Riot Games disciplined 296,416 rank-manipulation accounts in VALORANT and League of Legends: four penalty tiers and one empty column

11 PM Seoul time, after my shift at the desk, I reopened the spreadsheet I use to track enforcement disclosures from game publishers. Four columns: publication date, title, accounts actioned, and the appeal mechanism available to players who were wrongly punished. The first three columns are filled in. The fourth column, on the two rows for VALORANT and League of Legends, is empty.

I did not skip it. Across everything Riot Games has published about its Anti-Boost system, the machine built to hunt boosting and rank manipulation, there is no passage describing an independent appeals channel, or any tolerance threshold that a player caught by accident could rely on.

Riot cites 296,416 accounts with rank-manipulation behaviour, pooled across VALORANT and League of Legends. When the transfer window goes quiet, I hear the spreadsheet rustle. This time the rustling comes from somewhere else: the noise sits in the number, the silence sits in what the number does not say.

Let us lock the definition first. Boosting is a high-skill player logging into someone else's account to play ranked on their behalf, climbing on the owner's behalf without the owner playing. Riot draws the line reasonably clearly: secondary accounts created and operated by the player themselves count as normal activity; Anti-Boost targets the intent to manipulate rank. That is an intent-based standard, far narrower than a blanket ban on alt accounts.

To a Vietnamese reader this may sound distant. Placed next to the Korean market, it is very close. Here, in-game rank does not stop at the screen. It is an admissions premise for esports academies, a screening criterion for youth rosters, and in some cases something openly advertised on forums. The dense PC bang network turns logging into someone else's account into a simple physical act: sit in the same row of machines, hand over ID and password, swap seats after each game.

South Korea is also one of the few countries with direct legal provisions on trading game accounts. The Game Industry Promotion Act, in its amended form effective from 2026, sets a penalty framework for buying, selling or brokering game accounts, with maximums reaching tens of millions of won in fines or prison time. That produces a curious paradox: the same behaviour is handled by the publisher with an account ban and by the state with the law. Two systems run in parallel with two different yardsticks, and ordinary players rarely know which coordinate they stand on.

One thing to flag before the detail: the source material is one-directional, originating from Riot Games itself. No voice from punished players appears. No independent monitor verifies it. All data on enforcement scale is self-reported and unaudited by any third party. That does not make the number wrong, but it raises the question of what the benchmark actually is.

The penalty structure Riot describes has four tiers, and how it operates is worth reading slowly.

Tier one applies when the system detects manipulation: ranked points and rewards gained from cheating are cancelled, the account is returned to its pre-intervention rank, and a temporary suspension follows. The phrasing matters. The penalty is not the removal of everything a player earned, only the portion established as coming from cheating. By design, this is a restorative penalty: strip the wrong part, keep the right part. The system works on a detect-then-restore model, meaning there is always a lag between the manipulation and the remediation. During that lag, clean players already lost matches, already lost points, and nobody gives those hours back.

Tier two is the escalation mechanism: repeat offences lengthen the ban. The existence of an escalation ladder is itself a data point. If recidivism were negligible, nobody would design multiple severity levels. Riot building that ladder signals an expectation of recidivism at a statistically meaningful rate. In other words, a first ban does not deter enough.

Tier three is the heaviest: buying, selling or transferring accounts, or intentional deranking to drag others down, can trigger a permanent ban. This is where Riot places the economic centre of gravity. Boosting exists because there is demand. The buyer needs a clean account. Permanent bans target the infrastructure of that market rather than individual matches.

Tier four is the most contested part and the largest blind spot: joint liability. The booster's main account can be actioned, and players who frequently queue with that person fall inside the scope. The rationale is tactically clear: ban only the boosted account and the booster moves to another within minutes; block the network around them and the operating cost of a boosting service rises. The price is a grey zone.

I have watched a fair number of ranked matches streamed out of PC bangs in Seoul, and what is easy to notice is that most players queuing together are simply classmates, flatmates, or people who happened to land in the same queue a few times in one evening. A duo playing three games together this weekend looks identical to a duo playing three games together next weekend. If matchmaking co-occurrence is the basis for inferring complicity, then the error ceiling of the system is set not by detection quality but by how frequency is interpreted.

