Inside Raph Koster's Philosophy — The Moment a Game Is Solved, It Becomes a Puzzle

Introduction

Raph Koster is known as the lead designer of Ultima Online (1997), the creative director of Star Wars Galaxies (2003), and above all as the author of A Theory of Fun for Game Design (2004). The book was translated into Japanese by O'Reilly Japan in 2006 (his own CV), so his name has reached Japanese readers even where his games have not.

But the reason to profile him on a puzzle site is not MMO history. He is that rare writer who has spent more than twenty years publicly re-defining a single question: what is the difference between a game and a puzzle? As he put it in 2025: "If a problem basically has one answer, we often call it a puzzle. There's not a lot of uncertainty built into a binary structure" (his blog, 2025).

Read once, that seems to file puzzles below games. But his argument turns the same blade on games too. This is not a review. It is a study of Koster the designer—his philosophy, his obsessions, his failures and how he absorbed them, his dilemmas, and his influences—built only from what he has himself put on the record.

From the thumbnail of “Practical Creativity” (GDC Festival of Gaming, YouTube)

Background: a designer trained as a poet

By his own CV and About page, Koster was born on Long Island, New York, on 7 September 1971. His mother worked for UNICEF, so he spent six years in Peru and time in Barbados and Haiti before finishing high school in Florida (About Raph). He took a BA in English and Spanish at Washington College in 1992 and an MFA in Creative Writing at the University of Alabama in 1995; his thesis was a poetry collection, Housebuilding, and he served as assistant poetry editor at the Black Warrior Review (CV). He was trained as a poet first.

The games career starts with LegendMUD in 1993, then Origin Systems from 1995—designer and creative lead, later lead designer, on Ultima Online. In 2000 he moved to Sony Online Entertainment as creative director of Star Wars Galaxies, managing a design team that peaked above one hundred, and served as the company's Chief Creative Officer from 2003 to 2006. In 2006 he founded Metaplace, a platform for user-created worlds that at its peak ran over 70,000 of them; it was sold to Playdom/Disney in 2010, and he left in 2013.

He founded Playable Worlds in 2018 and, as of 2026, is its CEO and creative director, building the sandbox MMO Stars Reach. His books are A Theory of Fun for Game Design (2004; second edition 2013), the poetry collection Sunday Poems (2015), and the essay collection Postmortems (2018). He is also a musician with an album, After the Flood, and keeps galleries of pencil and pen-and-ink drawings on his site. That a designer, theorist, poet, musician and illustrator share one body matters for everything below.

Philosophy: fun is progress on prediction

His central claim has barely moved in twenty years. In "An atomic theory of fun game design," written in 2004, he already states it flatly: "I've reduced it down to a cognitive challenge, the notion that fun is the feeling you get when you are exercising your brain by solving a cognitive puzzle" (written 2004, reposted 2012). In the same paragraph he adds that victory "usually comes from identifying the pattern, then correctly executing on some action that the pattern does not account for."

In "Game design is simple, actually" (2025) the vocabulary has been refreshed but the skeleton is identical: "In Theory of Fun I basically asserted that the useful bit for game designers was 'mastery of problems.'" And the section closes: "Bottom line: fun is basically about making progress on prediction" (his blog, 2025).

The line most readers skip, I think, is the qualifier attached to it: "Fun often shows up after an activity." That sentence exists to cut the naive equation of fun with moment-to-moment pleasantness. If the theory's centre of gravity is the "that was great" you feel on the far side of a hard passage, then bitterness along the way stops being a defect. For a puzzle designer that is an unusually practical claim.

Obsessions: the moment a game is solved, it becomes a puzzle

The boundary between game and puzzle is the theme he keeps returning to. Writing about the New York Times' Connections in 2023: "The main difference between a game and a puzzle is that a puzzle has one real solution, an optimal way through the challenge." Then he immediately dissolves his own line: "This means that really, puzzles and games are terms that are matters of degree, not kind" (his blog, 2023). In the same piece he recalls that in A Theory of Fun he termed games "a series of puzzles."

From that gradient comes his sharpest formulation: "It happens quite often that we used to think something was a game, and it turned out it was actually a puzzle. Mathematicians call that 'solving the game'" (his blog, 2025). Tic-tac-toe and Connect Four fall to the puzzle side once solved. So for him puzzles and games are not opposites; they are the same object before and after being solved. Which is why his measure of depth never changes: "Bottom line: the more uncertainty, indeterminacy, ambiguity in your game, the more depth it will have."

