296,416 Accounts and the Limits of a Crackdown: How Riot Runs Anti-Boost in VALORANT and League of Legends

PART 1 — A NIGHT ON THE LADDER Around 2 a.m., I reopened the replay of a Diamond-tier VALORANT ranked match. The account on the left side of the screen had 214 matches in 11 days — close to 20 a day. Its win rate over the last 40 matches was 87.5 percent. Its kills per round rose from 0.61 to 1.34 in six days. That curve does not have the shape of a player improving. It has the shape of two people sharing one account. I have covered esports for ten years, most of it reading match data rather than watching highlights. My trade taught me something uncomfortable: the biggest distortions in data rarely live in the mean. They live in the tail of the distribution, where a few individuals run faster than the rest of the sample in a way that makes no sense. That account was in the tail. Riot Games saw it too. According to what the publisher has disclosed about its Anti-Boost system, 296,416 accounts have been actioned for rank manipulation across VALORANT and League of Legends, covering the window from late last year to the time of publication. But this is where I stop and put a hat on the data, following the habit I built in 2026. A total is not a trend. A publisher announcement is not an independent audit. And a system designed to protect ladder integrity can generate a new kind of risk that it has no instrument to measure. This article does not retell Riot's announcement. It reads that announcement as a governance document — and it looks for the places where the document says less than it believes it is saying. PART 2 — CONTEXT: A SYSTEM DESIGNED AT ANOTHER LAYER Before the detail, one methodological boundary must be fixed. The source document I am analysing is a summary of Riot Games' official communication about Anti-Boost. It contains no patch information, no professional tournament information, no teams, no players, no transfers and no club finance. That means the entire analytical weight has to sit on two axes: rules and governance, and industry transmission. This is not a small limitation. It changes the reading entirely. Anti-Boost operates at the account and behavioural layer, not at the gameplay-balance layer. That has one very concrete technical consequence: the patch cadence of VALORANT and League of Legends is largely irrelevant to the effectiveness of this system. You can ship three patches in a month, or none, and the capacity to detect boosting stays the same. Two different clocks are running in parallel, and the community routinely conflates them. I have made that mistake. In 2026, while analysing an unusual climbing stretch in a regional circuit, I blamed a balance patch and ignored the simpler possibility: a group of accounts being operated by someone else. That lesson made me separate the two analytical layers — gameplay balance and system integrity — before drawing any conclusion. By definition, this system belongs to what I call online ladder protection mechanisms. Its subject is not a bracketed tournament with a group stage, qualifiers and a schedule. Its subject is a continuous ecosystem: millions of accounts, running 24 hours a day, with no clearly terminating season in the traditional sense. That difference matters. In a bracketed event, cheating affects a finite set of matches and can be handled by administrative sanction. On a continuous ladder, rank manipulation affects millions of matches, has no identifiable plaintiff, and the penalty must be issued by the operator of the system itself. Riot is simultaneously the legislator, the investigator and the judge. That is the starting point for everything that follows. PART 3 — CORE ANALYSIS 3.1 Four lines that cannot be crossed The document I read classifies violations into four main groups. The classification matters more than it appears, because in system governance, how you define a violation is how you decide who gets punished. Group one is boosting in the narrow sense: a higher-skilled player logs into someone else's account and plays ranked matches on their behalf, earning rank points for the account owner. Group two is buying, selling or transferring accounts. This is the most clearly commercial behaviour, and according to the document it can lead to a permanent ban. Group three is intentional deranking — deliberately losing matches to lower one's own rank. The usual motive is to create a larger skill gap in later matches, or to help another account climb more easily. Group four is smurf-assisted climbing, where a high-skill player uses a secondary account to face weaker opponents, or coordinates with a main account to pull rating. One design detail stands out: all four groups are defined by intent and by observable behaviour, not by the existence of an account. That is a deliberate choice, and it produces both the system's greatest strength and its greatest weakness. 3.2 The alt-account safe harbour Riot states something many readers skim past: creating and operating your own alt account is normal activity. Anti-Boost targets the intent to manipulate rank, not the existence of alt accounts. This is a deliberately narrow boundary. It protects a large population of legitimate players — people with multiple accounts for sound reasons, such as learning a new agent, playing with friends at a different skill level, or simply wanting a lower-pressure space. It also narrows the system's target to the group that actually causes harm. But narrow boundaries carry costs. An intent-based standard is far harder to apply consistently than a bright-line rule. With a bright-line rule, players know exactly where they stand. With an intent-based standard, players know exactly where they stand only after they have been actioned. I have seen this model in a different domain. In football transfer analysis, financial fair play rules also operate in two modes: hard thresholds and intent-based assessment. The first is transparent and litigious. The second is flexible and vulnerable to accusations of favouritism. Neither option is free. Data does not lie, but it learns how to hide the most important thing. In this case, what it hides is the criterion separating a legitimate alt account from an alt account being exploited. The source document offers no specific threshold. 