Vietnamese Chess and the Lagging Indicator Problem: Re-reading 4,812 Games for the Real Signal

Vietnamese Chess and the Lagging Indicator Problem: Re-reading 4,812 Games for the Real Signal

2:47 a.m., April 9, 2026. I rewound that game for the eleventh time, to count rather than to find the best move — the engine had finished answering on the second pass. The Vietnamese player on the left side of the board broke the pawn structure on his own initiative exactly four times in the final thirty-eight moves. Four times. In three of those four, the position tilted his way.

I saved that number into a file named do_tre — the lag. For seven years I have been chasing one thing: the gap between what the board says and what the Elo list says. I once believed in feeling, until a number knocked on my door at three in the morning. That night I understood that what I was reading belonged to a measurement system, and that system runs slower than reality.

Context: an archive nobody wants to keep

My archive started as a small habit. After every open tournament held in Vietnam, I downloaded the full PGN file and logged the results into a spreadsheet. The habit lasted seventeen years. As of June 2026, the merged file holds 4,812 games by 137 Vietnamese players, spanning 2026 to early 2026, mostly from international opens in Ho Chi Minh City, Quang Ninh and Hanoi, plus public data on Chess-Results and the periodic FIDE rating lists.

I call it an archive nobody wants to keep for a simple reason: it serves no broadcast. No sponsor pays to learn how often a twenty-two-year-old from Da Nang holds a balanced position past move forty-five. Yet those columns are the only thing that preserves what actually happened on the board.

Elo is an estimate drawn from the past. It updates game by game, and the size of each update depends on two things: how many games count and how strong the opponents were. A player who logs twenty games a year against 2,400-level opposition will hold a stable rating around that level, no matter how much stronger he has become in the training room. The rating list does not lie, but it answers a narrower question than the one fans are asking.

Two milestones show the lag most clearly. In 2026, Le Quang Liem became the first Vietnamese player to cross 2700 Elo. In June 2026, in Khanty-Mansiysk, he won the blitz section of the World Rapid and Blitz Championship — Vietnam's first individual world title in chess. On another branch, Nguyen Ngoc Truong Son won the World Under-10 Championship in Spain in 2026 at the age of ten, yet it took more than a decade for his rating to reflect what specialists had seen long before.

The evidence chain

In my dataset, between 2026 and 2026 a leading Vietnamese player averaged just 9.3 games a year against opponents rated 2,500 or above. Between 2026 and 2026 that figure rose to 21.6. But raise the opponent threshold to 2,600 and the density collapses below five games a year for most national team members. When the sample is that small, the estimation error is wider than the actual strength gap between two players in the same bracket — and that is the blind spot of every rating list.

The first consequence sits with the players themselves. Someone whose true level is around 2,650 but whose rating reads 2,590 gets seeded lower at open tournaments. Lower seeding means meeting strong opponents earlier, losing earlier, playing fewer quality games, and the loop closes exactly where it began.

The second consequence sits with organisers and club teams. Professional chess runs an employment market that behaves much like a transfer window: club leagues in China, Germany and the Czech Republic pay per-game appearance fees, contracts are signed season by season, and extension and release clauses are negotiated between seasons. The most common selection criterion remains the individual Elo. A lagging indicator is therefore pricing human beings directly. The transfer market does not buy the past. It buys what the data has already forgiven.

The third consequence concerns record-keeping infrastructure. Most games played by Vietnamese players in domestic events are not stored in a standard format, or survive only on paper scoresheets. A game that is never digitised does not exist in any model. This is why many young players improve for real yet leave no trace in international datasets — and by the time they draw attention, three or four developmental years are gone.

By way of comparison, India and Uzbekistan are the two cases I track most closely. India solved the small-sample problem by adding more domestically rated events, then pushing its fourteen-to-eighteen cohort into European opens at high frequency. Uzbekistan chose differently: concentrate resources on a small group, train deeply, and keep them competing in foreign club leagues. Both roads end in the same place — more quality games. In my file, Vietnamese players who log 25 or more rated games a year show average rating gains 2.4 times higher after five years than those who log fewer than twelve.

The cost of that path is not mysterious. A slate of four European opens a year — flights, lodging, meals and entry fees — typically runs into a few hundred million dong per player. Set against the budget of a youth training centre, that is a measurable, recoverable investment, provided administrators are willing to read the data table alongside the medal table.

There are players who have been forgotten, but the data never forgets them. My file contains names that never wore the national shirt, whose rate of holding balance against 2,500-plus opposition is higher than that of some Olympiad participants. On the women's side, Pham Le Thao Nguyen has held board one for years, and the depth behind her is a question the data table still cannot answer.

The counterintuitive angle

A familiar trap waits here: mistaking correlation for causation. SEA Games medals and Elo are two different currencies. Medals determine funding, salary structures and standing inside the training system. Elo determines invitations to international events. A player who wins gold in a regional arena may barely move his rating, because the opponent pool there is not strong enough to shift the calculation meaningfully. Conversely, someone competing steadily in European opens can gain points steadily while his trophy cabinet stays empty.

The result is two optimisation targets pulling in opposite directions. Coaches want medals. Players want points. Tournament organisers want names that sell tickets. Nobody is wrong in his own choice, but the system as a whole is skewed.

And here is the surprise. When I regress rating progression between 2026 and 2026 across the 137 players in the file, the strongest predictor is neither current Elo nor medal count, but an index I built myself: the share of games in which a player holds a balanced position past move forty against opponents rated 2,600 or above. That index explains more than most remaining variables, including win count. In other words, what forecasts the future is not whom you beat, but whom you refuse to collapse against.

The next cycle

The 2026-2027 cycle will answer three concrete questions. First, does the number of Vietnamese players logging twenty or more games a year against 2,600-plus opposition pass five? Second, is the current Under-16 cohort sent onto the open circuit early enough? Third, do domestic events add rounds that count for standard Elo?

I light a candle for the data. But I always let the flame of feeling light the way to the question. Because after all these tables, what I still want to know is not who is strongest. What I want to know is how many names are still waiting to be called — and whether this time we will count them earlier.