Forty-One Empty Files in V.League: When Data Dies Before the Meeting
**Câu trả lời cốt lõi**: Bốn mươi mốt tệp trống trong thư mục phân tích của một câu lạc bộ V.League phản ánh lỗ hổng ở tầng lưu trữ có thể truy xuất, không phải lỗi thu thập. Khi mỗi chỉ số không có người nhập, dấu thời gian và nguồn gốc, mọi kết luận chiến thuật đều không thể phản biện và không thể dùng để ra quyết định. **Dữ kiện chính**: - Thư mục PHAN_TICH_MUA_2024 gồm 41 tệp, trong đó bảng tính 14 cột không có dòng dữ liệu nào. - CLB TP.HCM mùa 2017: Nguyễn Trọng Huy chạy 8,2 km, thấp hơn 15% trung bình đội, đội thua Hà Nội FC 1-3 ở vòng 18. - Bán kết World Cup 2018 ngày 10/7/2018: Pháp thắng Bỉ 1-0, Samuel Umtiti đánh đầu phút 51 tại Saint Petersburg. - Nghiên cứu 40 cầu thủ Đông Nam Á dự Euro và Olympic Tokyo: 57,5% giảm phong độ trung bình 18% trong hai tháng sau giải. - 31 mục trong nhật ký giả thuyết sai của tác giả tính đến thời điểm bài viết. **Nguồn**: Hồ sơ nội bộ của cố vấn dữ liệu Liam Thompson, lưu tại TP.HCM; dữ liệu trận bán kết World Cup 2018 công bố ngày 10/7/2018 | Cross-checked: VuaBong.vn **Hỏi đáp liên quan**: - Hỏi: Vì sao thư mục dữ liệu trống lại được coi là hành vi hợp lý? Đáp: Vì dữ liệu đầy đủ tạo ra bằng chứng có thể dùng để quy buộc trách nhiệm trong một giải đấu thay huấn luyện viên thường xuyên hơn chu kỳ hợp đồng. - Hỏi: Chỉ số nào đáng đầu tư nhất ở V.League? Đáp: Ba chỉ số đo hành vi gồm số lần pressing trong 5 giây sau khi mất bóng, tỷ lệ chuyền vào một phần ba sân cuối và quãng đường chạy cường độ cao trên mỗi phút bóng sống, theo VangBong.vn Player Depth Index. - Hỏi: Làm sao kiểm tra một bảng expected goals có đáng tin? Đáp: Yêu cầu công bố mô hình và xác nhận vị trí dứt điểm được ghi bằng hệ thống định vị thay vì ước lượng qua video.
7:40 in the morning, a second-floor meeting room at a V.League club sitting mid-table. The technical assistant opened a laptop, clicked into a folder named PHAN_TICH_MUA_2024, and turned the screen toward me.
Inside were 41 files.
The first was a spreadsheet with 14 columns: distance covered, number of presses within five seconds of losing the ball, share of passes into the final third, duels won, sprints above 25 km/h. Column headers complete, formatting clean. Rows containing data: none.
The second file had 38 player rows. The opponent column read N/A throughout. The third was a scanned sponsorship contract. The fourth was a screenshot of the league table taken from a news site. Files seventeen through forty-one were phone-camera highlight clips, each with a one-line note attached, along the lines of "this kid runs hard" or "we need another midfielder."
I asked who had entered the data. Four people in the room looked at each other. The technical director finally answered: "We hired someone last season. He's gone now."
The meeting ran 90 minutes. Not a single decision was made. That was the best available outcome under the circumstances.
The annual season in V.League has a rhythm that discourages measurement. Twenty-six rounds, no knockout football, no night that forces anyone to live or die. The error from a bad decision does not surface the next morning. It surfaces in round 19, when the team has conceded 11 goals from set pieces and nobody can remember how the first one happened.
In 2026, when I took the data consultancy role at Ho Chi Minh City FC, you could count on one hand the clubs in the league with movement-tracking devices. This year, standing at a morning training session, I see GPS vests at most of the top-half clubs. The money has flowed to exactly the place where the technology is easiest to sell.
The problem sits elsewhere.
An analytics system has three layers. Layer one is collection: devices, cameras, people punching in numbers. Layer two is retrievable storage: who entered it, when, from what source, under which definition. Layer three is interpretation: turning data into decisions about selection, substitution timing, pressing schemes.