3.3 The four-tier penalty ladder The penalty structure Riot describes has four tiers, designed on an escalating principle. Tier one applies to a first detection of manipulation: rank points and rewards obtained through cheating are cancelled, the account is returned to its original rank, and it receives a temporary suspension. Tier two applies to repeat offences: ban duration increases. This is a highly informative detail, and I will return to it. Tier three applies to account buying and selling or intentional deranking: possible permanent ban. Tier four extends the scope beyond the directly violating account: the booster's main account and the teammates who frequently queue with them may also be actioned. As an internal legal design, this model is fairly hard-edged. It combines two mechanisms common to mature anti-cheat systems: rollback and escalation. Rollback has a feature few notice. It can only work after detection. That means there is always a lag between the moment manipulation occurs and the moment it is neutralised. During that lag, every affected match has already been played, every defeated opponent has already lost points, and no mechanism returns those points to them. This is the point I want to stress: the system protects the manipulated account, but it does not protect the victims of the manipulated account. Those victims are players who happened to be matched with an account being boosted. They lost, they dropped points, and nothing is refunded. No model solves this perfectly. But the degree to which a system acknowledges that limit is a marker of its maturity. The source document does not mention any compensation mechanism for victims. That is a gap to track, not yet a conclusion. 3.4 Joint liability — the biggest blind spot If I had to pick one detail in this whole document to flag in red, it would be the clause about teammates who frequently queue with a booster. Logically, the clause makes sense. Boosting is rarely a wholly solitary activity. An account being carried usually comes with a familiar group of players, and punishing only the carried account would leave most of the operation's structure intact. Broadening the enforcement scope is how you reach that structure. In enforcement terms, the clause creates a very wide risk zone. Take a completely ordinary case. Two friends play together three evenings a week. One of them, for reasons neither of them knows, is using an account that was purchased or is being played by someone else. The other did nothing wrong. Under the joint-liability clause, that person still sits inside the actionable zone. The source document gives no specific threshold for the term frequently. How many matches over how many days? Is there a distinction between randomly being matched together repeatedly and deliberately grouping up? Is there an appeal mechanism for wrongly sanctioned players? No answers appear in the document. This silence does not necessarily mean such mechanisms do not exist. They may well exist at an internal operational layer. But in a system where the publisher is simultaneously investigator and judge, not publishing the threshold and not publishing an appeal path creates a legitimacy problem. An affected player has no way of knowing whether they were judged correctly or incorrectly. Variance is not the enemy — it is the mirror that reflects the arrogance of prediction. Here, the variance lies in the possibility that the system punishes innocent players. And the central question of any automated anti-cheat system is always the same: how many false positives are you willing to trade for one true positive? Riot has not published that number. 3.5 Detection architecture and the escalation gap A significant part of the document concerns future expansion. Riot says it will continue to scale Anti-Boost and is developing the ability to detect signs of boosting at match level, not only at account level. That shift deserves careful analysis. Account-level detection rests on relatively simple signals: sudden performance shifts, unusual play-time patterns, changes in input behaviour, rank movement inconsistent with a skill history. This is the easiest detection layer to build and the easiest to counter. Match-level detection is far harder. It requires the system to recognise the signs of two different people alternating on the same account, or of a group coordinating to pull rating for one member. That is a behavioural pattern-recognition problem in a high-noise space. In principle, the match-level layer is far stronger. In practice, it also carries higher false-positive risk, because the line between a well-synced group of friends and a group coordinating a carry can be very blurry at the data level. There is a structural asymmetry here that the source document does not directly admit, but implies clearly through Riot's continued work on detection methods. The defending side must be right in every case to maintain legitimacy. The attacking side only needs to be right once to find a way around. Esports is not slower than football — it is simply running on a different clock. In football, a fraud investigation can take months and end in an administrative sanction. On an online ladder, the investigation never ends, because the subject of the investigation is constantly reborn. 3.6 A total is not a trend This is the section I want to give the most space, because it touches my trade directly. The source document states that 296,416 accounts were actioned for rank manipulation across VALORANT and League of Legends between late last year and the time of publication. The number itself tells us very little. It does not say whether enforcement is rising or falling. To answer that, you need at least two reporting periods using the same measurement method. A cumulative total is not a time series. It does not say what share of active accounts that represents. If the combined monthly active account base of these two titles is in the tens of millions, then 296,416 is a very small fraction. If the base is much smaller, the fraction means something entirely different. The source document provides no denominator. It does not give a breakdown by region, by rank tier, or by title. It does not give a recidivism rate after enforcement. And it has not been independently audited. This is data published by the very party imposing the penalties. I stress this not to cast doubt on Riot. I stress it because it is the data-reading discipline any analyst must keep. The same standard I apply to a midfielder's running distance must apply to a count of banned accounts. If I do not accept a transfer metric without an independent source, I should not accept an enforcement metric without one either. There is one indirect inference I consider reasonable and still in need of verification. The fact that Riot built an escalation mechanism based on repeat offences is a signal that recidivism is significant. If most violators were caught once and stopped, an escalation mechanism would be nearly unnecessary. Its existence is a signal, not proof. A season is a statistical sample. A decade is evidence. A single enforcement report, by the same logic, is a sample. It becomes evidence when a second and third period arrive, measured with the same ruler. 