Competitive Systems

Modern ranked runs on two numbers. League canonized the split in 2013: a hidden matchmaking rating that builds fair matches, and a visible rank (tiers, divisions, LP) that delivers the feeling of a climb. Purists call this corrupt. I don’t. A ladder that is pure measurement is a medical chart, and most players didn’t queue up for a diagnosis. The hybrid is fine (arguably it’s the whole job). What’s not fine is lying about which number is real.

Season 17 of Apex is the public record on what happens when the disguise gets caught. In May 2023 the ranked system moved to a hidden MMR with visible Ladder Points layered on top. The stated rationale, from the dev blog, was that “RP was not a true representation of a player’s skill.” Which was true! It just wasn’t the right fix. The visible rank became, functionally, a progress bar decorating an invisible rating, the LP tuning ran hot, and Masters inflated past 100,000 players inside a month (community trackers counted more players in Masters than in Diamond 1). By the July 2023 ranked dev blog the team was saying it plainly: “Arsenal’s LP tuning was intended to be generous, but not this generous,” and the tuning was “causing a series of cascading issues with the rest of the Ranked system.” Season 20, February 2024, reverted to the visible-RP architecture from Season 13.

Look at what players actually revolted against. It wasn’t the hybrid; they’d been climbing a hybrid since 2019 without complaint. The revolt came at the moment the two numbers were caught contradicting each other: a shiny rank on screen, and a lobby full of opponents the hidden number knew were better. That’s not really rank inflation, it’s a contract violation, and the inflation is just the receipt.

And before anyone files this as a Respawn problem: Riot, fifteen-plus years into running the most-studied ladder on earth, hard-reset its apex tiers across six regions at the start of the 2026 season, opening with “We made a number of mistakes in how we configured the Ranked system for the 2026 Season start.” Nobody has solved this, which is exactly why the contract has to be explicit: this number builds your matches, this number marks your season, and here is how they relate. In my experience players can hold two numbers in their heads just fine. A disguise is the thing they can’t hold.

Sirlin’s line is aimed at players, and it’s earned. But flip it around and it points at us: a dishonest ladder handicaps players with designer-imposed rules the player knows nothing about. Sirlin’s player respects exactly and only what’s real. Hide the real rules and you’ve built a game that punishes the exact people who take it seriously.

When we launched Ranked Leagues, you paid to sit down: an RP entry cost (one to five points by tier at launch) that you earned back through kills, capped at five, and placement. Every queue was a small wager, and that was deliberate. Stakes are what separate a season from a playlist. Loss aversion isn’t a dirty trick when the player walked up to the table, read the price, and chose to play. The launch blog said the goal out loud (“Create a true measure of skill in Apex Legends”), and a measure with no downside isn’t much of a measure.

This is the Hextech trade from chapter 05 again: agency for cost. One trade, priced in the open, opted into. I’ve shipped that trade in loot, in ranked, and around commerce, and the line is the same everywhere: consent and clarity. The player has to know the price, and the price has to mean what it says. Loss aversion with consent is stakes; without consent it’s a trap.

The public record backs the distinction. The per-match wager survived everything: Season 13 hardened it with scaling entry costs and demotions, Season 17 abandoned it, Season 20 brought it back. Meanwhile, episodic friction died across the industry. Apex removed promotional trials in Season 20 because, in the team’s words, “they weren’t as fun or celebratory as we wanted them to be,” and League deleted promotion series outright in 2024 after more than a decade of manufactured drama. The lesson I take from all this: friction works when it’s continuous and legible (a price on every hand) and curdles when it’s episodic and binary, like a toll booth at the tier gate. Players find meaning in stakes they chose, and mostly just dread in tolls they hit.

