Riot, Anti-Boost and the 296,416-Account Problem: When Punishment Reaches the Bystander

At three in the morning in Nha Trang, I stopped on the eleventh match in the log of a VALORANT Radiant-tier account. The owner had won eleven of the last twelve matches, averaged a 1.94 K/D, and posted a 78% first-gunfight win rate. What kept me from closing the laptop was not those numbers. Six weeks earlier, the same account sat in Gold with a 0.83 K/D and a 41% first-round win rate. In VALORANT's ranked system, a jump like that does not come from practice. It comes from someone else sitting in the chair. I logged the case in my notebook with a single line: "suspected boosting, cross-check against penalty data." Fourteen months later, Riot Games published the figure of 296,416 accounts engaged in rank manipulation across VALORANT and League of Legends. The number did not surprise me. What made me pause was how Riot defined the boundary of punishment: not only the manipulated account, but the booster's main account, and the players who frequently queue with them. The match ends, but the data remains. I write my blog from a rented room in Nha Trang; now probability takes me everywhere. But there is one kind of data I still record by hand: data from matches with no audience, no broadcast scoreboard, only two anonymous teams and a ranked system quietly logging every round. Boosting belongs to that kind of data. Three concepts need to be separated before we reach the numbers, because the Vietnamese community tends to merge them. Boosting is a high-skill player logging into another person's account to play ranked on their behalf, climbing the ladder for the owner. A smurf is a second account that a player creates and operates themselves, usually to face weaker opponents or test strategies. Intentional deranking is deliberately losing to lower one's own rank, often to enable boosting or find easier matches. Riot draws the line clearly: self-created, self-operated alt accounts are normal activity. Anti-Boost targets the intent to manipulate rank, not the existence of alt accounts. This is a narrow, intent-based standard, and it has direct consequences for how the penalty system works. A bright-line rule like "account trading is banned" is easy to enforce and easy to accept. An intent-based rule like "targeting the intent to manipulate rank" requires the system to infer from behavior. Inference from behavior always carries error. Notably, Riot pools VALORANT and League of Legends into a single report. The two titles are fundamentally different: one is a first-person tactical shooter, one is a multiplayer online battle arena. Rank-inflation pressure, regional boosting demand, and how players price rank prestige all differ. Pooling them strips the 296,416 figure of any ability to be analyzed by title or by region. That is a methodological weakness I will return to later. Now let us get to the core: the penalty system. Riot operates a four-tier penalty ladder. Tier one, upon detection of manipulation, ranked points and rewards earned through cheating are cancelled, the account is returned to its pre-manipulation rank, and a temporary suspension follows. Tier two, repeat offenses escalate ban duration. Tier three, account buying, selling, or transferring, or intentional deranking, can result in a permanent ban. Tier four, associated parties, including the booster's main account and players who frequently queue with them, may also be actioned. Reading this ladder through the logic of data, three points demand analysis. First, the system is reactive-with-rollback rather than purely preventive. Points and rewards are cancelled after detection, meaning a lag exists between the moment of manipulation and the moment of remediation. During that lag, other players have already faced a manipulated account. Probabilistically, if a boosting account plays an average of twenty matches before being flagged, each match places five other players into an unfair test. Twenty matches times five players is one hundred affected experiences per account. Multiplied by 296,416, the affected experiences could reach tens of millions, though this is my own estimate from observation, not official data. Second, the system escalates and tiers by severity. Permanent bans are reserved for the most commercially driven violations: account trading and intentional deranking. This is a reasonable design choice, since both behaviors tie directly to the black-market account economy. But it also implies an assumption: that a recidivism rate exists large enough to require escalation. If every offender offended only once, escalation rules would be redundant. Their existence is a signal about the structure of the problem. Third, and this is where I want to spend the most time, is the joint-liability model. Riot extends punishment to the booster's main account and to players who frequently queue with them. Theoretically, this creates a radiating deterrent: not only the direct manipulator is punished, but the ecosystem around them. In enforcement terms, however, this is the highest-risk zone of the entire system. Picture a duo that plays together every night. One of them, for financial reasons, quietly hires someone to boost their account. The