DEVBro
← Writing
Essay · 23 min read · Jun 17, 2026

What makes a game a good game?

I've been a gamer my whole life, and I've tried to make my own game more than once — always quitting. This is a summary of my research into what actually separates the games that stick.

D DEVBro
game-designplayer-psychologyindie-dev

I’ve been a gamer my whole life, and like a lot of people, I’ve tried to make my own game more than once. I always ended up quitting, for one reason or another.

Even now, with generative AI able to handle the code for me and sometimes the assets too, I still can’t build a game I actually like. So this article is a summary of my research, an attempt to work out how you make a good one in the first place.

All the sources I used are listed at the bottom of the page.

Three geometric shapes — a circle, a triangle and a square — interlocking into one balanced form.

1. The psychology underneath every good game

A staircase climbing a narrow diagonal channel between a dark zone above and a coral zone below.

The first thing I learned is that “good” isn’t really a mystery. Strip away the genre, the budget, and the art style, and most great games pull the same few psychological levers. Once you can see them, you start seeing them everywhere.

The clearest starting point is Self-Determination Theory, from psychologists Edward Deci and Richard Ryan, applied to games by Ryan, Rigby, and Przybylski in 2006. It says people have three basic needs, and games happen to be very good at feeding all of them. There’s competence, the feeling of getting better and beating a challenge. There’s autonomy, the feeling that you’re making your own choices instead of being dragged down a rail. And there’s relatedness, the feeling of being connected to other people.

This is the part that changed how I think about UX. Clean controls and a readable screen aren’t decoration. They’re the pipe that competence and autonomy flow through. Clunky controls don’t just annoy people, they physically block the reward the game is supposed to hand you.

The second idea is flow, from Mihaly Csikszentmihalyi. Flow is that state where you’re so absorbed the clock disappears. The condition for it is easy to say and hard to hit: the challenge has to sit just slightly above your current skill. Too hard and you get anxiety. Too easy and you get bored. Since your skill keeps climbing, the difficulty has to climb with it. I’ve since noticed I make the rookie mistake all the time, softening things too much after watching a friend struggle, which quietly slides the game into the boredom zone.

The third is the MDA framework (Mechanics, Dynamics, Aesthetics) from Hunicke, LeBlanc, and Zubek. You build the mechanics, the rules and actions. Out of those emerge the dynamics, how players actually behave. And out of those come the aesthetics, what players end up feeling. The thing I keep reminding myself is that you only ever touch the first layer directly. Everything else emerges, which is why one small rule change can blow up into completely different behavior. Players, meanwhile, meet the game backwards: they feel the emotion first and notice the rules later.

No single game pleases everyone, and the player-type models explain why. Richard Bartle’s old split was Achievers, Explorers, Socializers, Killers. The newer, data-heavy version from Quantic Foundry, built on surveys of nearly two million players, sorts motivation into action, social, mastery, achievement, immersion, and creativity. Someone who lives for competition and someone who wants to wander and relax are chasing opposite feelings. Usually you have to pick.

All of this gets compressed into one old piece of advice, Bushnell’s Law, named after Atari’s Nolan Bushnell: the best games are easy to learn and hard to master, and they should reward your first minute as much as your hundredth. It’s basically Self-Determination Theory and flow fused into a slogan. Early ease gives you instant competence, a high skill ceiling keeps the flow channel open for hundreds of hours. (Designer Ian Bogost has argued the “easy to learn” half is overrated, since plenty of great games like Spelunky aren’t easy at all. What matters more, he says, is the second half: giving people a reason to come back.)

2. Accessibility and UX, or the two kinds of difficulty

That competence need leads straight into a question I used to get wrong: should a good game be accessible? Yes, but accessible doesn’t mean easy. The trick is to separate two different things that get lumped together.

One is challenge difficulty, the intentional kind. That’s the point of the game. The other is access difficulty, the friction nobody asked for: confusing menus, unreadable icons, awkward controls, a tutorial that teaches nothing. Good design kills the second and protects the first. Dark Souls is brutally hard, but its controls are crisp and fair. The difficulty is all challenge, almost no friction.

