Vietnamese Chess Enters the Data Era: Lessons From an Empty File

At three in the morning I opened the live rating table and saw an empty Elo column. Not a single number moved. The game had finished four hours earlier, my editor had messaged me three times, and I was staring at a page of data that contained nothing but rows stating there was insufficient information to assess anything. In thirty-seven years of covering chess, it was the first time I sat in front of a completed game with nothing to read.

The first reflex of a reporter is to write. The first reflex of a data journalist is to check whether the gap is real. Overnight I called three sources: a supervising arbiter, a member of the organising committee, and a technician responsible for the electronic scoreboard. All three confirmed the game had ended and the result had been announced on site. None of them held a move record. The tournament's input system went down at 23:40, the data stayed in a local buffer, and it was never pushed to the server. A real tournament, a real result, an empty file.

It is a silent failure, and in chess it is far more dangerous than in football. When a data system breaks at a football match, I still have video, still have the referee's report, still have the minutes played. In chess there is no reliable footage: a move exists only if someone wrote it down. Losing the record is losing the game.

That night I did not file. I sent the desk one line: not enough data to conclude, and I will not guess.

Chess is the most data-rich sport I have ever covered

The paradox is this: chess carries the highest level of verifiability of any sport I have worked on. Every move can be checked against an analysis engine and expressed in centipawns, the unit that measures the value of a position. That produces the metrics any data journalist must know by heart: ACPL, average centipawn loss per move, where lower is better; the share of moves matching the engine's first choice; and the evaluation bar tracking the course of a game.

In football I still tell colleagues that xG does not replace emotion, it explains why our hearts race. ACPL does exactly that job on a chessboard: it does not tell the story, it points to where the story broke.

Vietnam already has the material to tell it with numbers. Le Quang Liem won the men's blitz event at the 2026 World Blitz Championship in Khanty-Mansiysk, Russia, according to data published by the International Chess Federation (FIDE), a milestone that put Vietnamese chess on the map of elite events. Nguyen Ngoc Truong Son was among the youngest grandmasters in the world when he earned the title at fourteen. The next generation, including Tran Tuan Minh, Pham Le Thao Nguyen and Vo Thi Kim Phung, has held steady positions in regional rankings.

Yet most of what Vietnamese readers see about these players is still storytelling: a win is called inspiration, a loss is called a dip in form. That style is not wrong, but it cannot be verified, and so it expires in about three months.

Four data systems that cannot be mixed

This is what I repeat whenever I train young reporters: chess has four official time-control families, classical, rapid, blitz and bullet. Data from one cannot substitute for another. A player rated 2,700 in classical is not guaranteed to beat a 2,500 blitz player in bullet, and conversely a brilliant blitz player can collapse in a six-hour classical game.

The same applies to over-the-board and online results. Over-the-board outcomes do not extrapolate directly to online play. Many Vietnamese articles merge the two sources and then draw conclusions about a player's level from a streak on an online platform. Statistically, that is an input error, not an interpretive one.

The third source to separate is the rating table. FIDE's official Elo list is published periodically, while live ratings update game by game. The two numbers differ in nature and in timing, and should only be placed side by side when both carry an explicit timestamp.

The three-source rule and the trap of three copies

My rule is simple: before any conclusion, every number must appear in three independent sources. One source is an official document from FIDE or an organising federation. One is a live rating table. One is an archived game database for cross-checking moves.

But there is a trap I once fell into, and it cost me a week of corrections. Three different outlets published the same figure, I counted them as three sources, when all three had copied the same organiser's scoreboard. Three copies are not three sources. If three data streams flow out of one pipe, they will fail in the same way.

Since then I verify provenance before I verify quantity. The right question is not how many outlets reported it, but how many independent measurements exist behind those outlets.

Eight analytical dimensions and one empty column

Back to that night's file. When I ran the deep analysis routine across eight dimensions, from game technique, player data, tournament system, competitive landscape, rules and governance, risk, public narrative, through to the industry's transmission chain, the output was eight identical blank rows: insufficient information to assess.

Numbers are asceticism: you must give up convenience before you can see the truth.

What is frightening is not the eight blank rows. What is frightening is that an automated pipeline can read eight blank rows as eight completed checks, then conclude the game carried no risk. Finding no evidence in a failed extraction does not mean no evidence exists.

In my trade, the gravest mistake is not missing data, it is letting an empty file look like a clean file.

So I label all eight dimensions as unassessed rather than low risk. To a reader those two lines look almost identical. To a decision-maker they are worlds apart.

Anti-cheating: where statistics must outrun rumour

There is one analytical dimension where I never allow myself to speculate, even with complete data: allegations of cheating. The reason is practical. A name attached to suspicion follows that person longer than any correction.

Chess has anti-cheating detection built on statistical models, and online platforms have announced waves of account bans. The affair between Magnus Carlsen and Hans Niemann at the Sinquefield Cup in September 2026 showed what happens when an allegation outruns the statistics: a global controversy erupted while most readers had no way to verify the underlying data.

The professional lesson is not to take a side, but a principle: when discussing cheating, speak only in models, thresholds and samples, with an explicit falsification condition. If index X exceeds threshold Y in sample Z, the conclusion holds. Anything else is rumour with a name attached.

Age and the curve that never lies

For older players I keep a habit that has become a rule: every piece about form must include a comparison with that same player a season earlier. Without a baseline there is no assessment.

Age is the only variable that never lies.

The decline curve in chess is not the one in football. Speed goes first, accuracy goes later. In bullet, the gap between a twenty-five-year-old and a forty-year-old usually appears in the last moves of a game, with under thirty seconds on the clock. In classical chess that gap narrows, sometimes disappears, because experience compensates for reflexes.

Based on my experience following matches across many seasons, I always separate two metric families before writing a single line: the speed-dependent family and the position-dependent family. Mixing them is the fastest way to write something wrong that still looks data-backed.

Correlation is not causation

This is where I routinely disagree with the rest of the newsroom.

A rising average Elo for a national team does not mean that team will win a medal. Elo is a cumulative measure; a medal is the outcome of one week of play, where the schedule, the opponents in your pool and the team format matter enormously. Likewise, a high draw rate does not automatically make a tournament dull. To judge an event's pull, I need additional data: viewership of long games, the average number of moves before a draw, and the anti-draw rules the organisers enforce.

Another example: the majority assume that to win you must chase whatever opening theory is fashionable. I once tested this the other way round and found that the players who win most are not those who match the engine most often, but those who best understand the positions the engine undervalues.

The truth is simpler: a beautiful correlation cannot pay the bill for a wrong conclusion.

Signals to watch in the next cycle

For Vietnamese chess I am tracking three signals, and I state the falsification condition for each so readers can check me later.

First, the age flow. If the number of Vietnamese players under twenty appearing regularly in international open events rises for two consecutive seasons, that indicates the development system is working. If that number flatlines while the average Elo still rises, the results are coming from a small group rather than a foundation.

Second, data infrastructure. A country can produce grandmasters without a public game archive, but coverage of them will remain dependent on narration. Once a Vietnamese-language chess database with clear provenance for every game exists, the quality of domestic analysis will change within a season.

Third, how domestic tournaments handle short draws. If organisers add minimum-move rules, the share of watchable games rises, and the data behind coverage becomes more trustworthy.

I still remember the game I could not write that night. No move was saved, yet it left me with something more useful than any game I have watched: a hard rule. If the input data is empty, the writer stops. People call that a shock; I call it data that has not yet been read.

The data era of Vietnamese chess will not begin with a medal. It will begin on a morning when someone sits down, opens the archive, and writes the correct source next to every number.