The Evidence Ladder: A Five-Tier Filter for an Esports Transfer Window Drowning in Noise

01:40 in the morning, and twelve articles born from one sentence

01:40 in the morning, third day of the transfer window. I am sitting in my apartment in Shanghai, the desk lamp still on, the second monitor showing the tracker I built at the start of the month — one row per name, one column per evidence tier. Someone in a group chat sends exactly one English sentence, no source, no citation, not even the handle of the original account: "He's leaving."

Forty-five minutes later I have counted twelve articles about that same name. Twelve different headlines. Twelve different opening paragraphs. Three carry a specific fee, two mention a "contract length," seven contain the phrase "a source close to the deal." I follow every link, open every tab, paste every URL into the sheet. All twelve trace back to one post from an account created three weeks earlier, and that post cites no one.

What kept me awake was not the possibility that they were wrong. It was that they were written very well. Each one contained the four components of a complete transfer report: a subject, an action verb, a hedged number, an unverifiable attribution. Not one was missing a slot. A flawless template. A hollow interior.

Technical literature has a concept I borrow often: a fail-open system. When the input is empty, a system designed to always produce output does not raise an error — it produces a complete, correctly formatted, content-free artefact. A reader downstream cannot distinguish that report from a real one. The esports transfer industry operates on exactly that principle, and it has done so for at least seven years.

Why the esports transfer window is noisier than football's

I entered the industry in 2026, starting as an athlete, moving into tournament operations, and only then into media. Eleven years of watching this business taught me one thing: the esports transfer window is not noisier because insiders lie more than they do in football. It is noisier because esports lacks something football has — a registry.

Football has FIFA's transfer matching system, international clearances, windows fixed by confederations, a central body recording every transaction. A football transfer leaves an administrative footprint. You may not know the fee, but you know the deal happened, on what date, between which two legal entities.

Esports has nothing equivalent at scale. What it does have is a contract database partly published by the publisher — in League of Legends, Riot's global contract database. It shows the player name, the registered team, and the contract start and end dates. No fees, no release clauses, no salaries. It contains one field that actually matters: the expiry date.

That single field produces an effect analysts call a step function. The contracts of hundreds of players across the major leagues expire within the same narrow stretch of November. The free-agent market does not open gradually. It opens like a switch. Before that timestamp, every claim about a player's future sits at the bottom of the ladder. After it, a batch of questions is answered automatically by a single line of registration data.

From my own experience tracking matches and transfer windows, the noise curve of a transfer window does not correlate with the number of real deals. It correlates with the distance to the lock date. The further from the lock, the more articles. The closer to the lock, the fewer articles — and the sharper their accuracy. That is a structural consequence, not a moral failure of the media.

Anatomy of a transfer report

A standard transfer report has four slots, and all four can be filled without a single real-world event occurring.

Slot one is the subject. Slot two is the verb. Slot three is a hedged quantity — "around," "reportedly," "believed to be." Slot four is the source: "a person close to the situation," "regional media," "a member of the organisation confirmed." Those four slots are enough to produce something readable, shareable, and impossible to refute within the first twenty-four hours — because refutation requires an official statement, and official statements always arrive late.

The fifth slot is the amplifier: a context paragraph copied from the previous article about the same player. This is where the fail-open design shows itself. The context paragraph adds no information. It adds volume, it manufactures a sense of depth, and more importantly it creates a circular citation chain: article C cites article B, article B cites article A, article A cites a status update nobody verified. By the third rotation, the reader sees three independent sources. In reality there is one.

The transfer news industry does not lack information. It lacks a circuit breaker. And because it lacks a circuit breaker, the default product always looks finished.

The evidence ladder — the five tiers I sort every claim into

After the 2026 window I dropped the habit of reading transfer news by feel and started sorting every claim into five tiers.

Tier one — registered state. This is the only tier with a documentary basis. An expiry date displayed in a public database. A termination notice. A retirement statement issued by the player himself. A free-agent list published by the league. These are not news; they are data. At this tier I do not ask whether it is true, only what it means.

Tier two — two confirmations that are genuinely independent. The condition is not "two sources." The condition is two sources that do not depend on each other. If source B merely repeats what source A said, it is still one source. Most "confirmed by both sides" language in the esports window is one source published twice.

Tier three — behavioural signals. A player removed from the team's marketing material. No longer appearing in joint streams with teammates. A practice schedule that has changed. A team page admin unfollowing an account. A visa process or a registered address changing. Weak evidence on its own, strong evidence when the timing lines up.

Tier four — a single claim from a credible figure. Still news, but unverified news. Its value depends on the reporter's record, and that record is shorter than you think, because a transfer window happens once a year.

Tier five — aggregation of aggregation. An article citing an article citing an article. At this tier, information can no longer be traced to its origin.