In 2017, researchers at EA published a formal proof that fair matchmaking is not engagement-optimal: fairness is just one point in a design space whose real axis is retention. The paper is good math, and the data since says the machinery works. Activision’s 2024 matchmaking whitepapers A/B-tested loosening skill-based matchmaking and found that even “attempts to protect only the bottom 25% of players” still “had clear negative effects on player counts in two weeks, with increased quit rates, and reductions in total hours played.” Engagement optimization works, and that’s exactly why the line has to be hard. A match arranged to maximize your engagement isn’t a contest anymore. It’s a treatment plan with a scoreboard.

Players can’t audit the matchmaker, so they price in the worst case. Activision filed a patent in 2015 (granted in 2017) for matchmaking that could steer players toward microtransactions; combine that with the EOMM paper’s existence and you get a permanent folk theory. By August 2025, NetEase was compelled to put it on the record: “We want to be clear: Marvel Rivals does not use EOMM.” That’s a major studio formally disavowing a nine-year-old conference paper by name. Every queue in this industry is now poisoned by suspicion of techniques most studios swear they don’t run. You pay for papers other companies published. I wrote it in an old design outline years before any of this: lack of transparency leads to mistrust. I wasn’t being clairvoyant; that’s just how players work.

So name your objective function in public. Riot’s 2018 matchmaking retrospective named its triangle outright (fair matches, primary position, short queues), all three properties of the match itself, which is why that post aged well. Activision publishing actual A/B data in 2024 was the right move too; I’ll take receipts over a decade of vibes from any side of the SBMM war. And for what it’s worth, XDefiant marketed itself on no-SBMM casual play in May 2024 and was gone inside thirteen months. I won’t claim that closes the case, but I notice nobody is rushing to rerun the experiment.

Balance patches ship every couple of weeks, but the meta moves daily. A draft phase with real bans closes that gap: when something is overtuned, ten thousand lobbies can vote it off the island that same night, every night, until the hotfix lands. That isn’t a side effect, that’s meta governance, and you should design for it on purpose. I contributed to the Picks & Bans updates that brought the 10-ban system to League’s normal and pro play in 2017, and the stated goals in Riot’s dev post were governance goals: “to make champ select fairer for everyone, to give everyone more control over the ban phase, and if possible, to lower the time it takes to get to game.”

The interesting part was the fork. Pro play got sequenced split-draft: counter-bans mid-draft, mind games, broadcast drama. The ladder got ten simultaneous blind bans: individual agency at no added time cost. Riot’s Draggles called it “one of the only times we’ve ever made a fundamental difference between pro and regular play, and it’s not a decision we made lightly.” It wasn’t, and I think it was the right call. The constituencies were different, so the instruments diverged: pros needed strategy, ranked players needed a veto, and the broadcast needed theater. That’s what you get when you treat pick/ban as a governance instrument: one system, tuned three ways, serving all three.

The fork keeps widening. Fearless Draft (champions consumed across a series) went global across tier-1 leagues in 2025 and stayed for 2026, which makes pro League a meaningfully different drafting game from ladder League. I’m measured on this: divergence serves each constituency better, but the aspirational loop (watch the pros on Friday, queue the same game on Saturday) is a real asset, and every fork spends a little of it. So diverge deliberately, and know what it costs.

StarCraft II players coined “ladder anxiety” around 2010: the discovery that a perfectly accurate 1v1 ladder drives players away from the very mode built for them. Nearly every ranked feature invented since (placement matches, demotion shields, loss forgiveness, soft resets) is anxiety management dressed as mathematics. I don’t say that as an indictment, more as a recognition: rank is identity. The season is a story players tell about themselves, and the leaderboard is where they tell it to each other. Design it as a public square, not a spreadsheet.

This is also where competitive design earns its keep on behavior. Most lobby toxicity is frustration turned into blame more than it is a character defect, and nothing manufactures frustration faster than a system that feels rigged. An honest contract, visible stakes, a matchmaker that serves the match: those are queue-health features as much as they are competitive ones. What the scoreboard chooses to display is its own chapter (that’s 06).