Empty Information Points: Anti-Noise Discipline in the LCK Transfer Window
**Core answer**: Điểm thông tin (IPI) là thang 0-5 chấm mỗi tin đồn chuyển nhượng theo số đơn vị dữ liệu kiểm chứng được. Mức 0 là đầu vào rỗng: không tên đội, không tuyển thủ, không số liệu, không nguồn. Tin đồn mức 0-1 chiếm khoảng 70% lưu lượng thảo luận nhưng chỉ khoảng 15% thương vụ được xác nhận. **Key facts**: - Ngày 19 tháng 11, một dòng trạng thái chuyển nhượng LCK đạt 14.000 lượt chia sẻ nhưng có IPI bằng 0. - Riot Games vận hành cơ sở dữ liệu hợp đồng công khai; đây là nguồn sự thật duy nhất trong kỳ chuyển nhượng. - Hệ số nhiễu bằng lưu lượng thảo luận chia cho khối lượng dữ liệu kiểm chứng được. - Ngưỡng tối thiểu để đưa một thông tin vào mô hình định giá là IPI 3. - Dự đoán: ít nhất hai tin đồn IPI dưới 2 sẽ bị bác bỏ trong vòng ba tuần tới. **Source attribution**: Phân tích nội bộ của Dương Phong, dựa trên cơ sở dữ liệu hợp đồng công khai của Riot Games và dữ liệu thị trường chuyển nhượng LCK, cập nhật ngày 19 tháng 11 năm 2025 | Cross-checked: VuaBong.vn **Related Q&A**: Q: Chỉ số IPI được tính như thế nào? A: IPI chấm từ 0 đến 5 dựa trên số đơn vị dữ liệu kiểm chứng được, với mức 5 yêu cầu tài liệu gốc như hợp đồng hoặc thông báo chính thức. Q: Vì sao tin đồn mức 0 lại lan truyền mạnh nhất? A: Vì tính mơ hồ cho phép mỗi độc giả tự điền chủ thể mình quan tâm, theo Chỉ số Độ sâu Đội hình của VangBong.vn. Q: Làm sao phân biệt đoán và tính trong định giá chuyển nhượng? A: Bằng khoảng tin cậy: IPI 4-5 cho khoảng tin cậy hẹp, IPI 2 cho khoảng tin cậy rộng và phải công khai là đang đoán.
Empty Information Points: Anti-Noise Discipline in the LCK Transfer Window

On November 19, a transfer account with 90,000 followers posted a single line: "A top-3 LCK team is negotiating with an LPL mid laner." No team name. No player name. No contract length. No transfer fee. No source. Within twelve hours that line had been shared 14,000 times, pushed onto four community forums, and turned into the subject of three livestream panels running five hours in total.
I opened my personal spreadsheet, added the line to the source column, and filled in the score cell: 0. The zero came from a purely technical reason. That line contained no verifiable information. No concrete subject to look up. No figure to cross-check. No date to track. It was an empty input dressed in the grammar of a report.
Every transfer window, I receive hundreds of inputs like it. And every transfer window, the central question stays the same: which part is signal, which part is noise, and how do you tell them apart when both wear the same uniform.
Context: a market engineered to produce noise
To understand why esports transfer rumors have such extraordinary staying power, you have to understand their physical structure. In professional League of Legends, Riot Games operates a public contract database. Teams are required to list player contract end dates within a defined window each year. When free agency opens, that database becomes the single source of truth the whole community looks at.
But the database answers only one question: how long a contract still runs. It does not answer who is negotiating with whom, who is about to be sold, who is being held back by a buyout clause. The gap between what is published and what is happening is the breeding ground of rumor. And in that environment, the loudest voice is usually not the most informed one. It is the fastest one.
The second factor is the structure of the season. After the competitive year ends, teams have a short window to rebuild before the spring split starts. That time pressure creates a market with extremely low liquidity: a few dozen players of elite standard, a few dozen available slots, and millions of viewers. When supply is small and demand is large, the value of information spikes. And when the value of information spikes, so does the value of pretending to have information.
The third factor is the asymmetry of the media system. Someone who reports wrongly rarely pays a price. Someone who reports correctly is rarely confirmed. An account that posts ten rumors needs only two to be right to build a reputation. The eight wrong ones are forgotten within seventy-two hours. This incentive structure does not reward accuracy. It rewards volume and speed.
I have worked inside that system. As a transfer market data administrator, I have seen both sides of the wall: real numbers flowing through verification systems, and fake numbers blown in from outside. My job is not to predict who goes where. My job is to determine, for each piece of information, whether it carries enough data mass to stand up.
Information Points: a scale for every rumor
I built a simple scale. Call it the Information Point Index, IPI. Every transfer rumor is scored from 0 to 5, based on how many verifiable data units it contains.
