Evidence First, Story Later: A Chess Report That Began With an Empty File

Hook

Singapore, 12 December 2026. After fourteen games the scoreboard read 7.5–6.5. Gukesh D had beaten Ding Liren and taken the FIDE World Championship. That night two files were open on my laptop — one with the moves, one with six years of FIDE rating lists stacked in columns. In the same week a third file arrived. It had eight rows, and every row said N/A.

It was not anybody's scoresheet. It was a source file with no article title recorded, no author, no date. Domain: chess. Everything else blank.

The temptation to write is strongest at that exact moment. The reader wants a name, an opening, a "revelation" about some federation decision. I opened the ledger expecting an answer and found a better question. Before the argument, the audit trail: who recorded this, and when.

Context: the news is not what happened, it is what got logged

Chess journalism is built from three places — the crosstable, the rating list and the circular. Each runs on its own clock, and each trails the others by weeks. When somebody says a player "is back in form", what has usually happened is that three readings from three different clocks have been folded into one sentence.

In 2026, in Ahmedabad, I logged 3,412 games from Gujarat school and club events into a database, coding club affiliation, tournament level and opening family. A sociology MA earns its keep here: I was coding the structure around the player, not the player. That work showed where local "prodigy" claims break — eight wins and a junior is declared the next Anand. So I put the rule on myself that year: nothing called "the next" on a sample under 50 classical games.

This article is that rule under test. When the source file arrives empty, a writer has two doors — invent a room, or report that the room is empty. The second door is read less. It is also the only door that lets tomorrow's correction repair today's claim.

Core: six clocks ticking inside one report

Clock one: rating-list latency. FIDE's standard rating list publishes around the first of each month. A title decided on 12 December 2026 does not appear in that day's rating — it waits for the next list. The gap between an event's final round and the list that reflects it is routinely several weeks. Latency is a stat line, and somebody has to read it. A commentator who ignores that gap turns stale numbers into proof of a fresh win.

Second thing that must travel with any rating: the K-factor. FIDE runs it in three bands — 10, 20 and 40. New players and under-18 juniors move fast; anyone who has crossed 2400 moves slowly. The same eighteen-point jump is two entirely different stories for two different players. Every rating is a promise about the past, dressed as a forecast.

Clock two: separate ledgers for separate time controls. In 2026, during Tata Steel Chess India's first Kolkata rapid-and-blitz invitational, I built a move-time model from 1,120 rapid games that included Viswanathan Anand and Hikaru Nakamura. I refused to let blitz games into the classical baseline. The sheet showed Anand's time variance rising sharply after move thirty in faster controls, and the model flagged two of his three time-pressure errors before they were annotated.

The clock was the variable nobody had bothered to log. Every report needs the control, the increment and the tiebreak type written down. An Armageddon game and a 90-minute increment game cannot share a column. Rapid and blitz do not need to be ranked below classical — they need their own register, held to the same rigour, because that is where the Indian audience is actually watching.

Clock three: the server's clock. In the 2026 FIDE Online Olympiad final, server failures disrupted the India–Russia match. I logged 96 games across the event and built a latency-adjusted sheet separating genuine blunders from disconnection artifacts. My report advised no protest on one contested position because the sample was too thin. FIDE declared India and Russia joint winners. I never called that final a clean result, and my rule since then has been simple: no single online game counts as evidence of form for six months.

Clock four: board order, pre-registered before round one. With India-2's open team at the 2026 Chennai Olympiad I insisted the board-order rule be written before the first round and not revised after results. The model, back-tested on 600 Olympiad games, put Gukesh on board one despite his age. He scored 9/11, took individual board-one gold, and India-2 won open bronze. The same sheet warned against moving Nihal Sarin down after one bad round. Match points and game points are two different accounts — rebuilding a board order off one loss is the most expensive habit in team-event coverage.

Clock five: ACPL and performance rating are two faces. Average centipawn loss measures how much a player gave up against the engine's best move. Performance rating measures the level a tournament represented. Different questions. A player can post a 2,650 performance at 15 ACPL and another at 45 — the first is stable, the second is living on the edge of results. Preferring stability over noise is an editorial decision, not a statistical luxury.

Clock six: the circular's clock. In Indian chess, administrative paper arrives either ten days before a player's career or ten days after, and that delay does the damage. From Anand's 2026 Grandmaster title through India's double gold in Budapest in September 2026, every leap rests on a paper chain — the Velammal school pipeline, Tamil Nadu's state awards, public-sector patronage — and all three need checking against records rather than press releases. A suspension becomes news when it costs a selection, not before.

What the eight rows actually demanded

The empty file carried eight dimensions — technical, player data, tournament system, competitive landscape, rules and governance, risk, public narrative, industry transmission. Every row asked for the same three things: date, time control, source. Where none of the three exists, no conclusion exists.

That is not a failure. That is a guardrail. The model was clean. The match was not. I kept both.

Contrarian: the empty file is safe; the full file is dangerous

The comfortable belief is that a missing source is journalism's biggest problem. My experience is the reverse. The bigger problem is the report where every column is filled and no column carries a source. Print a figure to one decimal — 2,741.3 — and the reader accepts it, because specificity looks like authenticity.

Three traps recur. First, correlation dressed as cause: a junior's rating rose and he got a new coach the same year, so the coach gets the credit, though both events merely shared a calendar. Second, cross-contaminated time controls: explaining a classical career off a blitz evening. Third, cross-contaminated eras: folding the 2026 Chennai loss and the 2026 Singapore win into a single redemption arc when eleven separate ledgers sit between them.

And a fourth trap catches the cleverest writers — over-caution. Caveat every sentence and no claim survives. The cautious writer is never wrong because he never lands. The fix is pre-registration: publish the control, the threshold and the date you will grade yourself against, then publish the call.

Takeaway

Watch three things next round. For any rating-based claim, ask for the list date and the latency. For any online result, ask for the disconnection count. For any administrative announcement, ask whether the circular reached the district association. What was not logged was not played — as true in chess as in a ledger. The question is no longer what sat inside that empty file. It is this: when the next one arrives full, which number will you reconcile it against first?