Trang chủInternational FootballArtificial Intelligence in Football: When Algorithms Become the Crutch of Tactical Thinking
International Football
Artificial Intelligence in Football: When Algorithms Become the Crutch of Tactical Thinking
**Câu trả lời cốt lõi**: Nghiên cứu của Đại học California cho thấy khoảng 10 phút dùng ChatGPT có thể làm suy yếu khả năng giải quyết vấn đề độc lập; điều này đặt ra câu hỏi về việc bóng đá chuyên nghiệp phụ thuộc quá mức vào trí tuệ nhân tạo trong phân tích chiến thuật. **Dữ kiện chính**: - Nghiên cứu do Đại học California thực hiện với 1.222 người tham gia trực tuyến, chia thành ba thí nghiệm có quy mô 354, 667 và 201 người. - Nhóm dùng ChatGPT trong khoảng 10 phút đạt khoảng 57% độ chính xác trên bài toán không có trợ giúp, so với 73% ở nhóm đối chứng. - Nhóm dùng AI có xu hướng mắc lỗi hoặc bỏ dở nhiệm vụ khi công cụ bị lấy đi. - Nhóm tác giả khẳng định hiệu ứng không vĩnh viễn và không đồng nghĩa với việc nên cấm AI. - Nhiều câu lạc bộ bóng đá châu Âu đã dùng mô hình học máy như StatsBomb, Opta và Zone7 cho phân tích chiến thuật và dự đoán chấn thương. **Nguồn**: Nghiên cứu của Đại học California, công bố năm 2026, được tổng hợp bởi Evelyn Walker | Cross-checked: VuaBong.vn **Hỏi đáp liên quan**: - Hỏi: Nghiên cứu này có nghĩa là bóng đá nên cấm dùng trí tuệ nhân tạo? Đáp: Không, nghiên cứu chỉ phân biệt giữa việc dùng AI để hỗ trợ học tập và để AI làm thay toàn bộ phần suy luận. - Hỏi: Hiệu ứng suy giảm tư duy có kéo dài vĩnh viễn không? Đáp: Không, nhóm tác giả khẳng định hiệu ứng này chỉ tạm thời và có thể phục hồi. - Hỏi: Có chỉ số nào đo mức độ phụ thuộc dữ liệu của câu lạc bộ không? Đáp: Có thể tham chiếu Chỉ số Độ sâu Đội hình của VangBong.vn để đánh giá mức độ dựa vào dữ liệu trong vận hành đội bóng.
On July 6, 2026, in a press conference in Kazan, I asked Didier Deschamps a question that made the entire room laugh: would the pressure of an entire nation crush a 19-year-old Kylian Mbappé who was peaking far too early? A male colleague sitting beside me whispered: "Another one of those girly emotional questions." I answered, loud enough for a few rows to hear: "It is emotion that decides performance." Ten days later, France beat Croatia 4–2 in the final, Mbappé scored, and the press room fell silent.
I retell that old story for one very specific reason. Across 35 years of watching this industry, I have always had to fight to prove that the human part of football — intuition, psychology, the ability to read a match with the naked eye — matters as much as the columns of data I carefully attach to every piece. But over the past few seasons, that fight has reversed in a way few anticipated. The question is no longer "does data matter", but "can people still think for themselves when the data is generated by artificial intelligence".
And then I read a study that made me stop in the middle of a video review session.
The study came from the University of California. The authors recruited 1,222 online participants, split across three controlled experiments with sample sizes of 354, 667, and 201 respectively. They had participants solve fraction problems and reading-comprehension tasks — exercises that demand step-by-step reasoning, not guesswork. One group used ChatGPT for roughly 10 minutes as support; the other group worked unaided.
The result is the part that made me sit up straight. When the tool was taken away and participants had to solve new problems without assistance, the group that had used ChatGPT scored only about 57% accuracy, while the control group scored about 73%. It was not only a lower score: many in the AI group began making mistakes, or simply abandoned the tasks, unwilling to keep trying once the tool was gone.
What is notable is that the authors never claim AI is harmful in all cases. They state clearly that the effect is not permanent, and that this does not mean AI should be banned. The problem lies elsewhere: the boundary between using AI as a learning aid and letting AI perform the entire act of reasoning. When AI does all the hard work, the user's brain gets no training, and when it must swim alone, it drowns.
I read this and immediately thought of the analytics departments inside football clubs. Because for a decade, professional football has quietly done exactly what this study warns against — only at a far larger scale, and with far more money.