My operating rule is simple: tiers four and five are leads, not events. And never cite upward.

Three deals and their real tiers

Choi Woo-je leaving T1 after the late-2026 window is the cleanest example of a fail-open system running at full capacity. The entire community analysis apparatus knew his contract was expiring months in advance — tier one, a public data field. But the possibility of leaving, the probability of leaving, and having left are three different states, and most of the content in between sat at tier four with destinations assigned at random. When the move was announced, every prior tier-four article was retroactively counted as a correct call, including the ones that named the wrong destination but the right verb. That is the most common reputation-building mechanism in this industry, and it only works because audiences collapse multiple evidence tiers into one.

Zhuo Ding's move from JDG to BLG in late 2026 shows a different shape: a deal where tier one had almost already stated the answer. When a player enters free agency at a top-tier salary and only a handful of organisations can afford him, the plausible destination set shrinks to two or three. The remaining noise is not about whether he stays, but about where he goes — and that question can be answered through roster structure rather than through rumour.

Lee Sang-hyeok re-signing with T1 taught me the inverse lesson. The volume of rumour around a name is proportional to that name's audience, not to the probability of change. The most-followed player in the sport generates the most articles every window, including windows where the final outcome is that nothing changes at all. The noise there is not wrong. The noise is amplification.

The Vietnamese angle: a small market inside a shared system

Vietnamese fans plug into that machinery from a particular position. Lê Quang Duy played a world final in 2026 with Suning — the only Vietnamese player ever to reach the last match of the biggest tournament. Đỗ Duy Khánh has been tied to GAM Esports across multiple cycles and multiple international appearances. Those two names are enough to show something: Vietnamese players sit inside the same contract registration system as Korean and Chinese players, share the same expiry field, and obey the same switch-like law of the free-agent market.

But the volume of Vietnamese-language content about them passes through tier five more than in any other market. Most news about a Vietnamese player moving abroad is translated from a Chinese article, which translated a Korean article, which cited a status update. Language distance turns tier five into the default tier. That is a structural disadvantage, not a failure of the reader.

A one-variable experiment: delete the database

I run a small exercise on myself: change one variable, watch the whole system.

Suppose the publisher deletes the contract expiry field from the public database. No transfer rule changes, no budget changes, no clause changes. One data field disappears.

What follows is predictable. Tier one vanishes. The free-agency opening date stops being a public milestone and becomes something leaked in fragments. Fans lose the ability to know in advance who will be available, so every claim has to be filed at tier four. Demand for "insiders" spikes, because personal credibility becomes the only remaining data source. And the price organisations can charge for a leak, as a negotiating instrument, rises accordingly.

One data field is holding most of this market at a verifiable tier. This is what I want the sceptics of data transparency to understand: disclosure is not the opposite of secrecy in sport. Disclosure is what lowers the price of noise.

Where I could be wrong

The five-tier filter has four blind spots, and I have to name them before recommending it to anyone.

First, the base rate. A meaningful share of tier-four claims eventually turn out correct. If I discard tier four entirely, I also discard the fastest reporting on real deals, because several of the biggest moves of recent years leaked that way before any announcement. The filter has a real cost; it is not free. The correct use is not to reject tier four but to price it — print the tier next to the claim.

Second, tier one is not truth. An official announcement is a marketing document with a negotiated narrative. "The two sides agreed to part ways" often conceals a benching decision. "We thank him for everything" often conceals a buyout. The contract is the only place truth lives, and nobody outside that room reads it. So this ladder ranks traceability, not truth.

Third, the ladder rewards institutions and punishes individuals. The early leaker takes the risk, and my filter stamps them "unverified," while a team's statement three weeks later gets promoted to tier one. There were times I was right because of a tier-four item, wrong twice before being right, and a final label alone does not show the price that was paid.

Fourth, survivorship bias. I can only tier claims that exist. A silent deal, negotiated cleanly over four weeks without a single leak, leaves no trace in my dataset. So my record — and the record of anyone called a transfer expert — over-represents loud deals. In a certain sense, the best reporter is the one you never read during a transfer window.

What this window is actually missing

What I want after this window is not more accurate rumours. What I want is a public transfer registry, modelled on football's, with a filing date, a termination notice, and a record between two organisations.

A registry like that would erase most of this industry's noise within a single season. It would also erase most of the leverage clubs currently use in negotiations, and most of the value held by people who live off leaking. So the right question is not whether the rumour mechanism is bad. The right question is who would pay to shut it down, and whether the payer wants it shut.

I will let my evidence ladder run through the end of this window and publish the full classification after the lock date. But the metric I care about is not hit rate. The metric is how many tier-five items eventually resolved into tier-one events, and how many days that took. If that ratio falls over the next two seasons, this industry is improving. If it rises, we are merely writing more carefully for a system with no circuit breaker.