A few habits show up in games that get this right. You can read what matters at a glance, even mid-fight. Icons are consistent, so a gear always means settings and you never relearn a symbol. Every action gets immediate feedback, visual and audio. The game teaches one thing at a time, ideally through play instead of walls of text (Portal 2 is the gold standard here). And there are real accessibility options, colorblind modes, text scaling, remappable controls, which quietly widen your audience for very little cost. The best UX is the kind nobody notices, because everything works the way they expected.

3. Why Geometry Dash refuses to die

Theory is one thing. I wanted a real example of these ideas surviving for a decade, and Geometry Dash is almost too perfect.

It came out in 2013, made by one Swedish developer, Robert Topala (RobTop). More than ten years later it’s still pulling tens of millions of monthly players on mobile and breaking concurrent-player records on PC well into 2026, with basically no marketing budget. So what’s holding it up?

The core loop is Bushnell’s Law in its purest form. One button. Anyone starts in seconds, but a hard level can take thousands of attempts to clear. When you die you know exactly why, it was your mistake, and the restart is instant, so failure turns into “one more try” instead of rage-quitting. People literally brag about how many attempts a level cost them.

There’s a detail about the music I had wrong, and it’s worth getting right if you ever want to build something nearby. People call it a “rhythm platformer,” and that’s roughly true. Music was there from the very beginning. RobTop started prototyping in 2012, inspired by browser games like The Impossible Game and Bit.Trip Runner, and his two big additions were syncing obstacles to the music and adding a level editor. The official levels were each built around an electronic track, and the music works as a timing cue, most players memorize a level through the song rather than the visuals.

But it is not a rhythm game in the strict sense. In Guitar Hero or Crypt of the NecroDancer, your inputs are scored on how well they land on the beat. In Geometry Dash the obstacles are fixed in space, you react to the level, and the music is a synchronized guide and a mood, not the scored mechanic. You could play with the sound off and still win. It’d just be much harder. So it’s a precision platformer paced by music, not a beat-matching game. That distinction matters later.

The real reason nobody has dethroned it is the level editor. Geometry Dash isn’t only a game, it’s a no-code tool that has produced millions of community levels. That ecosystem feeds itself: players make content faster than RobTop ever could, which means an endless supply of free, fresh levels, plus a whole culture of difficulty rankings, legendary hand-made levels, verification videos, and awards. A competitor can clone the game in a weekend. What they can’t clone is thirteen years of accumulated community content and the people who keep making it. That’s the moat.

4. Why some online communities are warm and others are poison

Three small emblems in a row: a pair of opposed arrows, an inward spiral, and a scatter of dots.

Geometry Dash is a solo experience, so the moment I started thinking about playing with other people, a different problem showed up, and it’s the one that fascinates me most.

Here’s the contradiction that got me started. In Deep Rock Galactic, a co-op game, one player’s mistake can wipe the whole team, yet its community is famous for being one of the kindest in gaming. Meanwhile League of Legends, CS2, Dota 2, and Valorant are infamous for toxicity. Same kinds of people, wildly different behavior. Why?

The answer I kept arriving at is that toxicity comes from the design, not from bad people. Players behave the way the system trains them to. And the cleanest way to see it is to ask one question: when you lose, where can the blame go?

Think about three competitive games. In League of Legends, you blame your teammates. The blame has a human target, which is what makes it toxic. The loop is well documented: frustration, blame a teammate, flame them, they tune out and flame back, everyone spirals down together. There’s even a hidden ingredient. A forty-minute match you knew you’d lose by minute ten gives you thirty minutes of compounding anger. Long matches breed more poison than short ones.

In chess, you blame yourself. No teammates, no luck, so a loss is unambiguously yours. That kills interpersonal toxicity completely, there’s nobody to scapegoat, but it dumps all the frustration inward, which is rough on the ego. Tellingly, the main toxicity in online chess is accusing your opponent of cheating, the only blame-shift left when there’s no luck and no teammate to point at.

