When the Database Is Empty: A Chess Analyst Faces the Silence

When the Database Is Empty: A Chess Analyst Faces the Silence Three in the morning, Shanghai time. The city had not fully gone to sleep, but my study was silent as a sheet of paper. On the screen, a spreadsheet lay open. The header row was ready: player name, Elo rating, head-to-head record, win rate, share of moves matching the engine. Every cell of data, not one of them held a number. They were empty, like a chessboard someone had swept clean of pieces before I could sit down. That was the moment I understood: in the modern chess world, the most frightening thing is not a bad number. The most frightening thing is the silence of the number. A wrong figure can be corrected. A blank cannot. It sits there, polite and dangerous, waiting for someone to invent an answer to fill it. There are players the world forgets, but the data never forgets them. That is what I still tell my younger colleagues. But tonight, I myself had to face the reverse paradox: when the data remembers nothing at all, what gets forgotten is the truth itself. And I once believed in emotion — until a number knocked on my door at three in the morning. THE LESSON OF A DATASET THAT CONTAINED NOTHING I sat back, brewed another cup of tea, and did what I have done for forty years whenever I feel lost: I reread my own process. In chess analysis, everything begins with a step called extraction — pulling raw data from a source. If that step returns correctly, I have ingredients to cook. If it returns empty, I have only an empty pot and a flame. That night, my pot had nothing. No player name. No tournament name. Not a single footnote. No title. No time control. Not one number to cross-check. The only thing left was a bare label: "chess." Just one word. As if someone had handed me the map of an entire continent without a single place name written on it. I tried to list honestly what I could draw from that label. A topic label tells me the sport, not the subject. It cannot distinguish a single game, a player's technical profile, opening preparation, or the technical trend of a whole event. Four possibilities. No way to choose. That is the truth, however much I might want it to be clearer. My trade taught me that when the data runs dry, the analyst must have the nerve to say one thing aloud: "Insufficient information, no conclusion can be drawn." It sounds simple. But saying it in front of a newsroom waiting for copy, an editor pushing a deadline, readers anxious because a tournament is under way — that is a struggle with oneself. CONTEXT: THE DATA MACHINE OF MODERN CHESS To understand why a data gap is so dangerous, one must understand how completely chess has become a vast data machine over the past two decades. Every professional player lives within four main numbers. First, the classical Elo rating — a coefficient reflecting relative strength, updated with every official event published by the World Chess Federation. Second, the rapid rating. Third, the blitz rating. Fourth, the performance rating — a temporary figure expressing the level corresponding to a player's results in a specific event, regardless of the base rating. Those four numbers do not sit still. They run continuously. Some systems track ratings live, updating with each game as soon as the pieces are reset, so a fan anywhere can know whether a favourite is rising or falling after each decisive move. Beneath those four public numbers lies a far denser submerged layer. There are game databases holding millions of games across more than a century, from legendary championship matches to minor games digitised late. There are specialist bulletins launched weekly, systematically organising the results and events of the global chess scene. There are online platforms recording every game of tens of millions of players, generating a new layer of data: large-scale amateur data. And above it all sits the engine — chess machines stronger than any human champion. The engine evaluates every position on a scale. From that scale, a range of derived metrics is drawn, the most famous being average centipawn loss per move — a measure of move quality. The lower, the better. A player with a low loss figure makes few mistakes, or is rarely caught by the engine. Another derived metric is the share of moves matching the engine — what percentage of a player's moves coincide with the machine's number-one choice. This figure was once a basic fault-finding tool in analysis circles. But it is also a double-edged sword, because it later became the focal point of the game's biggest suspicions. Then there are countless smaller things outsiders never see. A move never before appearing in a database is called a novelty. The frequency of an opening line within a game repository helps measure the popularity of that system. Moves per game, thinking time per move, the moment one falls into time trouble — all of it is recorded. An analyst of my era has far more weapons than the previous generation. But the more weapons one has, the easier it is to forget a foundational truth: all those weapons are trustworthy only when their provenance can be verified. And at the exact moment a data machine falls silent, that entire trust is reset to zero. CORE: ANATOMY OF A TYPICAL DATA INCIDENT Let us treat that night as a clinical case to dissect. The patient is a chess analysis process. The symptom is an empty dataset. The question: what actually happened, and what is the most honest response? First, I must distinguish two entirely different situations that look identical at first glance. Situation one: the source genuinely contains no chess content. This could be a bare result stub — only a score table with not a single line of analysis. Or a headline leading to an empty page. This case is rare, but it exists. Situation two — and in my experience far more common: the source has content, but extraction failed. The page was blocked. The original sat behind a paywall. The parsing engine hit a syntax error. Or the source was not text at all, but a video or an image with no words to read. The difference between these two situations is not academic. It determines the entire next move. If it is situation one, I should find another source. If it is situation two, I should fix the process. Confuse the two, and a person will hunt for an article that does not exist while the real fault lies in