Portal

In-level commentary nodes turn a finished game into a design lesson. Valve placed audio nodes in Portal and Left 4 Dead where designers explain mechanics, teaching, guiding the eye, puzzles and playtests. It inspired Mark Brown's commentary mode in Mind Over Magnet.

Mark Brown: "this feature just completely opened my eyes to game design as a craft."

The restart left two lasting habits: small cabals and frequent playtests. A cabal is a tiny multi-disciplinary team that fully owns a small chunk of the game. Pairing it with constant playtesting became how Valve builds games, from Half-Life 2 to Left 4 Dead, Portal and Half-Life: Alyx.

Gabe Newell calls playtesting "our secret weapon" and former in-house psychologist Mike Ambinder calls it "the most important part of the game development process." A designer who joined in 2018 said he "was not prepared for the amount of playtesting this company does."

Treat each design as a hypothesis and each playtest as the experiment. The team sets a goal, such as a puzzle that is clearly readable and satisfying to solve, designs toward it, playtests, and repeats until the tests are "no longer excruciatingly painful to watch."

Valve frames game development as an engineering problem or scientific study. Ambinder: "We see our game designs as hypotheses and our playtests as experiments to validate these hypotheses." Playtesting here is distinct from QA, which hunts bugs, and focus testing, which is market research.

Portal ran on a weekly loop: test Friday, review Monday, fix all week. Portal was tested almost every week, with results discussed Monday and the build tested again Friday. Mark Brown copied the rhythm for Mind Over Magnet: new content Tuesday to Thursday, bug fixes Friday, then a build for a few testers.

In that rhythm new content starts rough and experimental, and early levels seen by many testers end up seriously polished.

Valve scrapped Half-Life two months before ship and restarted around playtests. The first Half-Life could not be played start to finish, its levels did not tie together and it had serious technical problems, so Valve threw it out. In the restart, playtesting began about three months in, with random gamers recruited from game shops and from returned registration cards.

Valve wanted to know right away if the game sucked, and each test produced dozens of fixes. Testers broke every crate in a level, so Valve filled some crates with ammo and health. The game went on to be hugely influential.

The restart left two lasting habits: small cabals and frequent playtests. A cabal is a tiny multi-disciplinary team that fully owns a small chunk of the game. Pairing it with constant playtesting became how Valve builds games, from Half-Life 2 to Left 4 Dead, Portal and Half-Life: Alyx.

Gabe Newell calls playtesting "our secret weapon" and former in-house psychologist Mike Ambinder calls it "the most important part of the game development process." A designer who joined in 2018 said he "was not prepared for the amount of playtesting this company does."

Treat each design as a hypothesis and each playtest as the experiment. The team sets a goal, such as a puzzle that is clearly readable and satisfying to solve, designs toward it, playtests, and repeats until the tests are "no longer excruciatingly painful to watch."

Valve frames game development as an engineering problem or scientific study. Ambinder: "We see our game designs as hypotheses and our playtests as experiments to validate these hypotheses." Playtesting here is distinct from QA, which hunts bugs, and focus testing, which is market research.

Test within days of prototyping, with programmer art and one rough room. Kim Swift's Portal team ran a playtest one week into working at Valve, with only one half-finished room. Early problems get real redesigns; late ones get band-aids like a character simply telling you a puzzle's solution.

Testing before art and audio also avoids polishing a mechanic that ends up cut. Swift called playtesting "the most important thing that we learned since coming to Valve."

Portal ran on a weekly loop: test Friday, review Monday, fix all week. Portal was tested almost every week, with results discussed Monday and the build tested again Friday. Mark Brown copied the rhythm for Mind Over Magnet: new content Tuesday to Thursday, bug fixes Friday, then a build for a few testers.

In that rhythm new content starts rough and experimental, and early levels seen by many testers end up seriously polished.

Watch players in silence; body language beats their answers and suggestions. Observers give no hints and answer no questions so the test simulates real play. Kim Swift says "you'll really tell by their body language whether or not they actually enjoyed themselves."

Testers' own proposed fixes are usually ignored because they don't know the designer's vision or constraints. One Portal tester spent half an hour pushing a shelf onto a button while ignoring a nearby box.

The designers who built a level run its tests themselves. Valve does not hand testing to a separate department or publisher. The cabal that made the level watches it, which shows exactly what to fix, motivates the team and sparks ideas.

In Half-Life: Alyx, players covered their real mouths to stop Alyx coughing near the blind zombie Jeff, and Valve turned that into a mechanic. Testers range from staff and children to pro gamers, but changes target the intended audience.

Use many testers so trends stand out from outliers. Each Half-Life 2 chapter had about 100 playtesters. Newell says after a couple hundred playtests you develop a sense of which design strategies work, yielding lessons like "players don't learn when stressed" and "players don't look up."

Valve's setup is hard to copy, and the method may suit linear story games best; heat maps may fit other genres better.

Valve checked live play data and patched a difficulty spike. In 2006 Valve tracked Steam data from Half-Life 2: Episode One, saw players getting stuck in one siege battle and patched it to be easier.

Play data confirms or contests what players say, but can mislead, so it works best as one source among several.

Many signature Valve features began as fixes for what testers did. Testers who called Portal a great tutorial led to GLaDOS, and players missing objects in cluttered rooms led to Portal's plain white walls. Players loved the gravity gun, so Half-Life 2 handed it out earlier, and Left 4 Dead got x-ray outlines because testers couldn't find teammates in trouble.

A wall-walking paint in Portal 2 was cut after several testers felt queasy, and Half-Life: Alyx sped up because VR players tolerated standing and watching people talk less. Robin Walker says GLaDOS began as "a straightforward process of us trying to solve the core gameplay problem in Portal."

