Inside Riot's Anti-Boost: 296,416 Accounts, a Four-Tier Penalty Ladder and an Unmeasured Blind Spot

Late on a Saturday night after work, I reopened the ranked match logs of a few accounts on my watchlist. One Platinum account had won 14 straight matches in six days; its headshot rate jumped from 18% to 31%; its handling of 1v2 situations had shifted decisively toward the pattern of a player two tiers higher. During the same window, the account owner posted photos from a trip to Da Lat. Everything matched a familiar behavioral template, but it was not enough to draw a conclusion. I took notes, closed the file, and waited for more data.

Three weeks later, Riot Games published a figure that forced me to reopen that file: 296,416 accounts had been actioned for rank manipulation in VALORANT and League of Legends. The match is over, but the data remains.

This is not a leaderboard, not a grand final, and not attached to any professional player. It sits at a deeper layer: the operational layer of the ranked system. For someone who works in esports data analysis, that layer deserves far more scrutiny than the headlines it can generate.

Context: when a rank becomes an asset class

VALORANT and League of Legends are two titles operated by Riot Games, but they share one ranked logic. Players climb from Iron, Bronze, Silver, Gold, Platinum, Diamond, Ascendant to Challenger. Rank is not just a badge on a profile.

For part of the player base, rank is credibility that attracts viewers on streaming platforms. For another part, rank is the target of boosting services — where a highly skilled player logs into someone else's account to play on their behalf and earn ranked points for the owner. That demand creates a gray market: people pay to be carried up the ladder, and service sellers collect payment to play the games. In between sit hundreds of thousands of accounts pushed to a level their owners could not reach on their own.

The consequences do not stop at a sense of unfairness. When rank loses value as a skill measure, every account around that zone gets placed into mismatched games. And at a deeper layer, academies and youth teams that use solo-queue rank to filter talent lose part of their input signal. An inflated ladder does not just annoy players; it blurs scouting data.

Riot calls its system against this behavior Anti-Boost. It is not a patch, not a champion stat change, not a map adjustment or map rotation. It operates at the account and behavioral layer — a completely different layer from game balance.

I wrote my blog from a rented room in Nha Trang; now probability takes me everywhere, but the principle has not changed: a mechanism is only trustworthy when you understand how it defines a violation, how it punishes it, and what it misses.

The data core: four tiers and an intent-based standard

Riot published four categories of violation within Anti-Boost's scope, each tied to a distinct level of action.

Tier one, detected violation: ranked points and rewards earned from cheating are cancelled, the account is returned to its original rank before interference, plus a time-limited suspension. This is a rollback mechanism. The system does not prevent the violation beforehand; it detects it afterward, then erases the benefit already created.

Tier two, repeat offense: ban duration escalates with the number of offenses. I consider this the single most important signal in the entire announcement. A system only needs an escalating rule when a recidivism rate large enough exists to make escalation necessary. If the recidivism rate were zero, tier two would be a meaningless line of text. Riot wrote it out, which means they measure it. That small detail says more than the aggregate figure.

Tier three, buying, selling or transferring accounts and intentional deranking: possible permanent ban. This is the heaviest level, aimed squarely at the supply side of the gray market — sellers, intermediaries, and accounts deliberately dropped in rank to make climbing back easier. Economically, a permanent ban is a tax on the expected cost of service sellers: every locked account is a sunk investment.

Tier four, associated parties: the booster's main account, and even players who frequently queue with that account, may also be actioned.

The subtlest point is in the definition. Riot states clearly: self-created and self-operated alt accounts are normal activity. Anti-Boost does not target the existence of alt accounts; it targets intent to manipulate rank. This is an intent-based standard, far narrower than a blanket ban.

As a design, this makes sense. If Riot banned every alt account, it would strike a widespread and legitimate behavior, triggering an enormous backlash from its most loyal players. Choosing an intent-based standard lets them protect that safe harbor while keeping the objective.

But an intent-based standard carries a cost. A clear-cut ban — for example, every alt account is locked — is easy to explain and easy to accept, even when harsh. An intent-based standard requires players to believe the system judges intent correctly. When Riot actions an account and the owner insists it was just an alt played for fun, the argument stops being about data. It becomes about trust.

At the implementation layer, the announcement also shows Anti-Boost expanding: new violation categories, plus added capability to detect boosting signs at the level of individual matches. This is technically notable. Account-level detection typically relies on anomalies in win rate and aggregate stats. Match-level detection requires models that recognize behavioral signatures within the game itself — harder, but able to catch things earlier, before the benefit is locked in.

Based on my experience tracking matches, this logic is familiar. When I analyze a team, I do not just look at the scoreline; I look at PPDA, high-speed running distance, contest positions. The same principle applies: results are the easiest thing to read, and also the easiest thing to be misled by. Moving from the result layer to the behavioral-signal layer is the right direction, even if it is slower on the communications side.

One scope issue deserves clarity: the 296,416 figure pools VALORANT and League of Legends together, with no breakdown by title and no breakdown by region. Analytically, this is a weakness. A tactical shooter and a multiplayer online battle arena have very different boosting economies — rank-inflation pressure, regional rental demand, and community tolerance all differ. Pooling two titles into one number destroys the ability to compare internally.

The contrarian angle: a total is not a trend

Here I have to separate the data from the interpretation, because this is where it is easiest to slip.

Riot's announcement offers a cumulative figure: 296,416 accounts. Many articles on this subject describe Riot as tightening its grip. But a cumulative figure is not a trend. To say tightening, you need at least two time points to compare. Here there is only one. A total tells you neither the rate of increase nor the effectiveness of the mechanism — it tells you only the scale of the problem as of the publication date. More disclosure is not automatically more effective enforcement; correlation between the two is not causation.

Moreover, the figure is self-reported by Riot, without independent audit. That does not mean it is wrong. It means we are reading a one-sided report with no cross-check.

The detail that gives me most pause is tier four — actioning players who frequently queue with a boosting account. Joint liability is a powerful tool, and it has a reason to exist: many boosting cases take the form of organized duos, where the booster and the boosted travel together to speed things up. But it also creates a risk zone. An ordinary player who happens to queue repeatedly with a boosting account — through matchmaking, mutual friends, or one long evening of play — could be swept in. The announcement states no specific threshold and describes no appeal mechanism.

I do not consider this a grave flaw. I consider it an unmeasured variable. And in analysis, an unmeasured variable is always where you place the error term.

One more point: the asymmetry between detection and evasion. Riot expands the system, but the boosting market adapts too. When the direct path is blocked, activity shifts to harder-to-detect channels — organized group deranking, off-platform communication, splitting playtime to avoid behavioral signatures. This is a familiar pattern in any confrontation between a control system and a gray market.

Finally, something I always note when discussing publisher data: most automated enforcement relies on behavioral signals, not direct proof of account ownership. Structurally, false-positive risk is therefore always greater than zero. This does not negate Anti-Boost's value. It only reminds us that the false-positive rate is an undisclosed variable, and therefore an unpriced one.

Takeaway

What matters is not the figure 296,416. It is the next disclosure. If Riot releases a new figure with a clear time window, we can build a real trend instead of guessing. If a publicly known wrongful punishment case emerges, we will learn how much pressure the intent-based standard can bear. And if Riot publishes a specific threshold for joint liability, that risk zone narrows — and we gain one more variable for the model.

An empty stadium does not need spectators; it needs an analyst willing to look. Here too: the ladder does not need more criticism, it needs someone who can read the data layer beneath the number.