was fiddly and crazy hard, but the core idea held. Paul Rabbitte remade it in 3D, and GIFs of it drew publishers.
  • Superhot, Titan Souls, Baba Is You and Keep Talking and Nobody Explodes also started in jams.
• Test new functions with a small set, then scale the content. Mark Brown, Game Maker's Toolkit3:00
  • Word Play's first version had 11 perks, built mostly to test functions such as boosting one tile or the whole word.
  • Once the architecture worked, perks grew to 52, 74, 90 and finally 160.
• Plan once the prototype shows what is fun. Mark Brown, Game Maker's Toolkit3:10
  • An early design document stifles experiments, but afterwards you need a plan to turn loose elements into something you can ship.
  • Mark found a plan essential to finishing Platformer Toolkit, but made none for Mind Over Magnet until two years in.
• Keep prototyping to settle later decisions, but gate what enters the game.
  • If you cannot choose between a double jump and a dash, make quick prototypes of both, test them and keep the winner Mark Brown and Luke Muscat, Game Maker's Toolkit15:43.
  • Nothing should pass from the workshop into the game until it has been tested in different scenarios Mark Brown, Game Maker's Toolkit13:10.
  • Nearly every Mind Over Magnet puzzle had a bug or oversight that let players skip it.

2.7 Where designers disagree

Some of this advice pulls in opposite directions, and the right call depends on your game and your team.

• Skip the juice, or keep the juice that is part of the fun.
  • Skip polish and keep prototypes disposable Mark Brown and Luke Muscat, Game Maker's Toolkit8:57.
  • For visceral games, feedback can be the core fun, so removing all of it can hide the truth Mark Brown and Luke Muscat, Game Maker's Toolkit9:20.
  • The Fruit Ninja prototype was simple but had splatter particles. The danger is spending a month on 20 effects that band-aid a core that is not fun.
• Time-box the test, or follow the fun wherever it goes.
  • A prototype that takes one or two days points to a game of one or two years Mark Brown and Jonas Tyroller, Game Maker's Toolkit20:17.
  • Open-ended following of the fun makes schedules hard to predict, so it is more common in indie work than in regimented blockbuster production Mark Brown, Game Maker's Toolkit7:37.
  • Naughty Dog's museum took about two solid years, and some sequences were redone around 25 times Mark Brown and Evan Hill, Naughty Dog14:58.
• Stop prototyping, or keep prototyping through production.
  • Prototyping is addictive, because the next one might always be better. Learn when enough is enough Mark Brown and Luke Muscat, Game Maker's Toolkit14:45.
  • It is also a cheap way to answer any question during production, such as new features, boss fights and levels Mark Brown and Luke Muscat, Game Maker's Toolkit15:43.
• Experiment in a low-risk side project, if your studio can afford one.
  • Link's Awakening was a handheld game, so the team felt free to try new things. Takashi Tezuka said it was like making a parody of Zelda Mark Brown, Game Maker's Toolkit0:53.
  • Mark argues Deathloop took Mooncrash's core idea but dropped the smart patches that made it work. A different Arkane studio built it, with little overlap in staff Mark Brown, Game Maker's Toolkit1:36.