the machine in their own room. Watching this trade for forty years, I learned a rule: when extraction returns entirely empty while the topic label survives intact, the highest probability is a fault in data retrieval, not a meaningless source. Because chess articles almost always name at least one player, tournament, or federation in their first two sentences. No name at all means the first link in the chain has snapped. Now imagine the wrong response. Facing an empty table, a discipline-poor analyst will do what? They will fill. They will take general chess knowledge, pack it into the void, and construct an analysis that reads very smoothly. An imaginary player. An imaginary Elo. An imaginary head-to-head. An imaginary game. In most fields, such errors might pass. In chess, it is a professional crime. For chess is among the most verifiable sports on earth. Elo ratings sit in the official list of the world federation. Games sit in public repositories. Prize funds are published. Head-to-head results anyone can look up. A fabricated number in chess is unlike a fabricated number in idle gossip — a reader can look it up, cross-check it, and expose it within thirty seconds. That is why, in my trade, rule number one is not "write well." Rule number one is "do not fabricate." Fabricating in a field where everything is verifiable is like digging a grave for your own credibility. PROVENANCE: THE UNWRITTEN LAW OF DATA WORK There is one detail I want to dwell on at length, because it is the spine of this whole story: every number put forward must be traceable to a source. In chess, sources carry wildly different weight. An official press release from the world federation is pure gold. A long-standing specialist bulletin carries high credibility. An article in a major sports desk has its own weight. A post on a free platform is far lighter. And an anonymous forum comment has almost no evidentiary value. What I want to stress is this: source triage is not a formality. It must happen before the analysis, not after. Because if you analyse first and check the source later, you have already placed your entire conclusion on an unconfirmed foundation. When the foundation collapses, ordinary people are disillusioned, but professionals lose their credibility. In my practice, every number shown to readers must meet one condition: clear provenance. For Elo, that is the official list. For live ratings during an ongoing event, that is the tracking platforms. For games, that is the database. For platform statistics, that is the platform itself. If a number lacks such provenance, my rule is to tag it "pending verification." Four little words, but they have saved me from more than a few embarrassments. I remember a time when the chess world was shaken by a famous cheating case involving two top players. That story later became a lesson in how data and oversight processes are weighed on the scales. People did not argue with emotion. They argued with algorithms, probabilities, engine-matching frequencies, and statistical anomaly detection models. However much each side's view differed, the common floor of the debate was numbers — and numbers must be traceable to a source, or they are merely accusations. That is the biggest lesson I drew for my own trade. Once you step into the territory of accusations, even technical ones, everything must be precise beyond rebuttal. Because after a false accusation, one loses not just an article. One loses the honour of the person named. AND THE COUNTERINTUITIVE ANGLE: ABSENCE IS NOT EVIDENCE Now comes the part I consider most important in the whole story, the part I have only truly absorbed in recent years. When an extraction step fails and returns empty, a person's natural reflex is to read that emptiness as a positive signal: "I see no problem, so there is no problem." This is the subtlest trap, and the deadliest, in the entire analysis pipeline. Picture it. An automated system scans data and reports: no issues detected. The operator breathes a sigh of relief and records "checked, clean" in the log. But in truth, the system never read anything. It was empty from the very retrieval step. The "clean" recorded in the log is not a conclusion. It is merely a blank mislabelled. In chess, this confusion can play out at several levels. A tournament with no news is not necessarily a peaceful tournament — it may be one no reporter attended. A player absent from the bulletin is not necessarily retired — the update may simply be missing. A missing statistic is not necessarily a zero — it is merely a number no one has filled in. When the stadium is empty, the true value of a person begins to speak. But an empty stadium can also be one that no one has opened yet. Telling these two apart is the boundary between an analyst and a guesser. And here is what I want to send to the younger people in this trade: the truth is that in a data pipeline, a completely empty result, while the topic label survives intact, is a strong sign of a technical fault, not a conclusion about content. It is like a broken alarm light. The absence of a bell does not mean the house is safe. It may simply mean the bell has been unplugged. I once watched a young colleague draw a conclusion about an entire chess trend based only on an empty dataset and a few enthusiastic guesses. The article ran. Readers read it. Then a few days later, the source was recovered, and the whole conclusion collapsed like a sandcastle hit by a wave. From then on, he learned something I can only pass on through story: the absence of evidence is not evidence of absence. In a sport where every number can be looked up, that is not idle philosophy. It is a survival rule. A CONFESSION TO DATA I was born in Vietnam, work in Shanghai, and report on chess for a market other than my homeland. This trade taught me that geographical distance is a lens that can bend the truth in both directions. Homesickness makes one sentimental. Distance makes one prone to inference. Both are enemies of the number. There was a time I believed in emotion. I believed a player shone in a game because he had a sacred moment. I believed a tournament was exciting because it had charm. But then the numbers began knocking. They knock at three in the morning, when I sit alone, and they do not care what I believe. A game that looks emotional is an upset, but open the engine and it is a chain of accumulated errors by