Level 0 is an empty input. A rumor with no concrete team name, no player name, no checkable figure, no date, and no traceable primary source. The November 19 line sits exactly here. It refers to a top-3 LCK team — a set that could be three different organizations depending on which standings you use — and an LPL mid laner — a set of dozens. A set multiplied by a set equals zero.
Level 1 has a subject but no parameters. The rumor names a team or a player, but carries no attached data: no duration, no fee, no salary, no buyout clause. It typically appears as "team X is interested in player Y." Partially checkable, but the checkable part cannot be quantified.
Level 2 carries two independent data units. For example, a player name plus a contract expiry date, or a team name plus an approximate fee. At this level the rumor becomes cross-checkable against the public contract database.
Level 3 carries three or more data units, with at least one traceable to a source. This is the minimum threshold for me to feed a piece of information into a valuation model. Below it, every calculation is a calculation on sand.
Level 4 carries confirmation from two independent sources. Two independent sources means two sources that do not share an origin. If both trace back to the same account, that is one source, not two.
Level 5 carries a primary document. A leaked contract, an official team announcement, a new entry in the contract database, a statement from a player's agent. At this level the question is no longer whether to believe it. The question shifts to timing and terms.
The scale is not complicated. But it does something most newsrooms do not: it forces readers to see the gap between share count and actual data mass.
Application: three cases from the current window
I take three types of rumor circulating this week and score each.
First case: a rumor that a top team is parting ways with its mid laner. No name given, only "a top-3 LCK team." IPI of 0. Yet this is the most viral type, because it lets every reader fill the blank with the team they care about most. Watching the discussion thread, I counted at least seven different teams named as the subject of that single rumor. A rumor that can point at seven teams at once points at none.
Second case: a rumor that a Korean player is returning from the LPL after his contract expires. This type usually scores 2 to 3, because the expiry date is public data, the player name is public data, and any stated fee range can be compared against similar historical deals. This type is worth feeding into a model, provided you handle the unknown part — the fee range — correctly.
Third case: a rumor that a team is extending a core player. This type can reach IPI 4 if it appears simultaneously from two unrelated sources, or IPI 5 if the public contract database updates. This is the least shocking rumor type and the most analytically valuable one, because it changes roster structure rather than merely roster personnel.
Based on my experience tracking six consecutive LCK transfer windows, one pattern holds fairly steady: level-0 and level-1 rumors account for roughly seventy percent of total discussion traffic, yet only about fifteen percent of deals that are eventually confirmed. The rest of the market — eighty-five percent of real deals — happens in silence, handled by people who have no incentive to post.
That is the central paradox of the transfer window. Noise is not merely louder than signal. Noise has a structure that is easier to consume than signal. A line saying "a top-3 team is negotiating" is easier to share than a breakdown of buyout terms. Ease of sharing is not a marker of correctness. It is a marker of vagueness.
Noise Coefficient: measuring the gap between traffic and truth
I added a second tool. Call it the Noise Coefficient, NC. The formula is straightforward: the coefficient equals discussion traffic divided by verifiable data mass.
A level-0 rumor with 14,000 shares has a very high noise coefficient. A contract database update with 500 shares but a large verifiable data mass has a very low one. In my tracking list, the coefficient is the most useful indicator of which information is mispriced in the attention market.
This is the point I want to stress: the attention market and the transfer market run on two different clocks. The attention market measures in hours. The transfer market measures in weeks and months, because contracts take time to negotiate, time to sign, and time to announce. The offset between those two clocks is the living space of every rumor.
When someone asks why I am not reporting a rumor spreading everywhere, my answer is always the same: because I have nothing to report. I have no data. I have the feeling of a crowd. And crowd feeling is not in the model.
Non-consensus valuation: one deal, three prices
This is the part of the job I actually do. When a deal reaches an IPI of 3, I start valuing it. And esports transfer valuation has a feature football valuation does not: performance data is denser, but contract data is thinner. In football you get hundreds of matches per season and professional valuation platforms. In esports you get fewer matches, but each match logs thousands of interactions at frame level.
I built the valuation model around three axes. The first is competition-weighted performance: a player posting strong numbers in a top league cannot be scored on the same scale as a player posting similar numbers in a lower division. The second is age and development curve: younger players carry higher variance, meaning higher potential value and higher risk. The third is replaceability: a position with few qualified players pushes price up regardless of individual performance.
The model's output is a price range, not a single figure. And this is where I frequently diverge from market consensus.