In Hearthstone, you blame the dice. Randomness hands you a target that isn’t a person: “I didn’t lose, I got unlucky.” That protects your ego without creating a human victim, which is why card-game communities tend to be calmer while staying just as sticky. The cost is a different complaint, the feeling that the game is rigged.

So the available blame target sets two dials at once, toxicity and stickiness. Teammates and randomness both cushion the ego, which keeps people playing, but teammates create human victims and randomness creates a faceless unfairness. Chess cushions nothing, which makes it the purest competition and the harshest.

That raises an obvious question I had to sit with: if these games make people miserable, why are they so popular? Because misery and fun aren’t opposites here. A win only matters because losing hurt. The uncertainty itself is the hook, an unpredictable reward is the strongest one there is, the same mechanism a slot machine runs on. Ranked ladders hand you a precise public number that proves your skill, and the sting of dropping is the price of that number meaning anything. “One more game” is loss aversion plus the belief that the next match is the one where you climb. And the ego buffer of “my team lost it for me” fuels the toxicity and the retention at the same time. The skill ceiling never arrives, so there’s always a next rank.

Deep Rock Galactic does the opposite on purpose. Everyone wins or loses together, the four classes only succeed as a team, and the game never posts a humiliating individual scoreboard telling you who carried. It gives you a cheap, almost ritual way to be nice, the “Rock and Stone” cheer mapped to its own button, which turns encouragement into a one-tap reflex. It uses pings instead of open chat, which removes a harassment channel. And the studio actually moderates and welcomes newcomers.

The research lines up with all this. An experiment by Ewoldsen and colleagues found people who played cooperatively behaved more cooperatively afterward than people who’d competed, the social context shaped their behavior more than the game’s content did. Riot’s own large-scale work on League of Legends reached a result I didn’t expect: most toxic messages didn’t come from a small group of villains, but from ordinary, usually-positive players having a bad day. (Those numbers are Riot’s own, so take them as self-reported.) Their fix wasn’t only punishment, it was reshaping the culture.

5. Why multiplayer is so much harder than it looks

A loose constellation of dots, some pulled into a dense cluster and others drifting toward the edges.

That toxicity problem is really just one face of a bigger truth I had to accept: building a multiplayer game is a different animal from building a solo one, and no amount of AI changes that.

The biggest reason is cheating. In a solo game, cheating only affects the cheater, it’s sometimes even a feature. In a competitive online game, one cheater ruins it for everyone and breaks the fairness contract that gave the game meaning in the first place. Worse, honest players who run into cheaters lose trust, leave, and stop spending. Cheating can kill an entire game, not just a match. That forces heavy design decisions and an endless cat-and-mouse cycle you can never really win.

Then there’s balance. A solo game can be balanced once, against predictable AI. A multiplayer game keeps evolving as players discover dominant strategies, so you’re rebalancing forever, and every change makes part of your audience unhappy. On top of that, a multiplayer game is a service, not a product. Solo games ship and you’re done. Multiplayer means servers, seasons, events, moderation, and support, indefinitely. And you need a crowd just to function, because matchmaking needs a critical mass of players. Below that line, queues drag, matches get lopsided, and the game dies even if it’s excellent. That cold-start problem is the one that scares me most as a solo dev.

But here’s the part that gave me hope, and it matters for the rest of this article. “Multiplayer” isn’t all or nothing. The critical-mass trap is only fatal when other players are the whole point. There’s a big middle ground where you borrow the feeling of company without the fragility. Asynchronous multiplayer is one route, leaderboards, ghost runs, the bloodstains and messages in Dark Souls, where you feel other people without anyone needing to be online at the same time. Optional co-op is another, Deep Rock Galactic plays solo with an AI drone filling in for teammates. And persistent worlds like idle browser MMOs treat other players as flavor, an economy and a backdrop, never something you depend on to play.

The principle I took from this: make the game complete and satisfying for one player first, then let the social layer enrich it as the population grows. Critical mass becomes a bonus instead of a life-support machine.

6. What players actually want now

Knowing how to build the thing is only half of it. I also had to update my sense of what people want, because it has shifted, and the old assumptions I grew up with are partly out of date.