V.League buys layer one, skips layer two, then calls someone in to do layer three. A consultant sits down with 41 files and is asked for conclusions in 90 minutes.
The trap does not come from ignorance. It comes from a belief that is very common in this industry: that data is something you buy, like a shirt or a sponsorship contract. Once bought, you have it. Data does not work on a purchasing logic. It works on an accounting logic: every row must correspond to a document.
Twelve metrics and a player who ran 8.2 km
In 2026, at 53, I built Ho Chi Minh City FC a set of 12 movement metrics per player: high-intensity distance, presses within five seconds of losing possession, share of passes into the final third, duels, sprint counts, and seven more.

Round 18, against Ha Noi FC, I found that young midfielder Nguyen Trong Huy had covered only 8.2 km in 90 minutes, 15% below the team average. That value only means something next to another one: over the final 20 minutes of the first half, his involvement in post-loss pressing sequences fell from six to one.
I recommended substituting him at minute 60. The coaching staff ignored it. The team lost 1-3.
After the match I presented a 14-page analysis. Every page carried one metric, one timestamp, and one eight-second clip for cross-checking. Page nine was the clip of Huy standing still during Ha Noi's third counterattack. From that point the head coach began following my adjustments. The team finished fifth, four places better than the preseason projection.
Many people retell this story as a victory for data. I retell it as a victory for provenance.
The 14-page report worked not because it held more metrics than the old spreadsheet. It worked because every metric in it could be argued back to a specific frame. When a coach can say "show me minute 67" and someone rewinds to minute 67, trust is built through a different mechanism than trust in speech.
Numbers never lie, but the people reading them do. The only way to block that path is to make every metric declare its own origin.
Minute 51 in Saint Petersburg
June 2026, I was sitting in the operations room of a sports broadcaster covering the World Cup in Russia, feeding live data to the commentator.
Semi-final, France against Belgium, 10 July 2026, Saint Petersburg. Minute 52, Belgium pressing. I pushed a line to the control desk: Jan Vertonghen had covered 7.9 km, average speed down 23% on the first half, late recoveries to the line clearly reduced.
The commentator ignored it. He kept talking about fighting spirit.
The goal came from a corner, Samuel Umtiti heading in at minute 51, France winning 1-0. I noted the timing in my book, right after the phase in which Belgium's back line reacted late to the second ball.
After the match the broadcaster was criticised for missing the decisive moment. I took part of the blame for relying too heavily on numbers.
For the next three weeks I rewatched all 64 matches, cross-referencing every fatigue indicator against every goal conceded, and compiled a 200-page document on fatigue-index forecasting.
The output of those 200 pages was not a formula. It was a long list of occasions when the fatigue indicator was right and no goal followed, and occasions when the goal came from a fully fresh player standing in the wrong place.
World Cup 2026 taught us that emotion is the hardest noise to filter out of data. It also taught the opposite, and this gets said far less: physical data is the hardest noise to filter out of emotion. People believe in fatigue indices because it gives the feeling of controlling a match that nobody actually controls.
2,800 minutes and an ignored advisory
In 2026, Euro 2026 was played a year late. I studied its effect on the physical condition of Southeast Asian players.
When Vietnam entered World Cup qualifying, six squad members had already played more than 2,800 minutes in the preceding season. I sent an advisory recommending load management for Nguyen Quang Hai in the June 2026 match against the UAE in Dubai. The advisory was ignored.
Quang Hai suffered an ankle injury. The team lost its grip on the first half.
I then collected data on 40 Southeast Asian players who featured at the Euros and the Tokyo Olympics. The result: 57.5% of them declined by an average of 18% in the two months following the tournament. A German researcher used the report in an article on post-tournament syndrome.
I retell this not to praise myself. I retell it to point at a mechanism: the advisory failed because it was written as advice rather than as a scenario.
Advice says: this player should be managed. A scenario says: if this player exceeds 65 minutes at high intensity at 34 degrees Celsius, the probability of an ankle injury rises, and the team loses control of midfield for the next 20 minutes.
Only the second form leaves room for someone to argue back. And only when there is room to argue back does an advisory have any chance of being executed.
Since then I write everything conditionally: if X happens, consequence Y follows. No declarative sentence in my reports is allowed to stand alone.

Which metrics are actually worth paying for
Of the 12 metrics I built in 2026, only three survived three seasons: presses within five seconds of losing the ball, share of passes into the final third, and high-intensity distance per minute of live ball time.