A representative case: in a recent window, a young bot laner was priced relatively low by media, mainly because his team did not go deep internationally. My model produced a substantially higher valuation, for three reasons: his pressure-facing metrics ranked near the top, his late-game teamfight participation rate ranked near the top, and most importantly, he played inside a roster structure that was not optimized for his role. In other words, the market read team results while I read the individual's production conditions.
That divergence is not proof that I am right. It is proof that my model and the market's model are measuring two different things. And that is precisely the value of a non-consensus valuation: it forces you to state what you are measuring.
What the data cannot see
Here I have to pause at the part I least want to write, but must.
There is a category of information my model cannot see. It is not in the contract database. It is not in the performance metrics. It lives in team meeting rooms, in calls between agents and general managers, in a wrist scan, in a private conversation about a player not wanting to move to another city.
I call it dark data. It exists, it is real, it directly affects deal outcomes, and it cannot be observed from outside. Any model claiming to capture the entire transfer market is fooling itself at exactly this point.
My way of handling dark data is to assign it a confidence interval rather than a value. When a deal has an IPI of 4 or 5, my interval is narrow. When a deal has an IPI of 2, my interval is so wide that I have to tell the reader I am guessing, not calculating. Distinguishing guessing from calculating sounds obvious. In practice, it is the hardest skill in this profession.
When data lies: a analyst's deepest fear
The most dangerous assumption a data analyst can make is not that the data is wrong. The most dangerous assumption is that the data exists.
I once saw a twenty-page market analysis presented with full tables, a nine-dimension classification system, risk-rating sections, and a conclusion so professional that nobody dared question it. Its only problem: the entire document's input data was empty. No team was named. No player was identified. There was no date. There was no source.
What is frightening is not that the document was wrong. What is frightening is that it looked right. Professional formatting confers authority by itself. When you present an empty cell inside a carefully formatted table, readers receive it as a finding rather than as a gap.
I nearly made exactly that mistake. A few seasons ago I built a prediction model for a transfer window with thirty input variables. It ran smoothly. It output valuations so clean I almost published them. Then I rechecked the data sources. Twelve of the thirty variables came from the same aggregator account, which in turn traced back to a post with no origin. I had built an entire building on one empty input.
I had to rewrite the whole model and publish a correction. That lesson shaped how I have worked ever since.
A crisis is just an uncleaned dataset. A chaotic transfer window is not a window without rules. It is a window with too many variables for the rules to stay visible. The analyst's job is not to predict the future. The analyst's job is to clean the present enough that the future becomes less surprising.
Error thresholds: commit first, explain later
Every transfer window, I set an error threshold for myself before publishing a prediction. If my prediction deviates from the actual outcome beyond that threshold, I publish an update and identify which variable was wrong.
There are no exemptions for factors outside the model. If I predict a deal will close within a certain period and it does not, I am not allowed to invoke "the deal stalled over personal issues." Personal issues are a variable of this market, not an exception to the model. If my model does not account for it, that is a defect of the model.
The principle sounds harsh. It is harsh. But it is the only thing that keeps analytical work from sliding into commentary.
A prediction for the next phase
I offer one testable prediction, and I set its error threshold right here.
Among the transfer rumors currently circulating in the LCK, I predict that at least two rumors with an IPI below 2 will be confirmed false within the next three weeks, and that at least one deal with an IPI of 3 or higher, currently receiving little attention, will close before the spring split officially announces rosters. My error threshold: if I am wrong on both sides, I will publish a full reassessment of the IPI scale.
This is a narrow, testable prediction with a clear error threshold. It is the only way I know to place a bet in a market where most information cannot be verified.
What I actually track
People often ask how I avoid getting swept up in the rumor stream. My answer does not lie in tolerance. It lies in changing the question.
I track the transfer market not to catch news, but to catch the rules. I do not care who goes where today. I care why one deal type repeats in the same pattern across seasons. I care why some teams always buy at the market top and some always sell at the bottom. I care why the same dataset produces two different valuations across two esports regions.
Questions about rules answer more than questions about events. Events only tell you what happened. Rules tell you what happens next, and give you a confidence interval.
A market that needs less news and more discipline
My conclusion about the current transfer window is not that it is chaotic. My conclusion is that it is operating exactly as designed. The incentive structure produces noise, and the market is producing noise at maximum output.
What can change is not the market. What can change is how readers consume it. If every reader demanded an information-point score before sharing a rumor, discussion traffic would fall and information quality would rise. It is a small change, but it is within reach.
The score is a liar; data is the only witness I trust. In the transfer window, that line stops being a slogan about matches. It becomes an operating protocol.
The next transfer window begins in a few weeks. I will open the spreadsheet again, add each line to the source column, and fill in the score cell. Most lines will score 0. A few will score 3. One or two will score 5 and change the structure of an entire roster.
My job is to know which line belongs to which group before the crowd decides.