The phrase that kept coming up in my reading was respect for the player’s time. The old model of grinding for hours has lost ground. Vampire Survivors is the poster child. Runs cap out around fifteen to thirty minutes, attacking is automatic so you only handle positioning, and no run is ever wasted because everything feeds your permanent progress. It became a massive hit and spawned a whole sub-genre, at one point beating much bigger-budget games on concurrent players.

The flip side is live-service fatigue. Daily logins, battle passes, limited-time cosmetics, they turn play into a chore. The day “I want to play” becomes “I have to log in,” the fun curdles and burnout follows. A lot of live-service games shut down recently as the market crowded and a few giants soaked up everyone’s hours. Some designs have pushed back, like battle passes that never expire so you finish them at your own pace.

Alongside that, cozy and shorter games are thriving. As the audience ages into jobs and kids, a complete four-to-six-hour game can feel more valuable than an endless one. There’s a generational split too. Players raised on MMOs once tolerated grinding as the price of achievement, while a lot of younger players, and plenty of older ones, now reject anything that feels like a second job.

7. Where people play changes what they want

The same idea can succeed or fail depending on the device, which is something I’d never really thought through.

On mobile, sessions are short and frequent, controls are thumb-sized, and you’re playing on the move. Money mostly comes from free-to-play and gacha, the random-draw mechanic where a few big spenders subsidize a large free audience, and regulators are circling those mechanics now. Worth noting: the survivors-like genre was actually born on mobile, yet most of today’s hits are PC-first, which is a gap worth thinking about. On PC, sessions run long, mouse and keyboard give you precision, and players tolerate dense, complex interfaces, strategy games, sims, MMOs. The usual model is buy-once, often with DLC. On console, you’re in the living room with a controller, leaning back, and the screen has to read clearly from across the room. Historically premium, with extras layered on rather than forming the core.

The deeper point is that input and context shape the design itself. A precise shooter is a different game on a mouse, a controller, and a touchscreen. What works in a five-minute subway session falls apart in a three-hour couch session, and the reverse. And since nearly half of players now move across several platforms, consistent controls and synced progress matter more than they used to.

8. The hardest part is getting anyone to look

One very tall bar on the left, then a long low tail of tiny coral bars trailing to the right.

Here’s the uncomfortable lesson, and the one that probably explains why my own attempts went nowhere. For most indie games today, the hard part isn’t making the game. It’s getting anyone to see it.

The scale is rough. Around twenty thousand games hit Steam in 2025, roughly sixty a day, almost all of them indie. The results follow a brutal power law, a tiny fraction of games take most of the money. By one widely-cited analysis (Chris Zukowski’s annual review), only about 3% of 2025 releases crossed even a thousand reviews, and nearly half got fewer than ten. Quality isn’t enough. Plenty of good games sink without a ripple.

The tactics experienced indie marketers agree on are fairly consistent. Wishlists are the currency, because at launch Steam emails and notifies everyone who wishlisted, the single most valuable free marketing moment you get. Put your Steam page up early, since pages live six-plus months before launch dramatically outperform last-minute ones and old wishlists don’t expire. Your trailer needs to land its hook in the first few seconds, using real gameplay, not cinematics. A professional capsule, the thumbnail, is often the one thing worth paying for, since it’s the first thing anyone sees. And a demo opens two doors at once, festivals like Steam Next Fest and streamers, because a streamer can’t play a screenshot but they can play a demo.

A good case study is Megabonk, a 3D take on the survivors-like made by one developer (“Vedinad”). It sold over a million copies in two weeks in 2025 and peaked above 117,000 concurrent players, more than Vampire Survivors itself. How? Partly a head start I should be honest about, the developer looks to be an established game-dev YouTuber with an existing audience, which not everyone has. But the method still transfers. He built in public, with regular funny devlogs and an active Discord that turned viewers into invested fans before launch. He had a hook you could repeat in one sentence, “Vampire Survivors but in 3D.” He gave it a memorable identity, a deliberately rough PS1 look and a joke that it’ll “run on a toaster,” which doubles as an accessibility pitch. And the game is streamer-bait by design, short runs and absurd broken-build moments that spread on their own. Whether or not you start with an audience, the lesson is that you build the crowd during development, not on launch day.