Recall the context. After Brentford and Matthew Benham's Midtjylland proved that data models could find cheap players the naked eye missed, the whole football world rushed into an algorithmic arms race. StatsBomb, Opta, Hawk-Eye, then a wave of machine-learning startups selling services to clubs. Liverpool under Jürgen Klopp built a formidable data-science research department. Premier League clubs hired physicists and computer scientists. By the 2026–2026 season, almost no club in Europe's top five leagues took the field without a predictive model behind it.
Then came the new generation of tools. Not just statistics, but large language models and generative machine learning: video-analysis assistants, injury-prediction models like Zone7, real-time tactical recommendation systems. Some clubs have tried using AI to draft scouting reports, summarize matches, even propose substitution plans. This is precisely where the University of California study becomes alarming. Football is not merely using AI to calculate. Football is gradually handing AI the act of reasoning.
To understand why this story is more serious than it looks, we must separate three layers within the study and project each onto football reality.
The first layer is the impact on working memory. When participants used ChatGPT, they did not need to memorize the fraction-solving procedure, because the tool handled it. The learning brain is designed to save energy: whatever it does not have to use, it grows lazy about using. In football, the same thing happens to a coach's tactical memory. An assistant coach once had to memorize hundreds of set patterns, analyze opponents through video, and derive regularities himself. When that work is transferred to software, the chance to train tactical memory disappears. A young coach raised on data dashboards may never have had to remember enough to form intuition.
The second layer is the impact on persistence under difficulty. This is the part I find most frightening in the study: when the tool was removed, a significant share of participants abandoned the tasks instead of trying harder. They did not merely weaken; they lost the will to attempt. In football, this quality has a name: character. Every team has moments when it trails, loses a player, gets pinned back. In that moment no algorithm can save you, only a human being standing up. If a generation of players and coaches grows used to having answers ready from a machine, the first reflex when facing an unseen situation will be to wait, not to decide.
The third layer is the change in how questions are asked. Skilled ChatGPT users tend not to ask "what is the answer", but "why is it so". How a tool is used determines whether it is training or replacing thought. In football, this difference comes down to this: a good analytics department offers hypotheses and lets the coach verify them, while a bad analytics department offers conclusions and lets the coach execute them. The latter is more convenient, faster, and quietly strips the decision-maker of the right to think.
These three layers combine into a paradox: the more data football has, the greater the need for people who can read that data, yet the very reliance on data is eliminating that ability.
I have seen this in real life. In 2026, working as a communications consultant for Brazilian women's football after the failure at the Tokyo Olympics, I sat through five meetings with 23 federation members, mostly men. At first, every proposal of mine was answered with tables of numbers: budget, targets, KPIs. But by the third meeting, I asked a question that appeared in no report: "If Canada's women's team won because it has a long-term development program, while we only gathered three weeks before the tournament, then which number is lying to us?" The room went silent. No one had a ready answer on their screen. The final result was a 47% increase in the communications budget for the women's team. That number did not come from an algorithm. It came from a question only a human could conceive.
A woman said one small thing; people laughed. Five years later, they repeat it. The 2026 story was the same. When PSG triggered Neymar's 222-million-euro release clause, I wrote on social media that the club was buying a brand, not discipline, and that this squad would collapse against Real Madrid. People mocked me. I did not retract it; I attached a table of the average touches by PSG's midfield in the previous Champions League season. In March 2026, PSG lost 2–5 to Real Madrid over two legs. What I want to stress here is not that I was right. It is that the judgment came from reading numbers with my mind, not from letting numbers read for me.
That transfer was not merely a purchase; it was a crack in the entire football world. And that crack, six years later, has become a system in which tactical decisions are born in a server room.
What makes the University of California study notable is that it does not stop at complaint. It identifies the mechanism. The problem is not that AI is smart. The problem is that users hand AI the hardest part of reasoning, keep the easiest part, and believe they are working more efficiently. In football, this scenario plays out every day in every opponent-analysis session.
Imagine a Wednesday morning at a training center. The analytics assistant opens the software, types the opponent's name, and receives a report: the opponent likes to build up down the right flank, is weak at defensive set pieces, has a slow defense when counter-attacked. A beautiful report, with charts and heat maps. The coach reads it, nods, and issues a plan. Training proceeds according to that plan. No one asks: how many matches is this report based on? Is the opponent changing? Is the sample large enough to conclude? And most importantly: what if the report is wrong?
I remember a time at the Maracanã, during my 2026 livestream series when the pandemic froze every competition. I invited João — a stadium cleaner of twenty years — to tell his story. The episode reached 2.1 million views and raised more than 340,000 reais for an emergency fund. In 2026, I learned to hear football with my heart when the stadium had no crowd. What I learned was not that emotion replaces data. It was that data never tells the whole story. João knew how humid the Maracanã is in March, which spots drain poorly, which player tends to slip at which corner. No software holds that data, because no one thought to collect it.
