Nine Filters for the Esports Transfer Window: Leave the Empty Cell Empty

The spreadsheet had 41 rows and 12 columns. The first column held player names, the last held an estimated transfer value. The ten columns in between were empty. One cell was filled, in the top-left corner: domain label — esports.

The person who sent it works as a junior scout for a mid-tier organisation in Seoul. He attached one line: "Can you rate this squad for me?" I opened the file, read every row, and told him I could not. He asked whether it was a matter of time. I said no. I could not rate it because across 41 rows there was not a single fact to rate.

No tournament name. No patch number. No minutes played, no win rate, no gold-per-minute, no damage differential, no deaths before minute ten. Only names, and one number at the end of each line — the number everyone wants to believe and almost nobody can verify.

An empty dataset is not a bad dataset. It is a dataset that has not been collected. Those are different things, and how a person reacts to the difference determines whether they are a scout or a rumour vendor.

This piece is the filter I run on myself every transfer window: nine dimensions, each with three questions — what to measure, what breaks it, and what the data cannot see.

Context: a window where noise outruns signal

The esports transfer market runs on a paradox. The volume of information produced in the six weeks between seasons exceeds the volume produced in six months of competition, while the share of verified information falls sharply. Rumours are cheap. Contracts are expensive. Readers are paid in rumours; clubs pay in contracts.

I track the transfer market not to catch news but to catch patterns. News changes daily. Patterns hold for three years: clubs buy at the top of the value curve, players leave at the top of the expectation curve, and fans see only the intersection of those two curves without seeing their shapes.

This window, the flow passes through three gates. The first is the real contract: release clauses, term, salary structure, buyout. The second is agent behaviour: an interview that mentions no contract can move a player's price by twenty percent inside forty-eight hours. The third is community speculation, repeated often enough to become the foundation for further speculation.

The structure of release clauses and the new salary cap is the actual story. Everything else is surface.

Readers today are drowning in rumour, and what they need is not more of it. They need a reliability filter, an injury update, one line of structural logic solid enough to say: this club cannot sign that player, not because it does not want to, but because its wage bill has hit the ceiling and his release clause is roughly three times this window's transfer budget.

The empty spreadsheet arrived at the densest point of the news cycle. And it made something obvious: most transfer analysis is being written on an input even emptier than that scout's file — the only difference being that the writer does not know, or knows and writes anyway.

A crisis is just a dataset that has not been cleaned.

I split the filter into nine dimensions in the order a transfer dossier actually travels: patch, format, roster, region, finance, rules and governance, risk, narrative, and industry transmission. Each has one central question and a minimum evidence list. Without the minimum, that dimension stays blank.

One: patch and meta

Every esports title lives on a patch cadence. Riot Games ships League of Legends updates roughly every two weeks, each one tilting the champion table by a small margin. Dota 2 runs on a different rhythm: a few large patches a year, sometimes large enough to invert the priority order of nearly the entire hero pool. Counter-Strike 2 replaced CS:GO on 27 September 2026, dragging a restructuring of utility, weapons and the map pool behind it. Valorant adds agents more slowly, but each addition can reshape team composition for months.

Three kinds of change need separating before anyone is priced. The first is stat tuning inside roughly five percent: low-to-moderate disruption, short adaptation window, usually not enough to invert a settled order. The second is a mechanic rework: high disruption, capable of deleting one playstyle and opening another. The third is map rotation or pick-rule change: the highest disruption, and the most common source of shocks that analysts mislabel as "form".

I never trust goals. I trust chances created. In esports the equivalent of a chance is the resource differential at minute fifteen and the count of deliberately forced fights. The scoreboard is a liar; data is the only witness I trust.

The minimum evidence here is four items: game title, patch number, the list of adjusted champions or weapons or maps, and win-rate plus pick-ban deltas across two known patches. Sources for the last group are typically OP.GG, U.GG and Oracle's Elixir for League of Legends; HLTV for Counter-Strike; VLR.gg for Valorant; WanPlus for regional circuits. No named source, no weight.

The biggest risk in this dimension is patch targeting: a publisher deliberately weakening a dominant style. Detection is methodologically simple but requires history — a team's or player's win condition must be stable across at least three consecutive patches, and each period of stability must coincide with an adjustment round. Without that history, every targeting claim is inference dressed as fact.

One item almost no transfer dossier records: the gap between tournament server and public server. It is the highest-frequency source of pre-event controversy in the industry, and it is only detectable if you know the tournament name and the patch number. Without both, any comment about a team "failing to adapt to the meta" is unfounded.

