I have tracked the ranked ladders of VALORANT and League of Legends across many seasons, and there was a moment that forced me to stop. A friend of mine, a coach for a youth team in Shanghai, sent me his list of prospective players. On that list were three names with a 68% win rate. But when I pulled up their match history, more than forty of those matches fell into hours when the account owner could not possibly have been present — three in the morning local time, exactly when a different account in another time zone was active continuously. This is not a detective story. This is a problem about how a system is designed to distinguish skill from fraud. And Riot Games has just released a figure that makes that problem more urgent: 296,416 accounts with rank manipulation behavior across VALORANT and League of Legends.

The crowd reads that number and nods. I read it and see a gap. Not a gap in the number, but in the way it is presented.

Context: When the publisher becomes police, judge and executioner

The war on boosting in esports is not new. What is new is that Riot Games has formalized it into a named system: Anti-Boost. This is an automated mechanism to detect and punish rank manipulation across the ladders of both VALORANT and League of Legends — two games in two entirely different genres, one a first-person tactical shooter, one a multiplayer online battle arena.

Before going deeper, I need to redefine a few concepts the way market observers understand them, because terminology and operational reality often diverge. Boosting is when a highly skilled player logs into someone else's account to play ranked matches on their behalf, earning rank points for the account owner. Intentional deranking is deliberately losing matches to lower one's own rank, usually to enable boosting or find weaker opponents. Account buying, selling and transfer is the black-market trade in accounts. And smurfing is a secondary account a skilled player uses to face weaker opponents.

What stands out in Riot's definition is not the list of violations. It is the safe harbor they actively carve out: alt accounts created and operated by the player themselves are considered normal activity. Anti-Boost targets intent to manipulate rank, not the existence of alt accounts. This is a narrow, intent-based standard, and like every other intent-based standard, it looks good on paper but is hard in the field.

I have spent years working with player-behavior data, and I have one principle: when a system is designed to judge intent, its quality depends entirely on the quality of the signal, not on the strictness of the rule. The tighter the rule, the cleaner the signal must be. And that is the point I want to examine closely.

Core analysis: A four-tier penalty ladder and a joint-liability model

Reading the Anti-Boost mechanism closely, I see Riot running a fairly clear tiered penalty structure. What is interesting is not that they have penalties, but how they escalate severity.

Tier one: upon detecting manipulation, rank points and rewards obtained through cheating are cancelled, the account is returned to its pre-manipulation rank, along with a temporary suspension. This is a rollback penalty — the goal is to restore a state, not to punish severely.

Tier two: on repeat offense, ban duration escalates. This detail deserves a pause. The very existence of an escalation rule implicitly concedes an assumption: the recidivism rate is not negligible. If the recidivism rate were zero, an escalation rule would be redundant. A publisher designs an escalation system because they have seen data on offenders coming back.

Tier three: account buying, selling or transfer, or intentional deranking, can lead to a permanent ban. This is where Riot applies its heaviest penalty, and it reserves it for violations with clear commercial motive. A behavior is deemed heavier not because it harms the match more, but because it sits within a chain of transactions that is measurable.

Tier four, and this is the tier that draws my attention most: joint liability. The booster's main account can be actioned, and teammates who frequently play with them may also be affected. This is a design point that extends the boundary of punishment beyond the individual violator, and it creates a risk zone that anyone who queues ranked in a fixed group falls into.

Let me draw this structure as a table, because it is my professional habit.

| Tier | Behavior | Penalty | Design Characteristic | |------|---------|-----------|-------------------| | 1 | Detected manipulation | Cancel points and rewards, restore rank, temporary suspension | Rollback, not punitive | | 2 | Repeat offense | Escalating ban duration | Escalation, concedes recidivism | | 3 | Account trade, intentional deranking | Possible permanent ban | Heaviest, targets commercial motive | | 4 | Joint liability | Booster's main account, frequent teammates | Boundary expansion, false-positive risk |

When I look at this table, what I see is not an anti-cheat system but a governance model. And every governance model must answer three questions: how to detect, how to adjudicate, and how to appeal. Riot controls all three. They are the detector, the adjudicator and the executor. The article on Anti-Boost describes no independent appeals body whatsoever. All governance authority is concentrated in the publisher's hands.

This is not necessarily bad. In an ecosystem owned end-to-end by the publisher, centralized authority can react faster than any external regulator could. But it also raises a question about the transparency of the standard. When the standard is intent-based, and the adjudicator also owns the signal, ordinary players have no way to check for themselves whether they fall into the danger zone.

In the course of tracking ranked matches for my analytical work, I have recorded a pattern. Rank manipulation behaviors are not evenly distributed over time. They cluster at the end of the season, when ranked rewards become an exchangeable asset — titles, frames, and in some markets, the commercial value of the account itself. This is why escalating penalties by behavior, rather than by period, may miss the peak of cheating right at the moment of highest demand.