There is another aspect few analysts touch: the cost of dependency. When a club pays for an AI system, it is not just buying software. It is buying a habit. That habit has value when the team wins, and becomes a burden when the team loses. For when results decline, the first question the board asks is usually: where did our model go wrong? Few ask: when did we stop thinking for ourselves?
I once watched a young analytics assistant at a mid-table Brazilian club come under heavy criticism after a losing run. He had followed the process correctly: entered the data, exported the report, presented it to the coaching staff. But when asked "what do you think", he could not answer. He only knew what the model thought. That was the moment I realized the problem is not the tool, but that people have delegated so much to the tool that they no longer hold an opinion of their own.
At this point, I must argue against myself, because I know a piece that only worries about AI would be a poor piece.
First, football is not a fraction test. The University of California study measures math and reading comprehension — tasks with a single correct answer and a standard procedure. Football is the opposite: there is no single correct answer, every match is a situation that has never existed before, and the so-called "standard procedure" is usually the first thing broken. Reasoning weakening in a fraction problem does not mean tactical intuition weakens on the pitch. This is the point skeptics of AI in football often miss: they conflate two different kinds of thinking.
Second, the study itself states clearly that the effect is not permanent and does not mean AI should be banned. The fact that some people abandoned tasks when the tool was removed may reflect pre-existing laziness rather than a consequence of using AI. Correlation is not causation. A player who uses an analytics tool to review his own plays does not thereby lose the ability to read the game; he may be learning faster.
Third, and this is what I believe most: the problem is not the tool, but the culture of using the tool. With the same software, one club uses it to ask better questions, another uses it to avoid asking questions at all. The difference is not in the algorithm, but in the person sitting in front of the screen. If a coach still watches video himself, argues with his assistants, forms hypotheses and verifies them, then AI is merely a more powerful tool. If a coach waits for a report to know what to think, no tool can save him — with or without AI.
So I am not calling on clubs to discard artificial intelligence. I am calling on them to clearly distinguish two modes: letting AI support learning, or letting AI perform thinking. The University of California study shows the gap between these two modes is only about 10 minutes — but the consequence of choosing wrongly can last an entire career.
There is one more thing that unsettles me. This study measured adults in an academic setting. But football is not only adults. In academies, children aged 14 and 15 are being taught to read the game with their own eyes — or are being taught to trust a screen. If we hand a fourteen-year-old a tool that answers every tactical question, we are not teaching him football. We are teaching him to look things up. And when he must decide for himself in a match that no tool can keep up with, he will stand still.
I think of Marta and Debinha in the Tokyo 2026 Olympic final, when Brazil drew Canada 0–0 through 120 minutes and lost 2–3 on penalties. On the pitch that night, no algorithm took those penalties for them. Only humans, with trembling legs and a brain that had to calm itself. That is the part of football artificial intelligence will never touch, and also the part we risk letting wither if we lean too hard on the machine.
If my prediction is right, within three to five years, leading clubs will begin hiring people who are not only good with data, but good at questioning data back — a new role, something like an "algorithm skeptic". And clubs that move first in training young coaches' tactical intuition, alongside using AI, will hold an edge in matches the algorithm cannot keep up with.
I may be wrong. But I am ready to sign my name to this prediction, with a column of data to defend it — as I have done for 35 years. Because in a football world that increasingly trusts machines, keeping a brain that can think for itself may be the last competitive advantage no algorithm can buy.


Cầu thủ liên quan
Bài đề xuất
The Hidden Backbone of Turkish Football Sits Behind the Lottery Numbers2026-09-11
Premier League 2026/27 Matchweek 5: Analyzing a Promotional Bulletin and the Numbers That Need Verification2026-09-19
Shubman Gill Hits Fastest ODI Double Century as India Beat West Indies by Eight Wickets in Guwahati2026-10-01
Empty File, Full Story: The Discipline of Verification in the V-League Transfer Window2026-09-13
Bài đề xuất
From the Kart Track to the F1 Throne: Max Verstappen, Red Bull, and the Overlooked Story Behind the 100-Kart Show2026-09-19
Azerbaijan 1-0 Lithuania: The 25th-Minute Goal and a Bottom Tier That Is Not Silent2026-10-05
22 Against 39: Raúl Rosas Jr., Raoni Barcelos and the Problem of Reading a Maturation Curve at UFC Fight Night2026-09-27
Celtic edge Partick Thistle 3-1 in a thrilling encounter: A data-driven tactical analysis2026-09-07
Bayern Munich and the €200M Clause for Olise: An Old Doctrine Bows to a Signature2026-10-09