What the data cannot see: coaching adaptation speed. No public dataset measures whether a team needs four days or four weeks to understand a patch. That is why I keep this cell as a rough estimate with a confidence interval, never a single number.

Two: tournament format and variance

Format is the single largest lever on upset probability, and the most ignored variable when players are compared. A BO1 series carries several times the variance of a BO5. Single-elimination inflates the apparent strength of weaker teams. A lower bracket rewards adaptation and depth. Swiss pairing reduces meaningless games but makes each team's sample non-uniform.

The transfer-market consequence is concrete. If a player's playoff win rate is high but most of his sample comes from BO1 series, that rate cannot be compared with a player who played the same number of games under BO5. Placing the two numbers side by side without normalisation is a methodological error, not a viewpoint.

I apply a simple correction: inverse variance weighting by format. Each game is multiplied by a factor inversely proportional to that format's upset probability. After normalisation, the gap between players typically contracts by thirty to forty percent against the raw ranking. The contraction is the noise that has been removed.

Schedule density is the second variable. A compressed calendar immediately after a patch rewards teams with strong analytical staff over teams with strong individual mechanics. For a transfer dossier this means: signing a player from an organisation with a good analysis room into one without it means buying someone accustomed to being fed information, not someone who generates it. Those two things carry different prices.

Minimum evidence: tournament name, tier, format, series length, number of participants, qualification route, dates and venues. Without a tournament name, any claim about "strength of draw" is unverifiable.

What the data cannot see: a player's comfort with format. Some perform better with two lives, some with one. No dataset measures this at scale, because single-elimination samples per player are too small.

Three: roster and players

Roster-move magnitude is the key classifier. Replacing three or more starters is a rebuild, not a reinforcement. The synergy cost of a rebuild far exceeds the nominal transfer cost, and the adaptation window typically runs one to two splits. In the transfer window, this is the number clubs undervalue most, because it never appears in any headline.

I have watched rebuilds across both the LCK and the VCS over five years. The pattern is stable: teams replacing two players usually keep the old shot-calling structure and stabilise in three to five weeks; teams replacing four usually lose an entire split, and during that period every individual's metrics deteriorate against their own baseline at the previous organisation. This is not a form slump. It is coordination cost being charged back to individuals.

Role fit is title-dependent and cannot be generalised. In first-person shooters, the in-game leader is the position with the highest rebuild cost, because the calling system must be built from zero. In multiplayer online battle arenas, the macro link between mid, support and jungle is the expensive part, and it usually does not sit with the highest-rated player.

The age curve is role-specific. Reaction-heavy roles — marksman, assassin, entry — decline earlier than calling and support roles. A dossier applying one age adjustment to every position is structurally wrong, even when every number inside it is correct.

Cross-region signings add two costs that never appear on the payroll sheet: language and the rebuilding of the calling system. I have seen enough Vietnam–Korea contracts to state a pattern: imports usually reach their previous baseline after six to eight weeks, and only recover their previous peak if the team has a translator who understands both the language and the title's technical vocabulary. Without that person, the most expensive contract on paper becomes the cheapest on stage.

Bench depth and academy output are two metrics the market prices near zero, while they determine survival through an injury season. A roster with no substitute in the jungle position is holding a timed device, and it detonates in the most important split.

Minimum evidence: team name, full starting roster with positions, recent roster moves with dates, coaching and performance staff, recent results with opponent strength.

What the data cannot see: scrim results. It is the most cited and least verifiable number in the industry. I do not use it, and I advise anyone reading "this team is winning every scrim" to treat it as a story, not a fact.

Four: regional landscape

Regional standing is title-specific and must never be generalised. A region can rank high in one title and low in another at the same time, several tiers apart. Any sentence claiming "this region is weak" without naming the title is analytically meaningless, even when it feels true.

For Vietnam, this is where I focus, and where media in both countries misprice most. Vietnamese teams at international events are routinely placed in the weak bracket on the basis of market size rather than head-to-head data. Normalise by games won against major-region opponents in play-ins, and the gap between the middle of a minor region and the middle of a major region is far smaller than the nominal ranking suggests.

Talent flow is an early indicator. When domestic teams begin importing from a region, it usually signals that region is producing more talent than it has domestic slots. When the flow reverses — young players leaving early — it signals academy output has outgrown the domestic ceiling. Both are structural signals, not news.