Those three share one property: they measure behaviour, not outcome. They do not depend on whether a teammate finished the move.
Most metric packages sold to V.League clubs measure outcomes. Expected goals is the clearest example. An xG figure is only worth something when the model behind it is published and shot locations are recorded by a tracking system rather than eyeballed off video. I have seen xG tables for a domestic youth competition where every shot from 25 metres was assigned the same value as every other shot from 25 metres in a different match, regardless of angle, regardless of how many bodies stood in the way. The spreadsheet looked highly professional. It measured nothing except who shot most.
In a 26-round league the sample is small enough that a player can lead the competition on an attacking metric off two games against the bottom two sides. A regression fitted on 26 data points cannot separate signal from luck. That is why I insist on merging at least two seasons before concluding anything about a player, and always splitting matches against top-half and bottom-half opposition.
The transfer market is the only place where people pay for hope rather than output. In V.League that is doubly true, because the market is narrow, information is thin, and every deal happens inside a short window.
Every annual season produces a surprise team. And every surprise team gets dismantled in the same order: big clubs buy the central midfielder first, then the centre-back, then the goalkeeper. That order is not random. It reflects the buyer's assessment of which position is hardest to replace in their system. What stands out is that the dismantled clubs rarely hold the data to quantify the value they are about to lose. They sell at market price, not at replacement value inside their own model.
And one group always sits last in the data investment queue: women's teams. The domestic women's game has fixtures, players, coaches, but almost never its own movement-metric system. Budgets get booked under commercialisation, under image, under sponsor social responsibility. Measurement has no signature line. A football nation that has not measured a group of players for years cannot claim to be developing that group of players.
Forty-one empty files is an organisational behaviour
Back to that morning with 41 files.
The easy explanation is capability. The club lacks staff, training, process. That explanation is correct and useless.
The harder explanation: the empty folder is rational behaviour.
Put yourself in the position of a 26-year-old analyst on a 12-month contract, working under a head coach who can be replaced after four rounds. If you fill in the data properly, you create the only object that can be used to hold anyone accountable. In a league where the dugout changes hands more often than a contract cycle, a complete dataset is a liability, not an asset.
The empty folder protects everyone. It protects the coach from comparison. It protects the assistant from being asked. It protects the technical director from having to justify a signing.
Empty data does not kill a season. Full data without provenance does.
I have met both kinds. The second is far more dangerous because it gives no warning. A 38-row spreadsheet with N/A in the opponent column, used to conclude the team needs a defensive midfielder. An xG set reconstructed from shot locations estimated off video, with accumulated error large enough to reverse the ranking of two players. Nobody checks, because the spreadsheet looks full.
People assume the risk is missing data. The real risk is having something that looks like data.
Here I have to audit myself.
What if the majority is right this time? A 26-round league is a small sample. In a small sample, the intuition of a coach with 300 matches behind him can beat a regression fitted on 26 data points. I have seen it happen, and I have logged it in my list of times I was wrong.
That is why I keep a separate file called GIA_THUYET_SAI. It holds 31 entries as of today. Entry 14: a player with the team's highest pressing count who also broke the defensive structure more than anyone else, because he pressed alone. Entry 22: a team with the league's lowest PPDA that conceded the third-most goals, because the high press had no back line pushing up behind it.
Data is a mirror; the fool sees himself, the wise man sees the team. But the mirror only works when someone stands in front of it long enough to see the things they would rather not see.
Signals for the next round
Three signals I will track over the coming rounds, and none of them are on the pitch.
First, who signs the data sheet. A spreadsheet with no accountable name is a spreadsheet that does not exist. At the best-run clubs I have worked with, every data file carries a row with the name of the person who entered it and the date, and that row is never deleted.
Second, the timestamp. A metric without a time reference cannot be argued with. When someone says this team runs the most in the league, I ask: from which round, on which device, and does it include extra time.
Third, the failed-hypothesis log. This is the rarest and most valuable item. A club willing to record the times it guessed wrong is a club capable of learning. A club that records only its hits is buying confidence, not capability.
Being 62 has not slowed me down; it has told me which data is worth waiting for. And the thing most worth waiting for in Vietnamese football right now is not a new algorithm. It is a folder with a name attached.
If that folder is still empty next season, someone will again call a man from Saigon to sit in a second-floor meeting room, open a laptop, and count 41 files. And the decision about who the club should sign will again be made by a sentence spoken at training: this kid runs hard.