His other obsession is refusing to over-promise for his own theory. Closing the 2004 "atom" checklist he writes: "Is this an algorithm for fun? No, but it's a useful tool for checking on the absence of fun, in that you can identify systems that fail to meet all the criteria" (written 2004). Not a recipe for making it—an instrument for finding where it is missing. When he laid out a four-layer model in 2006 (mechanics, statistical content, metaphor and presentation, narrative), only one layer was non-negotiable: "The one that feels irreducible to me, of course, is the core mechanic. That is what makes this medium be games, and not something else" (his blog, 2006).

Failures, and how he absorbed them

He has a habit of taking failure in the first person. Closing his long 2015 re-examination of Star Wars Galaxies, he writes: "in that sense, SWG didn't fail. I failed it" (his blog, 2015). The causes he names are his own decisions: "This is my fault for failing to convey the vision adequately to everyone," and "I didn't really train a replacement." Even about the notorious NGE he opens a section titled "The error of good intentions" and defends the people involved: "They wanted to make the game more fun. This includes the Holocron drops, and the NGE, and CU too."

He says outright that this defeat produced A Theory of Fun: "it drove me to explore both plain old 'fun,' something that I felt I had failed at." The rest of the sentence matters more: "it made me try to be way more practical, and also made me reluctant to just manage, something which actually hurt Metaplace badly because I spent too much time on raw implementation." A lesson drawn from one failure becoming the cause of the next—few designers write that down so flatly. Metaplace's consumer service closed in December 2009, and his stated reason was blunt: "The reason? Well, it just hasn't gotten traction" (his blog, 2009).

His corrections about Ultima Online are equally concrete. In 2018, after the GDC postmortem, he revisited the decision to solve player-killing through design rather than switches: "I was wrong about that; it turns out it doesn't get cost-prohibitive until at least an order of magnitude more players than we had." And: "Business-wise, PKing was unquestionably a mistake. When Trammel went in, the population roughly doubled; our churn rate fell massively" (his blog, 2018). Back in 2000 he had already written, "One of my biggest disappointments in UO is that we never found a way to have the good guys win" (UO postmortem, written 2000).

One correction for the record. A line beginning "I regret lobbying for the combat upgrade" circulates online as a Koster quote. It is not in his 2011 post; it was left in that post's comment thread by a reader using the handle Jebadiah (comment thread, 24 June 2011). Without the habit of checking where a quotation actually sits, I would have been fooled myself.

Dilemmas: is depth ever rewarded?

The postmortem he wrote around leaving UO in 2000 has a section headed "Things that I have changed my mind on." Its first item: "I used to think that a richer, more challenging game would be rewarded. I am no longer sure that is the case." He pushes further: "most people want mere entertainment, stuff that is easy to cope with. Stuff that doesn't make them ask questions of themselves" (UO postmortem, written 2000).

But he does not get off there. On the option of simply entertaining the agreeable majority, he writes in the same passage that it "may be better business or may be enough to bring in the money, [but] it's not enough to make me feel like this is a field worth being in." Between commerce and authorship he seems to choose not a compromise but the discomfort.

A second dilemma sits between designer intent and player freedom. Explaining UO's approach to ethical behaviour: "We wanted to challenge players to act ethically, in the spirit of the Ultimas previous, without making it a set of quests that would be 'gamed'." Bolt the intent to mechanisms and you will achieve it—and the meaning drains out the moment you do. That is the spot where he keeps stopping.

In 2018 he extended the same conflict beyond UO. On failing to pursue distributed, player-run handling of trouble: "So I wish that path of development, figuring out ways to handle the issues in a distributed way, had been pursued more, our Internet today might not be nearly as horrible" (his blog, 2018). The man who wrote in 2000 that richer and harder work goes unrewarded is, eighteen years later, describing the bill for that unrewarded road arriving at society's door. I cannot stop reading those two statements side by side.

Influences he names himself

In a 2000 VaultNetwork interview he prefaces a reading list with "I do a lot of reading on things that I think can help me in game design," then names them: Scott McCloud's Understanding Comics and Reinventing Comics ("these books really help explain how players relate to the environment, and in the case of online, to other players"), Chris Crawford's The Art of Computer Game Design, Edward Tufte, Hofstadter's Gödel, Escher, Bach ("I loved [it] when I read it years back and it gave me many design ideas"), Amy Jo Kim's Community Building on the Web, Jared Diamond's Guns, Germs, and Steel, Julian Dibbell's My Tiny Life, Janet Murray's Hamlet on the Holodeck, and Howard Rheingold. He also credits the MUD-Dev mailing list: "That's where a lot of my writings on my website come from" (VaultNetwork interview, 2000).