The talent-gap risk is a medium-to-long-horizon signal: a retirement wave among veterans outpacing academy production. Measuring it requires age distribution across an entire region over at least three seasons. Without that, every "talent crisis" forecast is a feeling written as a sentence.

Minimum evidence: named regions and leagues, international results over a two-to-three-year window, import and export transactions, academy debut counts.

What the data cannot see: training infrastructure quality. Two regions with identical player populations can have entirely different output because one has stable connectivity and the other does not. It is a physical variable, and it appears in no ranking.

Five: club finance

Esports clubs commonly run salary-to-revenue ratios above eighty percent. That is a structurally loss-making shape, not a bad year. The consequence for transfer analysis: the analytical value of any financial figure depends on whether it describes a new event — a fresh capital injection, a new sponsor, a delayed payment — not on the size of the number.

I distinguish three source types. A club press release is a controlled source: usually right on structure, vague on amount. A journalist's estimate is a motivated source: right on order of magnitude, wrong on detail. A leaked contract is the highest-value and highest-legal-risk source, and also the easiest to falsify. These three cannot be placed side by side without weights.

Sponsor concentration is a standard red flag: one sponsor exceeding half of revenue. Without a revenue breakdown it cannot be measured, and any claim about a club's financial health is speculation.

Parent-company contagion is a recognised systemic risk: organisations backed by real-estate or streaming-platform capital can lose funding quickly when the underlying business runs into trouble. Assessing it requires naming the capital backer.

For a transfer dossier I always convert every figure to total cost of ownership across the full contract term, not the nominal transfer fee. A contract with a low fee, high salary and long term can cost more than one with a high fee, average salary and a clean release clause. Most transfer reporting skips this calculation because it generates no headline.

The risk signals to track here are few: wage delays, dissolution, slot sales, or mass contract restructuring inside a short window. Four signals, all document-verifiable given a source.

Minimum evidence: transaction type, amount with currency, sponsor identity and tier, contract term and clauses, any reported payment delays.

What the data cannot see: the signatory's motive. A contract can be signed to hold a slot, to satisfy a sponsor, or to create liquidity for another deal. No dataset records that.

Six: rules and governance

Esports governance is more publisher-centric than most entertainment industries: the publisher writes the rules, owns the IP and holds commercial stakes, with no independent third-party arbitration. That structure shapes every punishment scenario, and it is why cases in this industry rarely end in a predictable ruling.

The checklist has five items: competitive integrity, transfer and registration rules, contract compliance, minor protection, and publisher governance controversies. Each has its own question, but all require one shared thing: an alleged act.

Without an alleged act, compliance analysis has no object. This is where transfer-window writing fails most often: using a sanction from one league to infer the outcome of a case in another league, another title, another region. Comparing punishment severity without naming the applicable ruleset is an invalid comparison.

I always draft three scenarios for any dossier with a legal element: worst case, middle case, optimistic case. The worst case is usually not the heaviest sanction but the duration. A case running six weeks inside a transfer window can ruin a team's season even if the final finding is no violation.

Minimum evidence: the rule invoked, the alleged conduct, the governing body, and precedent sanctions within the same title and region.

What the data cannot see: negotiating pressure behind the scenes. A case can be settled by an undisclosed agreement, and the public dataset will then record an outcome different from the real one.

Seven: risk profile

I build the matrix across six groups: competitive, financial, personnel, rules, public opinion, systemic. Each has its own screening question, and the principle is that risk must be flagged first, whatever the tone of the source.

The three highest-frequency damaging risks in this industry are wage delays, injury to a key player, and competitive-integrity violations. All three are screenable given a club name, a player name, or a tournament name. Without any of the three, screening cannot be performed.

One point deserves emphasis because it is the most common misreading: the absence of red flags is not evidence of the absence of risk. It is evidence of the absence of data. When an analysis names no risk, the right question is not "is this team safe" but "did the writer have enough evidence to screen".

The most concerning systemic risk sits inside the analyst's own chair: making decisions on an analysis whose input was empty. That risk is high, and it appears in no conventional risk matrix because it belongs to the analyst, not the subject.

Minimum evidence: any one of game title, team name, player name, tournament name, or publication source. Any single one permits partial screening.

What the data cannot see: risk that has not fired yet. A roster with three players expiring in the same month carries a structural risk no metric reflects.