other does not know. When the system detects it, both fall into the "frequently queued" category. The innocent player may be actioned with no independent appeal mechanism described. Riot controls both detection and adjudication; there is no third-party tribunal. Governance authority is fully concentrated in the publisher's hands. Probabilistically, if I assume 80% of frequent-teammate cases are innocent, then with 296,416 flagged accounts, the number of wrongly affected players could reach tens of thousands. The 80% is my assumption, not data. But even if the wrongful rate is only 10%, the absolute number is still large enough to generate a backlash should a high-profile case emerge. The intent-based standard also raises a transparency problem. When error strikes a well-known account, community trust in the system's consistency is tested. And in most cases, the community has only one side to trust: the publisher, who acts as both prosecutor and judge. Here I want to address match-level detection. Riot states it is developing the ability to recognize signs of boosting at the match level, not only at the account level. This is a technical step forward, but also an indirect admission: current methods are not yet mature. If account-level methods were complete, no additional match-level detection layer would be needed. Riot's stated expansion roadmap signals that the race between detection and evasion is still running. To understand why this number matters, look at the market behind it. I once spent three months tracking account-trading channels in Southeast Asia, recording listed prices by tier. A Gold account might sell for a few hundred thousand Vietnamese dong, while a Radiant account could reach tens of millions. That price gap reflects what economists call signal value: rank is not just a number, it is a skill certificate priced by the market. When a skill certificate can be bought with money, the motive to buy is obvious. And when the motive is strong enough, supply appears. This is why boosting is not an isolated phenomenon but an organized industry, even if most of it operates out of sight. Riot's penalties, viewed this way, are an attempt to raise the transaction cost of that market. But there is a connection the article does not mention, and I consider it important: the link between boosting and betting. A manipulated account can be used to produce abnormal ladder outcomes, and in some cases those outcomes can be exploited in prediction markets. I have no direct evidence of this link, so I mark it as a low-confidence hypothesis. But it is a hypothesis any betting analyst must place on the table, because it extends the risk surface beyond the game's boundary. One more layer of impact deserves consideration: the scouting value of the ladder. Academies and professional teams still use high ranked play as one of the first signals to filter talent. A Radiant account can open a tryout opportunity. If that rank can be bought, the scouting signal is corrupted. A clean ladder therefore serves not only ordinary players but the entire talent supply chain of the industry. This is why I do not view Anti-Boost as a support feature, but as infrastructure. This is where I must be explicit about the limits of the data. The 296,416 figure is self-reported by Riot. No independent audit. No split by title. No split by region. No clear time window. Methodologically, a cumulative figure without a comparison point cannot prove a trend. The statement "Riot is tightening control" is the writer's inference, not a conclusion supported by the data. The data shows a total, not a trend. People call me a number-obsessed nerd; I take that as a compliment. Because distinguishing a total from a trend is precisely what separates analysis from propaganda. One more point belongs on the table: economic motive. Permanent bans for account trading target the supply side of the black market. If the expected cost of account trading rises, demand falls. This is basic deterrence-pricing logic. But the article provides no data on recidivism, market size, or the elasticity of demand to punishment. Without those numbers, we can only say the direction is sound, not that it is effective. And here is the counterintuitive part I want to give the patient reader. The community often reads a report like this linearly: Riot published a big number, so Riot is winning. There is another reading. A better detection system finds more violations, and therefore publishes a bigger number, even if the violation rate within the total player base is unchanged. A big number may signal detection capacity, not a spreading problem. Conversely, a small number may signal weak detection, not a clean community. This is the classic correlation-versus-causation trap. We see more accounts actioned, and we conclude boosting is rising. The data does not permit that conclusion. It only permits the conclusion that recorded cases increased. Two competing hypotheses, one about detection capacity and one about violation levels, both explain the same data. The article provides no information to distinguish them. I have said before that sports data models tend to overrate easily measured signals and underrate hard-to-measure factors. Here, the easily measured