More surprising is the literary origin he described in his 2006 Project Horseshoe keynote: "I was, to my everlasting dismay, mostly studying modernist literature while getting a creative writing degree. I really hated modernism, which was why I took so many courses in it, so I could argue with the professors." He then ports that training straight into design: "This is why I say that Minesweeper is actually an Impressionist game, because you never actually see it, you see everything around it and how it's been bumped and jostled." The same talk runs through Woolf, Stein, Picasso, Mondrian, Cage's 4'33", and Lorca's duende—"that elusive moment of transgression of the unreal into the real" (Project Horseshoe transcript, 2006).

He is explicit about his elders too: "Anytime I sit and talk with Randy Farmer, or John Taylor, or Richard Bartle, I'm the young pup among them" (interview by Celia Pearce, Game Studies 2005). On the reading behind A Theory of Fun: "I'd been reading a lot of nonfiction pop psychology books actually. They were leading me out of pop psychology and into the actual papers, particularly cognitive psychology and cognitive science… Obviously there were things like flow theory and the like" (Game Thinking podcast ep. 303, 2019). In the same episode the contemporaries he names are Dan Cook, Zach Gage and Michael Brough—all of whom we have covered here.

One caution. It is often said that the cartoon-plus-text format of A Theory of Fun derives from McCloud. I could not find a text in which Koster says so. What he does say is that he made his 2003 conference slides cartoon-like "because I had grown increasingly bored with bullet points" (his blog, 2006). I will not connect that lineage on speculation.

Kizuki's reading

What follows is my (Kizuki's) reading, not something he has said.—Koster writes that a game becomes a puzzle once solved, yet his own four decades read like a collection of counterexamples to that theorem. Player-killing, succession of vision, platform traction: none of the failures he discusses in public are problems with a single solution you can go and find. They are problems whose answers keep moving. And still, each time, he takes them in the first person—"I failed it"—and comes back a decade later to file a correction.

I read this as the habit of someone trained as a poet. In revising a poem there is nobody else responsible. His insistence on limiting his own theory—not an algorithm for producing fun, but an instrument for detecting its absence—looks to me like the same posture inverted: he will not promise to make it, but he will look for what is missing and sign his name to the finding. Which is why his 2025 definition, that fun is progress on prediction, reads to me as both the self-defence and the severe self-criticism of a man who spent twenty years getting his predictions wrong.

Where to start

The order to approach him is straightforward. Start with A Theory of Fun for Game Design (2004; second edition 2013). Then, especially if you play puzzles daily, read "Why NYT's Connections makes you feel bad" (2023) (his blog)—it tells you where the thing you solve every morning sits inside his theory. For the designer rather than the theorist, go to the 2015 Star Wars Galaxies re-examination and the 2018 Ultima Online postmortem. The most compact recent statement of the theory is "Game design is simple, actually" (2025) (his blog). His GDC 2024 talk, "Revisiting Fun: 20 Years of A Theory of Fun," is free on the GDC Vault (GDC Vault).

For neighbouring writers, we have covered Daniel Cook, Zach Gage and Michael Brough—all contemporaries Koster names himself. If you want theory brought down to implementation, Frank Lantz; if you want to think from the side of things that can be solved, Scott Kim.

Sources

Primary sources consulted for this article (every quotation was checked against the source text):

・Raph Koster, “Game design is simple, actually”, his blog, 2025-11-03

・Raph Koster, “Why NYT’s Connections makes you feel bad”, his blog, 2023-09-02

・Raph Koster, “An atomic theory of fun game design” (written 2004; reposted 2012-01-24)

・Raph Koster, “A bit on how I think games work”, his blog, 2006-03-05

・Raph Koster, “Did Star Wars Galaxies Fail?”, his blog, 2015-04-27

・Raph Koster, “SWG is shutting down”, his blog, 2011-06-24 (comment thread also checked for a misattributed quote)

・Raph Koster, “A UO postmortem of sorts” (written June 2000)

・Raph Koster, “UO postmortem from GDC2018”, his blog, 2018-03-28

・Raph Koster, “Metaplace.com closing”, his blog, 2009-12-21

・Raph Koster, “A Theory of Fun milestone and postmortem”, his blog, 2006-03-06

・Raph Koster, “Project Horseshoe: Influences”, keynote transcript, 2006-11-10

・VaultNetwork “Designer Dragon” interview by Jonric, 2000

・Celia Pearce, “A Conversation with Raph Koster”, Game Studies vol.5 no.1 (Oct 2005; interview 2004-03-22)

・Game Thinking podcast ep.303, “Raph Koster: The Nature of Gaming” (Amy Jo Kim, 2019-12-30; full transcript)

・Raph Koster, “About Raph” (biography)

・Raph Koster, CV (education, career, publications, awards)

・GDC 2024, “Revisiting Fun: 20 Years of A Theory of Fun” (free on the GDC Vault)