Another data point worth noting: Riot says it is expanding the system and developing the ability to detect signs of boosting at the match level. The most careful interpretation is this: the current mechanism relies largely on behavioral signals and telemetry, and they admit they need further improvement. In parallel, match results may be cancelled if cheating is detected. These three pieces — system expansion, match-level detection, result cancellation — combine into one thing: this is a reactive system with rollback, not a preventive one. That means there is always a lag between the moment of manipulation and the moment of remediation.

That lag is the window the crowd cannot see. The average player queues up, meets a match with a booster, loses points, and moves on. They do not know that weeks later, that match result may be cancelled and points returned. The entire loss experience sits in front of the curve, while the entire compensation sits behind it. In behavioral analysis, this is the most common form of bias: people remember the feeling of losing, but not the number that was returned.

Contrarian angle: The 296,416 figure does not prove what people think it proves

This is the part I want to spend the most time on, because it touches the data-reading habit of an entire industry.

When Riot announced 296,416 accounts with rank manipulation behavior, the narrative built around it was: the publisher is tightening the screws. But look closely at the data. It is a cumulative figure, not a comparative one. It tells us a total, not a trend. There is no prior period, no baseline, no split by game, no split by region.

A cumulative figure without a baseline is a figure that cannot conclude. If last year 500,000 accounts were actioned and this year only 296,416, that is a sign cheating is declining. If last year it was 100,000 and this year 296,416, that is a sign detection is improving. Two completely opposite readings, and the article provides too little data to choose either. The conclusion "Riot is tightening the screws" is an inference by the writer, not a fact of the data.

I have fallen into this trap many times in my betting-analysis work. Once I looked at a sequence of odds and concluded the market was leaning toward one outcome, only to realize I was comparing a number with my own memory, not with another number. Since then, my principle is: never read a single number as a trend. Every time, I force myself to find the baseline before I open my mouth.

There is a second problem, more serious in governance terms. The 296,416 figure pools VALORANT and League of Legends together. These are two games whose boosting economies differ in nature. Rank-inflation pressure in a tactical shooter is not the same as in a multiplayer online battle arena, where the number of champions, roles and coordination complexity creates an entirely different skill market. Pooling them into a single figure is like adding Hanoi's temperature to Moscow's and declaring it global weather. It is not arithmetically wrong, but it is analytically meaningless.

And there is a third risk that I consider the biggest blind spot of the entire design: joint liability. The clause actioning teammates who frequently play with a booster is a broad-stroke measure. It does not distinguish between a person who knows they are being carried and a person who unknowingly climbs alongside an account for one evening of ranked. The article describes no tolerance threshold — how many shared matches counts as frequent — nor any appeal mechanism. In governance risk analysis, this is a classic form of gap: expanding the scope of violations without expanding the scope of protection.

I wonder if something similar would happen in a market I know better. In China, where I live and work, platforms usually have multi-tier appeal mechanisms for automated punishment decisions. Not because they are kinder, but because the cost of a false-positive case spreading on social media is higher than the operating cost of an appeal channel. Riot operates at global scale with a centralized mechanism, and that is a structure I want to track further before concluding it is sustainable.

There is one specific fact worth citing here: all this enforcement data is self-reported by Riot, without independent audit. That does not mean the number is wrong. It means we are reading a single source, and in any serious analytical process, a single source is a source to be verified, not one to be believed immediately.

What the data really says, and signals for the next round

When I strip away the rhetoric and keep the structure, I see Anti-Boost as an investment in maintaining trust. A clean ladder is an input to the entire esports value chain: from scouts seeking amateur talent, to viewers believing that rank reflects skill. But an investment is not a result. It is a direction, and a direction must be measured against a baseline.

Based on my experience tracking ranked matches and working with player-behavior data, I believe the signal worth watching in the next round is not whether the new figure is larger or smaller. The signal worth watching is whether Riot publishes a figure with a baseline. If they can do that, the debate shifts from belief to analysis. If they keep publishing cumulative totals, then each subsequent announcement will remain a performance of belief, not a report of effectiveness.

The crowd sleeps through emotion; I stay awake with the spreadsheet. And in this case, the spreadsheet says something rather neutral: there is a system in operation, there is a publisher governing itself, and there is a data gap waiting to be filled. The question I leave behind is not whether Riot is winning. The question is whether we will recognize their winning through which data — or continue to recognize it by believing the number they themselves provide.

Falsifiable assumption of this article: if Riot publishes figures split by game and region in the next period, and if those figures show recidivism declining over time, then my argument about the data gap weakens considerably. Data has no offseason; it just waits for you to read it.