signal is the number of accounts actioned. The hard-to-measure factor is the true reliability of the detection system. And in most publisher reports, we are shown the number, not the confidence interval. There is one more point on the inter-market dimension. If Anti-Boost works effectively, it may push boosting into harder-to-detect channels, such as organized deranking rings, out-of-game communication, or intermediary accounts. This is a prediction, not an observation from the article. But it is a grounded prediction: when the cost of one channel rises, flow finds a detour. This is a basic law of black markets, and any detection system must face it. This leads to the question of what to track in the next cycle. If I could bet on a single metric to judge Anti-Boost's real effectiveness, I would not bet on the number of accounts actioned. I would bet on the percentage of successful appeals. That is the only metric that shows whether the system is detecting correctly or detecting broadly. Riot does not publish it. But the emergence of a transparent appeal process, with thresholds and breakdowns, would be a far stronger signal than any cumulative number. The second metric would be a split by title and region. If Riot publishes VALORANT and League of Legends separately, and breaks down by regional server, comparisons become possible. Only then can I speak about boosting dynamics by market. Otherwise, 296,416 remains an irreducible number. The third metric is a comparison point. A cumulative figure without a baseline is a photograph, not a film. To speak of a trend, you need at least two photographs at two moments. Riot has published one photograph. I realize I am painting a somewhat pessimistic picture of a system that, after all, is a real effort. Fairness is required: Riot publicly disclosing a penalty system, an escalation ladder, and a specific figure is something many publishers do not do. Clearly defining the boundary between legitimate alt accounts and manipulative behavior is a step in the right direction. Extending liability to associated parties, however risky, shows they understand boosting is not a solitary act but a network. But fairness in analysis does not mean leniency in judgment. A good system can still be poorly reported. And a poor report can still conceal a good system. The analyst's task is to show the gap between the two, not to fill it with faith. An empty arena does not need an audience; it needs an analyst willing to look. If I must apply a model, I would start by separating the data into three layers. Layer one, recorded violations. Layer two, actual violations in the population. Layer three, correctly actioned violations. We only have layer one. Layers two and three sit in a black box. Any claim about Anti-Boost's effectiveness can only carry medium confidence, because two-thirds of the data required for a firm conclusion is not published. The most dangerous phase of any enforcement system is not when it is announced. It is when the first wrongful case emerges and goes viral. That is when the community tests whether the system has a correction mechanism. A system with a correction mechanism can survive one wrongful case. A system without one can collapse in reputation from a single case. Riot has not yet shown us that mechanism. In the competitive context between publishers, Riot publishing this figure is also a commercial signal. It tells players and investors that the ranked system is being managed seriously. Against titles perceived as more lax, this is a competitive advantage. But the advantage exists only if players believe the number is accurate. And that belief, once again, depends on the transparency of the method. So what are my next-cycle signals? First, I await a second disclosure in the same format. If Riot publishes a new cumulative figure with the same structure, I will have a comparison point. That is when trend analysis becomes feasible. Until then, any trend claim is speculation. Second, I await a statement on the pairing threshold in the joint-liability model. If Riot clearly defines "frequently queued" as how many matches over how many days, the wrongful risk falls significantly. If not, that risk remains. Third, I await a public wrongful case and Riot's handling of it. This is the real test of an intent-based system. I do not wish for such a case. But in a system without independent audit, its probability is not zero. Fourth, I await data split by title. VALORANT and League of Legends have different boosting economies, and pooling them wastes information. A separated figure would tell me which title faces greater pressure, and in which region. Finally, I await a recidivism figure. If it is low, escalation rules are redundant. If it is high, we face a structural problem, not merely an enforcement one. Data does not lie. But data does not speak for the missing parts either. The gap between the 296,416 figure and the question "is the ladder actually fairer now" is a gap an analyst must endure, not fill with inference. Fourteen months ago, I closed my notebook with a line of suspicion about one Radiant account. Today the question remains, only at a larger scale. The match ends, but the data remains — and part of it has not yet been published.