Social Systems

Honor 2.0 shipped in July 2017, patch 7.13. The core bet was scarcity: one honor per game, cast at the end-of-game screen, accruing to a visible level on your profile. Riot’s public framing put the entire theory in one sentence: “Honoring one person per game prompts careful consideration of who you give your vote to” (Ask Riot, July 2017). Scarce recognition is sincere recognition. And it landed! Players gave more than 56 million honors in the first two weeks.

It was also coupled to loot. Honor progress fed key fragments and capsules, because rewards drive adoption, and adoption is how new systems survive. That coupling is where the failure got planted. Goodhart’s law is older than this industry: when a measure becomes a target, it stops being a measure. Once honor paid out in currency, it stopped measuring honor and started measuring attendance.

I want to be honest here: I’m not claiming I saw the drift coming. The system I worked on shipped the coupled version, and it took seven years of live data to see the results. Riot’s behavioral systems team wrote the postmortem in August 2024:

In that same update they reversed the 2017 design’s two most contested calls: opponent honor came back, and so did honoring more than one player. Seven years is a pretty slow feedback loop, but it’s still a feedback loop.

Here’s the principle I keep: recognition pays out in visibility and identity, and in nothing else. Once a badge can be melted into currency, that’s all it is: currency. If your sportsmanship system has an exchange rate, you’ve built a second economy instead of a culture, and economies get optimized.

The empirical backbone is Jeffrey Lin’s GDC-era research at Riot: roughly 1% of players are consistently toxic, and they produce only about 5% of the toxicity. The bulk comes from ordinary players having a bad day. If you spend your whole behavior budget hunting “toxic players,” you’ve targeted a rounding error and ignored the population that actually sets your game’s tone.

Lin’s own synthesis was that there is no silver bullet and you need both. The record since then backs him up on every count.

Don’t misread the stick swing of 2026, either. In patch 26.5 this past March, League began issuing gameplay bans (not chat restrictions) for severely abusive comms. Riot’s line: “If someone can’t play League without egregiously harassing other players, they aren’t going to be allowed to play it.” Some commentary read that as the nudge era admitting defeat. I read it as studios admitting they skipped the boundary half of the job. Carrots were never supposed to handle the 1%. That was always the stick’s work, and the industry let the stick rust for a decade because punishing paying customers feels bad in a quarterly review. The Call of Duty voice moderation data made the gap impossible to ignore: once proactive detection went live in 2023, it turned out 79% of daily code-of-conduct violators had never been reported by a single player. Report-driven enforcement was never really a boundary, it was a suggestion box.

Both halves were always the job. Recognition shapes what the silent majority believes is normal, and enforcement proves the floor is real. Carrots without sticks breed cynicism, sticks without carrots breed silence, and neither one gets you a community.

Discord won the guild hall. Voice, LFG, scheduling, the clan-drama backchannel: that war is over, and rebuilding those features in-game is usually vanity work. Discord also publishes the number that should be taped to every live-service roadmap: “When even one friend is present on Discord, players on PC spend a median of 6x longer in a game – with three friends, it’s 8x” (Discord, GDC 2026). It’s a little funny that the strongest published argument for social investment comes from the company that captured the social layer games gave away.

But there’s a line, and the line is identity. Profiles, reputation, history, honor, the visible record that you were here and you were good: all of that lives in your game, or your game doesn’t get the benefit. Raph Koster called the migration decades ago, when he noticed real social bonds drift out of the fictional world toward real channels. That’s fine! Let the bonds live on Discord. The resume stays home, because the resume is the retention moat. It’s the one part of a player’s social life that only your game can hold.

Match history is the case I keep returning to (chapter 05 covers its metagame half). A shareable record of every game turns sessions into a body of work, and the third-party ecosystem built on top of it did more for player identity than any first-party feature could have. The tracker sites outgrew the first-party client because players treat that record as theirs. The standing expectation now is that your history belongs to you, wherever you earned it. Identity artifacts outlive roadmaps, so plan for that (or plan to apologize).

And display turns out to be the easy identity problem. The hard one is recurrence. Daniel Cook put it well in 2017: “The hard part is getting people who played together once, to play together another time.” Matchmade games are repeat-encounter deserts, and clans, clubs, and persistent squads exist to fix the re-encounter rate. Most of them fail anyway, because they get funded like launch features instead of infrastructure. Which brings me to Black Ops 4.

This is the purest expression of this site’s thesis. Systems are the message, and the scoreboard is the system at its most legible: every player stands in front of it after every match and reads what counted. Put K/D at the top and players will leave a bleeding teammate to chase a kill, because you told them, in your own handwriting, that the kill is what counts and the teammate isn’t on the board. Show revives, assists, and time-on-objective, and the same lobby plays differently. The players didn’t change, the values statement did.

The evidence here runs pretty deep. Blizzard made positive behavior visible with Overwatch’s endorsement system and reported a 40% reduction in matches with disruptive behavior at GDC 2019. Riot’s Player Dynamics discipline (the successor to the Lin-era behavior team) put the underlying mechanism plainly: “We tend to attribute bad behavior in others to personal intent, when it is often due to situational constraints.” The scoreboard is a situational constraint. So is the comms wheel.

Which brings up pings. Chapter 02 tells the ping story as an accessibility dividend; here it matters as comms architecture, a full coordination channel that is structurally incapable of carrying harassment. I think it’s the most important social-systems artifact of its decade, and I say that with the strongest counterargument in full view:

Cook is right about the cost. Harassment-resistant channels are also friendship-resistant; a lobby that can’t insult each other can’t really befriend each other either. Where I land is that the constrained channel is the floor, not the ceiling. Ship the channel that’s safe among strangers, then open up richer channels as trust accumulates: squad, then friend, then regular. Almost nobody has shipped that ladder yet, and pings-versus-chat keeps getting argued as a binary when it’s really a gradient with exactly one sensible direction. In the end, your comms design is part of the scoreboard. What can be said and what gets counted are answers to the same question: who do you want your players to be?

Build the end-of-match screen yourself. Choose what it displays (kills, damage, assists, revives, honor) and watch a simulated lobby reorganize its behavior around whatever you decided to count. There’s a loot-coupling toggle too, so you can speedrun the Honor 2.0 lesson in about ten seconds.