3.7 A methodological problem: pooling two titles into one metric This is, in my view, the weakest technical point in the entire document. The figure 296,416 pools VALORANT and League of Legends. These two titles are not the same thing. VALORANT is a tactical first-person shooter. League of Legends is a multiplayer online battle arena. Their ladder structures differ, their tier counts differ, their point systems differ, and most importantly the economics of boosting differ. In a tactical shooter, the value of a high-rank account is tightly bound to mechanical skill — reflexes, aim, situation handling. That makes the skill gap between booster and account owner very visible and easy to detect. In a MOBA, the value of a high-rank account is bound more to game understanding, champion pool and team coordination. The skill gap can be concealed better, and detection becomes harder. Pooling these two genres into a single metric produces a number that looks stronger than it is. It reads like a global-scale claim, but it is really the sum of two datasets with very different distributions. In statistics this is a familiar error: add two different distributions and report the result as one unified distribution. I have seen exactly this error in football analysis. A metric that pools pass counts from a possession side and a counter-attacking side produces a meaningless average. It is arithmetically correct and analytically wrong. What I want to see in the next reporting period is separation. One figure for VALORANT, one for League of Legends, each with a denominator and with the violation definitions applicable to that title. Without that, we have no way to assess the system's real effectiveness on either surface. 3.8 The black-market economy and the price of risk The permanent-ban clause for account buying and selling has an economic meaning the source document does not explore. Every black market runs on one simple principle: price reflects risk. Boosters do not sell their service at a fixed price. They price it against the probability of detection and the expected loss if detected. When Riot raises detection probability, service prices must rise to compensate for higher risk. When Riot applies permanent bans to account buyers, the pool of potential customers narrows, because most buyers care more about keeping their account than about saving climbing time. But there is a paradox here I want to put on the table. Raising the price of boosting does not reduce total demand in the short run. It shifts demand from budget-limited players to higher-budget players. In many markets, hard crackdowns have the effect of concentrating the black market into a few suppliers with enough resources to bear the risk — and those suppliers are usually more professional and harder to detect. Every number on a transfer board is a confession by a manager. The same logic applies to the boosting market: every service price is a confession about the risk the seller accepts. When you see prices rise, you should not read that as a sign the crackdown is failing. You should read it as a sign the crackdown is working at one layer and generating a side effect at another. The source document provides no data on service prices, market size or recidivism. That means everything above is reasoning from principle, not conclusion from data. I mark it at medium confidence. 3.9 The industry transmission chain Anti-Boost affects the esports industry across four layers, with impact diminishing as you move outward. Upstream is the publisher. For Riot, Anti-Boost is an investment in trust maintenance. The ladder is the entry point of the whole esports funnel: ranked players produce viewers, viewers produce fans, fans produce a market for professional competition. If the ladder loses legitimacy, the whole funnel above it loses its foundation. Impact here is medium and positive, on a medium-term horizon. Midstream is the ladder surface and the boosting economy. This is where impact is most direct and clearest. It is also the layer whose outcome is hardest to measure, because no public metric tracks the size of the boosting economy over time. Impact is medium, medium-term. Downstream, the first layer is player experience, and this is the most underrated. A clean ladder improves the experience for millions of ordinary players — people who never appear in any violation statistic. That is the largest benefit and also the most invisible one. The second downstream layer is far less discussed: the ladder's value as a talent-discovery channel. Academies and professional teams recruit from the highest ranks. Rank manipulation pollutes that signal. A clean ladder makes it more reliable. The source document does not make this connection, so I mark it at low confidence. The peripheral layer is derivative markets, including grey zones adjacent to betting. Impact here is downward pressure in the direction of suppression, medium magnitude, medium term. What stands out is that at nearly every layer, the impact is marked medium-term. No large short-term impact appears anywhere. This is the nature of system governance measures: they operate slowly, and they are designed to operate slowly. PART 4 — THE CONTRARIAN ANGLE Now to what I consider the most important part of this article. All the storytelling around esports crackdowns rests on a hidden assumption: the stronger the system, the smaller the problem. I want to put that assumption under the light of data. There is an effect I call the visibility