Riot Bans 296,416 Boosting Accounts: Anti-Boost and the Limits of a Self-Audited System

Last month, while reviewing the Challenger ladder on an East Asian server ahead of a collegiate scouting report, I stopped at a familiar account. Same handle, same hours, but the match history kept changing roles: from Silver to Platinum, then back to Challenger within four days. No patch explained that curve. No champion change produced that jump. Only one possibility remained: someone else was sitting behind the account owner's keyboard.

Riot Games has just published the figure that forced me to reopen all my old notes: 296,416 accounts penalized for rank manipulation in VALORANT and League of Legends. This is not a tournament qualifier. This is the floor of the entire ecosystem — the ranked ladder every professional player, without exception, must climb before signing a first contract.

The context needs to be stated before the number.

Anti-Boost is Riot Games' automated enforcement system, designed to detect and penalize rank manipulation. It does not target playing well or playing badly. It targets intent: using someone else's account, or letting someone else use yours, in order to move ranked points in an unnatural way.

The violation taxonomy Riot published is fairly clear. First, direct boosting: a high-skill player logs into another person's account to climb on their behalf. Second, buying, selling, or transferring accounts. Third, intentional deranking to drop a tier for another purpose. Fourth, using a secondary account to assist climbing.

The point I consider most notable in the entire document: Riot draws a sharp line between self-created, self-operated alt accounts — normal activity — and rank manipulation. In other words, the system does not prohibit owning multiple accounts. It prohibits using multiple accounts to bend the ranked ladder.

That is a narrow, intent-based standard, not a bright line of the sort that says a second account is a violation. Based on my experience watching matches and anti-cheat systems across publishers for nearly two decades, I assess this as one of the narrowest definitions a major publisher has dared to publish in an official document.

On scale, the 296,416 figure is reported by Riot pooled across both VALORANT and League of Legends, with no split by title, no split by region, no comparison timestamp. That is the entirety of the enforcement data offered, on a cumulative basis.

Anti-Boost's penalty ladder runs across four tiers, and I want to read it as a pricing system rather than a moral system.

Tier one: on detection of manipulation, ranked points and rewards earned from that behavior are cancelled, the account is returned to its original rank, and a time-limited suspension follows. This is a rollback mechanism, not a prevention mechanism. It concedes that the system detects after the violation has occurred, and accepts that lag as an operating cost.

Tier two: repeat offenses lengthen the ban. The very existence of this escalation rule is a data point. If the recidivism rate were zero, nobody would need to design an escalation ladder. Riot putting it in an official document indicates the recidivism rate is large enough to be managed by law rather than by faith in players correcting themselves.

Tier three, and the heaviest level: buying, selling, or transferring accounts, or intentional deranking, can lead to a permanent ban. Riot places permanent bans exactly where the money flows. Account trading is a commercial transaction. Intentional deranking usually serves another commercial transaction behind it. The publisher has recognized that the most painful point of the gray market is not the player, but the cash flow.

Tier four is the tier I want to discuss most: joint liability. It is not only the manipulated account that is actioned. The booster's main account can be actioned. And players who frequently pair with the booster can also be actioned.

This is a very heavy-handed design decision. It turns a shared-play relationship into a risk variable. In the data I gather from East Asian and Korean ranked ladders, a top player typically has a small fixed group they queue with, pairing for several hundred matches each season. If that person is ruled a booster, the whole group enters the risk zone — even though most members may know nothing about what is happening behind the screen.

In esports, a single millisecond is a tactical gap. In governance, a single pairing relationship is a liability gap.

On the roadmap, Riot says the system will be expanded and will add detection of boosting signs at the level of individual matches, rather than only at the account level. This is the most important change in the entire announcement. Account-level detection means looking at long-term history. Match-level detection means looking at behavioral traces inside a single match: movement rhythm, skill correlation with teammates, decision-making patterns, reaction latency.

As someone who works in data analysis, I assess this as a shift from retrospective statistics to pattern recognition. They are two fundamentally different problems. Retrospective statistics answer the question "has this account climbed abnormally". Pattern recognition answers the question "is the person playing this match the account owner". The second question is far harder, and far easier to get wrong. A real player can have a bad day. A real player can change style after watching a guide. Behavioral pattern recognition always costs some false-positive rate.

Through the Vietnam–Korea lens, I see two different speeds on the same problem. Korea has long-established esports data infrastructure: analysis centers, scouting systems, player data stored to standard. Vietnam has enormous raw data potential — a large player base, a lively account market — but thin analytical infrastructure. The consequence is that for the same boosting behavior, in Seoul it is caught by an automated system, while in many other markets it is known only through word of mouth.

When the crowd falls silent, the data speaks in its own voice. The problem is that in some markets, that voice is not being recorded.

One more point I consider important but often left at the end of the document: result cancellation. When a match is determined to involve manipulation, that result can be voided. This sounds simple, but it has chain consequences. If I beat an account being boosted, I lose points unfairly. If I lose to an account being boosted, I gain points unfairly. Rolling back a result cannot repair the experience of the people who played that match. It can only repair the ladder at the aggregate level.

This is the structural limit of every retrospective enforcement system. You can delete a data row. You cannot delete a memory.

At this point I have to separate the data from the wording.

Riot published 296,416 actioned accounts. Media repeated that figure with an interpretation attached: Riot is tightening the screws. I read closely and found no basis for that interpretation in the published data. 296,416 is a cumulative total, not a trend. There is no prior-period baseline, no time series, no monthly or per-title breakdown. A total cannot prove an increase. It can only prove existence.

I nearly made exactly this error once. In 2026, analyzing World Cup Russia data, I almost concluded Croatia defended better on the strength of an impressive PPDA sample — until I compared them against their own qualifying campaign and realized the baseline was different. Without a baseline there is no trend. Only a snapshot. And a handsome snapshot says nothing about the direction of the current.

Second problem: Riot pools VALORANT and League of Legends into one figure. A tactical shooter and a multiplayer online battle arena have entirely different boosting economies. Different rank-inflation pressure. Different regional boosting demand. Pooling them hides the difference rather than clarifying it. For a data practitioner, this is a basic grouping error: mixing two populations with different distributions into a single figure.

Third problem: enforcement data is self-reported by the publisher, with no independent third-party audit. That does not mean the number is wrong. It means the number is a claim, not verified evidence. The gap between those two things is the gap between belief and proof.

And the fourth problem, the one I care about most: no independent appeals mechanism is described. Riot detects, adjudicates, and enforces. Governance authority is concentrated entirely in a single actor. With an intent-based standard — far harder to apply consistently than a bright line — the false-positive risk is structurally non-zero. Not because the publisher wants to be wrong. Because the problem is inherently that way.

Riot says the system is designed to help the ranked environment become fairer. I register that as an expectation, not a measured outcome. Expectations deserve respect. Outcomes require data.

I do not think Anti-Boost is a bad system. The opposite. It is one of the most transparent enforcement systems a major publisher has ever publicly described, with a clear violation taxonomy, a clear penalty ladder, and a clear upgrade roadmap. Compared with the days when I helped organize tournaments in Vietnam, when anything related to ranked fraud existed only as spoken rumor, this is a long stride forward.

The journey of data is the journey of humility. The publisher has dared to publish the figure. The next thing to do is publish the baseline too.

Sports culture needs people who silently count numbers, not people who shout. And the question I leave behind: if next period Riot reports 350,000 accounts, is that a sign of tightening, or merely a sign that detection has improved? Until there is a comparison baseline, both readings are correct. For an industry trying to build trust with numbers, "both readings are correct" is the worst possible answer.