Testing overturned the belief that Portal needed a complex final puzzle. Testers found the simple mid-point fire pit escape more climactic than any complex puzzle, thanks to time pressure, visual impact and high drama. The GLaDOS finale became a simple sequence too.

Kim Swift: "we've been holding on to this idea that we need a complex puzzle at the end and it simply wasn't true."

Feedback is data; Valve overrules testers when it conflicts with the goal. Testers loved Half-Life 2 getting to combat quickly, but Valve delayed it so players would witness the cops' cruelty first and the crowbar payoff would land harder. Hardcore shooter fans wanted a harder GLaDOS fight, but it frustrated Portal's audience, so that challenge went into optional content.

Writer Marc Laidlaw: "We wanted you to witness the cops doing something horrible, and feel like what you were doing was a response to that." Bending to every tester produces "design-by-committee sludge"; testers show whether the target is hit, they don't choose it.

Testers killed F-Stop, the first Portal 2, despite Gabe Newell loving it. F-Stop came from an internal game jam: players photographed objects and spawned them elsewhere at a different scale. After nearly a year, testers said Portal without portals didn't work, and Valve started Portal 2 over.

Newell had wanted each sequel to feature a different Aperture Science technology. Fans later rebuilt F-Stop from leaked assets and code.

Portal builds a puzzle game on familiar first-person shooter controls. It borrows FPS conventions players already know to deliver a non-violent puzzle game, blending first-person aiming with puzzles around one smart piece of interactivity.

FPS fluency is part of the puzzles, which Mark Brown says takes a lot of skill to pull off; in his own game players couldn't tell when a room needed thinking versus quick execution, so he cut the dexterity challenges. Miyamoto called Portal's core interactivity amazing.

A shift in colour marks Portal's turn from testing to escape. The white, sterile, low-detail test lab gives way to a warm orange, lived-in maintenance area as the player starts escaping an evil AI.

Portal's structure is strictly linear, so one hard puzzle can stop progress, unlike hub games such as Braid or Baba Is You.

Portal builds a puzzle game on familiar first-person shooter controls. It borrows FPS conventions players already know to deliver a non-violent puzzle game, blending first-person aiming with puzzles around one smart piece of interactivity.

FPS fluency is part of the puzzles, which Mark Brown says takes a lot of skill to pull off; in his own game players couldn't tell when a room needed thinking versus quick execution, so he cut the dexterity challenges. Miyamoto called Portal's core interactivity amazing.

Portal introduces its abilities one at a time and orders levels to build. Each puzzle relies on earlier tutorials and revelations, so shuffling Portal's levels would make it practically impossible for a new player to get in. Mark Brown copied the slow rollout, unlocking pick up, then polarity, then recall.

Powerful tools are held back because puzzles need walls; Half-Life 2's late gravity gun upgrade shows how an overpowered tool would trivialize earlier challenges.

Portal builds a puzzle game on familiar first-person shooter controls. It borrows FPS conventions players already know to deliver a non-violent puzzle game, blending first-person aiming with puzzles around one smart piece of interactivity.

FPS fluency is part of the puzzles, which Mark Brown says takes a lot of skill to pull off; in his own game players couldn't tell when a room needed thinking versus quick execution, so he cut the dexterity challenges. Miyamoto called Portal's core interactivity amazing.

Valve checked live play data and patched a difficulty spike. In 2006 Valve tracked Steam data from Half-Life 2: Episode One, saw players getting stuck in one siege battle and patched it to be easier.

Play data confirms or contests what players say, but can mislead, so it works best as one source among several.

A shift in colour marks Portal's turn from testing to escape. The white, sterile, low-detail test lab gives way to a warm orange, lived-in maintenance area as the player starts escaping an evil AI.

Portal's structure is strictly linear, so one hard puzzle can stop progress, unlike hub games such as Braid or Baba Is You.

Portal introduces its abilities one at a time and orders levels to build. Each puzzle relies on earlier tutorials and revelations, so shuffling Portal's levels would make it practically impossible for a new player to get in. Mark Brown copied the slow rollout, unlocking pick up, then polarity, then recall.

Powerful tools are held back because puzzles need walls; Half-Life 2's late gravity gun upgrade shows how an overpowered tool would trivialize earlier challenges.

Portal 2 seeds hints in the layout so hard ideas stay fair. In the laser, cube and button puzzle the cube starts in front of the laser, showing it can block the beam, and the only portal wall makes a light bridge right over the button, visible through the semi-transparent bridge.

The Turing Test's near-identical puzzle is far easier because it places the bridge already over the button and makes the cube effortless; Portal 2 makes you create and remove the bridge yourself. Mark Brown: "Portal's presentation is just so much more effective."

Lines from buttons to doors show how each room is wired. Portal draws lines between buttons and doors that change colour when powered, so players solve the puzzle instead of decoding the rig.

Portal builds a puzzle game on familiar first-person shooter controls. It borrows FPS conventions players already know to deliver a non-violent puzzle game, blending first-person aiming with puzzles around one smart piece of interactivity.

FPS fluency is part of the puzzles, which Mark Brown says takes a lot of skill to pull off; in his own game players couldn't tell when a room needed thinking versus quick execution, so he cut the dexterity challenges. Miyamoto called Portal's core interactivity amazing.

Portal builds a puzzle game on familiar first-person shooter controls. It borrows FPS conventions players already know to deliver a non-violent puzzle game, blending first-person aiming with puzzles around one smart piece of interactivity.

FPS fluency is part of the puzzles, which Mark Brown says takes a lot of skill to pull off; in his own game players couldn't tell when a room needed thinking versus quick execution, so he cut the dexterity challenges. Miyamoto called Portal's core interactivity amazing.