Eight: narrative and expectation

Narrative analysis requires locating the subject in a heat cycle: budding, accelerating, climax, backlash. Overhyping at the climax plants the seeds of a backlash, and in the transfer window that backlash usually lands in the first month of competition.

The primary test for organic versus manufactured narrative is cross-channel consistency: mainstream press, vertical esports media, and short-video community channels. When all three tell the same story with the same dataset on the same day, a single source is almost certainly pushing. When all three tell different versions, the story is still being shaped by the market rather than settled.

Expectation-gap analysis is the most practically useful part. For each subject I keep three columns: market expectation, objective assessment from data, and the gap. A large positive gap signals overpricing. A large negative gap signals underpricing — and that is where I look for opportunity, not for news.

Separating commercial value from competitive value is the core of this dimension. A player can carry commercial value many times his competitive value, and that is entirely rational for a club selling image-linked merchandise. Trouble only appears when the two are blended into one ranking without weights.

Minimum evidence: the narrative label, the publication channel, engagement metrics where cited, and odds movement purely as a market-expectation signal.

What the data cannot see: the intent of the person placing the story. A rumour can be released to raise a price, to lower one, or to mask a parallel deal.

Nine: industry transmission

The chain runs from upstream publishers and event licensing, through midstream clubs, organisers and streaming platforms, to downstream sponsorship, derivatives and mainstream adoption.

Each publisher produces a different ecological rhythm. Riot operates on a two-week cadence with a dense year-round calendar. Valve operates sparsely with long gaps between majors. Tencent runs a season model tied tightly to its domestic platform ecosystem. Three rhythms produce three speeds of talent formation and decay, and applying one framework to all three is a structural error.

Four contraction signals are worth tracking: publisher esports budget cuts, event outsourcing, franchise slot devaluation, and sponsorship tightening. All four are observable, but only once a publisher, league, or sponsor category is named.

Betting markets, where relevant, must be described purely as a market signal. Odds movement is an indicator of collective expectation, not advice. I never let it into the conclusion of a transfer analysis.

Minimum evidence: publisher name, title, event or league, sponsor category, broadcast-rights context.

What the data cannot see: revenue-sharing terms between publisher and clubs. That variable decides the health of the entire midstream, and it is almost never disclosed.

Data recovery requirements

From all nine dimensions I derive one minimum list, ranked. Level 0 is two items: game title, and at least one named entity — team, player, tournament, or publisher. Without both, no dimension is valid. Level 1 adds the patch number where the piece concerns gameplay, the tournament name with format and series length, and the publication source. Level 2 adds publication date, regions involved, and any monetary or contractual figure. Level 3 is the alleged rule violation, if any.

I apply a hard gate to myself: if the evidence list is empty, or the title is unidentified, I stop and write a note instead of writing on. Continuing in that situation is not analysis. It is production.

The contrarian angle: the market does not lack data, it lacks people willing to leave a cell empty

The most telling thing about that 41-row file is not that it was empty. It is that it was sent. A scout spent time collecting names, spent time building columns, and shipped a product whose substance largely did not exist. He did not do it deliberately. He was following the incentives the industry rewards.

The incentive structure in esports rewards the person who fills the cell. Someone who writes "insufficient data" is judged incompetent. Someone who writes a fluent paragraph about an unverifiable player is judged to have a viewpoint. Those two responses teach opposite lessons to the same person.

Correlation is not causation, and in the transfer window the confusion takes a very specific shape: a team signs a player and improves, so the signing is declared a success. The likelier explanation is that both events were caused by a third variable — a new coaching staff, a favourable patch, or simply an easier schedule in the first three weeks.

I have filled the empty cell myself. In my first year of transfer analysis I priced a young player off six games and published a number without a confidence interval. Three months later, new data showed my sample was selection-biased: all six games were against weak opponents. I published a public correction, stated the error margin, and set a hard threshold from then on — below twenty games, every valuation must carry a confidence interval at least forty percent wide.

Leaving a cell empty is not a failure of analysis. It is the correct output of analysis when the data is not there. And in a market where everyone fills the cell, the person who knows how to leave it blank holds a structural advantage: they never have to retract a published number for lack of sample.

What I am watching next

Before the first match starts, the number has already whispered the result. Next window I will track something other than rumour: which organisations begin publishing their own minimum-evidence requirements before signing. Those organisations will not buy cheaper, but they will pay less for empty cells. And if I am right, the competitive edge over the next three seasons will not come from a better prediction model — it will come from the discipline to sit with not knowing while the data arrives.