paradox. When a detection system becomes more effective, the number of reported violations rises, not falls. This holds for most monitoring systems: recorded crime rises when police are better equipped, recorded tax fraud rises when revenue authorities digitise their data, and detected doping cases rise when testing methods improve. That means the figure 296,416 cannot be read as a direct indicator of the state of violation. It is an indicator of detection capacity. The two are different, and they often move in opposite directions. The problem is that nobody reads it that way. Both the publisher and the community have incentives to read that number as a statement about the scale of the problem. The publisher wants to prove its system works. The community wants evidence that the problem is being taken seriously. Both sides are served by the same reading. This is the biggest blind spot in the whole story. There is a second consequence of this paradox, and it is subtler. When a detection system gets stronger, it does not only detect more violations. It also changes the behaviour of potential violators in ways the system can measure less well. Professional boosters do not confront the detection system head-on. They move to harder-to-detect channels: communication on platforms outside the game, coordinated deranking in groups rather than direct boosting, multi-layered intermediary accounts, or simply adjusting service prices to filter out high-risk customers. Each of these shifts reduces their visibility in the data. The result is a measurement paradox: the stronger the system, the more the behaviour it can observe tends to become the easiest behaviour to detect, rather than the most dangerous. The most sophisticated offenders — the group that genuinely threatens deep-layer ladder integrity — are the least visible in the published figures. I once wrote about a similar phenomenon in football. When advanced defensive metrics became public, teams began changing how they defended in order to optimise the metric rather than the outcome. The metric does not record reality. The metric records human reaction to being measured. Every sports data analyst should carry that in mind. This leads to my conditional conclusion. I believe Anti-Boost is effective at removing the commodity layer of boosting — players selling services with simple technique and clear traces. I mark this judgement at medium confidence. I believe Anti-Boost has no evidence yet of effectiveness against organised boosting. I mark this at medium confidence, based on the source document providing no data stratified by sophistication. And I believe the 296,416 figure will likely keep rising in coming reporting periods, regardless of whether the real problem is improving or worsening. I mark this at high confidence, because it depends on detection capacity — the one variable Riot is actively investing in. There is one detail in the source document I want to return to. The joint-liability clause, as I said, is the biggest enforcement blind spot. But seen through this paradox, it is also a positive signal in a narrow sense. Riot accepting an enforcement scope beyond the directly violating account shows it understands boosting as a structured activity rather than an individual phenomenon. That is a correct reading of the nature of the problem. But a correct reading of the nature of the problem, when unaccompanied by clear thresholds and an appeal mechanism, becomes a tool that generates risk for legitimate players. This is the kind of error I call rational over-expansion — a measure that correctly identifies the problem but is designed with too wide a scope. Fans remember the goal; I remember the probability before the goal happened. In this story, what readers remember is the figure 296,416. What I remember is the probability that an ordinary player gets actioned for queuing with an account they do not know is being boosted — and that probability has never been published. PART 5 — WHAT TO TRACK I will not close with a summary. I will close with a list of what I will be tracking. The first signal is the next enforcement report. The trigger condition is whether Riot separates the data by title and provides a denominator. If it does, we can for the first time calculate violation rate against active base, and shift from reading a total to reading a trend. The second signal is a publicly disclosed successful appeal. The trigger is the appearance of a case where a player proves they were wrongly sanctioned. If it happens and is handled transparently, it strengthens the legitimacy of the intent-based standard. If it happens and is handled quietly, it seriously weakens that standard. The third signal is any announcement clarifying the threshold of the joint-liability clause. The trigger is Riot publishing a match count, a time window, or an appeal mechanism. The appearance of any one of those would reduce the false-positive risk I raised. The fourth signal is new violation categories added to the taxonomy. The trigger is Riot adding a new class, such as more complex coordination models. Each such addition is a marker that the race between detection and evasion has moved to the next layer. The fifth signal is another publisher releasing comparable figures. The trigger is the appearance of a benchmark close enough in definition. Without that benchmark, the 296,416 will remain a lone claim forever. There is one detail in the source document I keep thinking about. Riot says it expects these measures to help the competitive environment become fairer. The word there is expect, not result. The document provides no metric measuring fairness before and after. That is the last point I want to leave. Ten years of following this industry have taught me that data does not lie, but it learns how to hide the most important thing. In the Anti-Boost case, what is hidden is not inside the figure 296,416. It is in the absence of any number that measures what an ordinary player feels at 2 a.m., winning a fair match on the ladder, believing the result is real. If Riot could measure that, it would not need to publish any other number yet. And during the pandemic, I built an empire out of numbers nobody was watching. It still stands — but it only stands because I learned to tell apart the number that can be measured from the number worth measuring.