one side. A defensive feat looks legendary, but the numbers show the opponent simply missed chances. Emotion builds the story, but numbers build the truth. I do not deny emotion. I simply place it correctly. I light a candle for data. But I always let the flame of emotion light the question. Emotion helps me know what to look for. Numbers tell me whether what I found is real. So when the dataset is empty, my emotion shouts: "Write, the readers are waiting!" But my discipline replies: "A conclusion built on zero is a wrong conclusion, however well written." That is a conversion without regret. I no longer want to be a teller of chess fairy tales. I want to be the chronicler of numbers. And an honest chronicler must be able to say: "This page is still blank, I cannot yet write anything." DATA CAN RESURRECT A LEGACY, BUT IT CAN ALSO BURY A TRUTH I once spent years using data to restore the names of players abandoned by the media. People who gave a lifetime to the board but never got a single line of memoir. For them, data is a resurrection. A record I rebuild can bring a name back to public memory. But precisely because I understand the resurrecting power of data, I also understand its dark side. Data can resurrect a legacy, and it can also bury a truth alive. One only needs to embellish a number, assign it a meaning it does not have, and the name mentioned will live forever with a distorted prejudice. For a player, the worst thing may not come from a defeat. It comes from a false article. An article that assigns him a number that is not real. An article that turns an ordinary loss into a moral tragedy. An article that turns an unproven suspicion into a verdict. So, in this trade, I set myself a strict standard: better a day slow than a number wrong. Better to cut an article in half than to insert a half without evidence. Better to say "cannot yet comment" than to say "I think it is so." That is also why I always reserve the end of my process for a second check. The first pass I read to write. The second pass I read to find errors. And the third pass I read to ask: if this number were taken away, would my reasoning still stand? If the answer is no, that conclusion does not deserve to exist. TIME PRESSURE AND THE TRAP OF LIVING MILESTONES One peculiarity of modern chess makes this problem more urgent than ever: its timeliness shifts very fast. An ongoing tournament can change shape after every game. A player on a rising rating curve can cross a milestone after one win. A record thought unbreakable can fall in a single afternoon. That means a wrong analysis can become obsolete within days, and in that window it spreads, influences how the public sees a person. If my source is an in-progress event or a just-set rating milestone, the shelf life of the analysis is counted in days, not months. Even in hours. And in that brief window, the lack of an accurate timestamp can turn a correct article into a meaningless one. So, in my process, a timestamp is mandatory. No date, no real-time analysis. That is not perfectionism. It is the condition for an article to still mean something tomorrow. AND THAT IS WHY I STOPPED Back to that night, to the empty dataset on the screen at three in the morning. After walking through the whole process in my head, I arrived at a decision that seems simple yet is the hardest to voice in my trade: to stop. I did not write the article. I did not build an imaginary player profile. I did not assign anyone a guessed Elo. I did not retell a game I never watched. I did not turn the blank into the clean. Instead, I did what I have told my students for forty years: when you hold nothing, be honest that you hold nothing. Because readers do not need a fast answer. They need a true one. I drew up a short list of what was needed to unlock a decent analysis. A title and publication date. Source name and type. At least one player name. Tournament name, round, and time control. One verifiable number — rating, result, prize fund, or viewership. And the original article's stance. That list is short. The recovery cost is low. Just one name, one tournament, or one federation would open nearly half the analytical dimensions. Chess is a surprisingly data-rich sport: one name plus one tournament is often enough to position a player on the rating map, on the career age curve, and on the qualification path of the world championship cycle. But until those exist, I keep my silence. A LIGHT IN THE EMPTINESS Hearing this, you might think that night of mine was a failure. But no. In a sense, it was one of the most valuable nights of my many years in the trade. Because an empty result, read correctly, is itself a signal. It tells me precisely which link in my chain broke. It turns a silent incident into a control signal. It teaches me that the greatest value of a system is not that it answers correctly, but that it knows how to say "I cannot read" instead of saying "everything is fine." An old lesson that never ages: the most dangerous system is not the one that sounds false alarms. The most dangerous system is the one that stays silent while it is blind. And this may be what I most want to send to everyone in chess analysis, whether twenty or fifty. The glamour of this trade does not come from always having an answer. It comes from knowing when you are not allowed to answer. END: THE NUMBER THAT KNOWS HOW TO BE SILENT Near dawn. I folded the empty spreadsheet, switched off the screen, and told myself I would return to work once the source was recovered. On the way to bed, I thought about the name I would have written that night — if the data had agreed to speak. Someone, somewhere, had just passed through a game they themselves may already have forgotten. And in the strange way of this trade, I am the one who remembers for them. Not because I am better than anyone, but because I chose to be the keeper of numbers. There are players the world forgets, but the data never forgets them. And when the data falls silent, an honest practitioner must have the nerve to stand inside that silence, rather than sow noise into it to please the ear. For a game can be forgotten. A wrong number can live forever. And the question left for all of us who write about the board is not how fast we write. It is: when the page before us is still blank, do we have the courage to write nothing at all? Silence, on some nights, is the most honest analysis of all.