There is another detail rarely noticed in how Riot reports its data. The 296,416 figure is pooled across two titles, with no regional split and no per-title split. Technically this is an incompatible aggregation. VALORANT is a tactical shooter where carrying depends on aim precision and round-by-round decision-making; League of Legends is a MOBA where carrying depends on wave management, objective control and long-horizon macro. The boosting economies of these two titles run on different logics: one sells individual skill, the other sells coordination. Merging them into a single number destroys exactly the information a reader needs most: which title is being manipulated more, and in which region.

For someone working out of the Korean market as I do, the missing regional breakdown is the most regrettable part. Korean servers have a distinctive ranked environment: high account density, a large share of PC bang play, and an account market that operates almost openly. With Korean-server-specific figures, we could measure the real effect of the Game Industry Promotion Act on in-game behaviour. Right now we can only speculate.

There is one more detail about the system's direction. Riot says it will expand Anti-Boost and add match-level detection of boosting signs, rather than stopping at account level. That is a significant shift. Account-level detection relies on login logs, IP addresses, devices. Match-level detection relies on behavioural patterns inside the game: movement, timing of decisions, mechanical error compared against history. The second is harder but less dependent on administrative traces.

Read together, the four tiers share a through-line: Riot is trying to shift cost from the booster to the buyer. Main account bans, associate bans, permanent bans for trading. All of it aims to make hiring a boost more expensive for the person paying. Agents do not read rumours; they read how often you are right. The boosting market works the same way: it will not read the press release, it will read whether the ban rate actually rises.

But there is a paradox I have not seen raised in the industry. Boosting is a transaction. To stop it, you must intervene in a trade, and a trade has two sides. Riot hits the supply side hard. The demand side is affected only indirectly: the buyer loses an account, loses a rank, but no data shows them being blocked at the identity layer, meaning nothing stops them from buying a new account and hiring again next week. Stripping a rank from an account is not the same as stripping the ability to pay from the person behind it.

This is the trap every account-based anti-cheat system falls into: the object of punishment is a thing, while the motive lives in a person.

The most contested part of Anti-Boost is not the heaviest tier. Permanently banning a traded account is easy to defend. The real blind spot sits in tier four, the associate-liability tier, where punishment is determined by relationship rather than behaviour. In any mature justice system, people are tried for what they did, and prosecuting a relationship requires proving that relationship had a purpose. Here the criterion is queue co-occurrence, a technical parameter, not evidence of intent.

There is a more generous reading. If Riot genuinely wants to dismantle networks, extending responsibility outward is a reasonable deterrent step, and the community might accept small collateral damage for a cleaner queue. But for that trade to be legitimate, three things are needed: a published frequency threshold, an appeals channel with a deadline, and a transparent post-audit report. The current disclosure contains none of them.

That leads to a judgement I consider more important than the 296,416 figure itself: Riot is simultaneously investigator, prosecutor, judge and executioner inside a system with no independent appellate body. That is a publisher's prerogative on its own platform, and operationally it makes sense. Nobody understands the game's data better than the people who built the game. But it means every error the system makes has nowhere to be corrected.

One more note on how the story gets told. Disclosures like this usually travel under headlines saying Riot is tightening the screws. That is an inference, not a fact. A cumulative figure with no prior-period comparison cannot produce a trend line. To claim enforcement intensity is rising, you need at least two consecutive reporting periods measured the same way. Right now we have one point, and one point is not a line. Forty-seven rumours to find one truth, and the truth always lies behind the number of times it was sent.

So what deserves watching? First, the next disclosure: if Riot returns a new cumulative figure in the same format, we begin to have a time axis, and only then can intensity be judged. Second, the first false-positive case large enough to become news. Every system built on behavioural signals will have that day; the only question is when. Third, a quieter change with long-term value: a clean ladder is raw material for youth scouting pipelines. When high rank is no longer diluted by boosted accounts, its signal value rises for academies and youth rosters hunting for talent.

One question stays in my spreadsheet, in the fourth column. If the tool that detects rank manipulation is built, operated and evaluated by the publisher itself, then who confirms it has never caught the wrong person, and will we ever get to read that number differently, as a rate rather than a cumulative total without a comparison sample?

A cleaner queue is what players deserve. But a clean queue upheld by a system nobody audits always carries a trust debt, and that debt only comes due on the day someone is punished wrongly and has nowhere to appeal.