9. So is the survivors-like still worth making?

I keep coming back to Megabonk for a reason. It’s a survivors-like, and that genre is exactly where a solo dev like me tends to land, because it quietly dodges almost every hard problem in this article. There’s no critical-mass trap, no live-service treadmill, no cheating to police, and the short, complete runs already respect the player’s time. It’s the most solo-friendly template there is. So it makes a good test case for everything above: if I understand why it works and why it’s hard, I understand the rest.

The catch is that every other solo dev ran the same math, which is why the genre is drowning in clones. The difficulty didn’t disappear, it just moved, from “get players to coordinate” to “stand out in a crowd.”

The genre is healthier than ever, in 2026 Valve even gave it an official “Bullet Heaven” tag, which only happens once something is undeniably big. Solo hits still break through. But the bar has jumped. As marketing analysts point out, it’s no longer a quick win, so much of it looks and plays the same that the most polished version wins and players spot lazy reskins instantly.

What’s worked recently is moving one specific thing while keeping the addictive core intact. Megabonk went 3D and added real movement skill, jumping and kiting, which answers the common complaint that the genre is too passive. Others mash it up with another genre’s hook, or pile on build depth that keeps theorycrafters busy for hundreds of hours, or pick a visual style nobody else has claimed. Co-op is on the table too now, The Spell Brigade turns friendly fire into a feature, and Vampire Survivors itself added co-op, though as section 5 warned, that drags the whole multiplayer tax back in.

If I were betting on an under-explored angle, and these are guesses I’d validate by searching Steam tags first, a few look open. A truly rhythmic one, where attacks land on the beat the way Crypt of the NecroDancer does, would move the scored mechanic itself, a sharper move than even Geometry Dash makes. A cozy one would cross two strong trends in a nearly empty space, since the genre is overwhelmingly dark and violent. A narrative one is wide open, since Hades proved narrative roguelites work and almost no survivors-like has a story. An asynchronous-social one would give you that feeling of company without the netcode cost. Or you could invert the whole thing and play as the horde. The mobile angle is interesting too, building portrait, one-thumb play with honest monetization, reclaiming the platform the genre came from.

The rule I keep landing on is that the angle matters less than being the most polished version of one clear idea, and proving it’s fun fast, before sinking months into it.

What I’m taking away

If I boil my notes down to the things I’ll actually try to remember:

Learn the frameworks before the tricks. If I can’t say which need my core loop feeds in the first minute, it isn’t ready. Decide early whether it’s solo, co-op, or competitive, because that single choice sets most of the difficulty. Solo and co-op are far easier to build and sustain, and they tend to grow kinder communities. If it is multiplayer, design the community like a mechanic, with shared goals, hidden individual stats, cheap ways to be nice, and real moderation, because toxicity comes from the design. Pick the platform before the design, since input and context shape everything. Treat getting noticed as the main constraint from day one, with an early Steam page, a clear hook, a good capsule, a demo, and wishlists above all. And respect the player’s time, with complete sessions and no manipulative pressure to log in.

A good game, in the end, is one that meets a real human need cleanly. It gives you mastery, or choice, or connection, without making you fight the interface to get there. Everything above is just a different angle on that one idea. Now I just have to go make one.

Sources

Grouped by section. Where a figure comes from a company about itself (like Riot’s moderation stats), I’ve flagged it in the text.

Psychology and design theory

Accessibility and UX

Geometry Dash

Toxicity and community

Cheating and multiplayer design

What players want now

Platforms and monetization

Marketing and discoverability

Megabonk

The Spell Brigade (co-op survivors-like)

One note on the numbers: some industry figures (Steam release counts, player counts, revenue shares) vary by source and method, and a few are projections rather than settled facts. Treat the orders of magnitude as solid and the exact numbers as rough.

#game-design#player-psychology#indie-dev#solo-dev#marketing
← Older
Starting a notebook in public
Newer →
Dashline: the first prototype, against my own advice