13 Goals, 8.6 xG and 18 Points: V.League Still Pays for Reading the Table Instead of the Shots

November 2026 closed the V.League season. Long An finished bottom with 18 points. Four months earlier I was sitting in a rented flat in Nha Trang, replaying their matches on a laptop and writing down one figure: 13 goals in 12 rounds against a total xG of 8.6. A gap of 4.4 goals. I published a short blog post arguing that when their luck regressed to the mean, they would be relegated. The comment section called me heartless. They went down.

I am not retelling this for credit. I am retelling it because the method has not changed, and because V.League 2026/2026 still runs on the same logic: the market prices goals and ignores shots.

In 2026 I had no tracking cameras, no Opta feed, no machine learning. I had twelve rounds of video and a spreadsheet. For every shot I recorded four variables: distance to goal, angle, number of defenders between ball and goal, goalkeeper position. From those four I assigned a conversion probability and summed them. That is the whole technique behind xG.

Reading xG is like reading current in open water. You do not stare at the far shore; you watch how the water moves around your shoulder. The goal is the shore. The shot is the current.

Germany 2026 did not collapse on luck. PPDA said it in the group stage. Against South Korea I calculated their PPDA at 13.2, meaning they allowed thirteen passes before applying pressure, and their forwards covered 6.3 km per match. Seventy-four per cent possession with pressing that loose is a team convincing itself with the ball rather than with pressure. The principle transfers to the V.League; only the scale changes.

Long An's 13 goals came from 8.6 xG. Broken down: four from outside the box, three from set pieces, two after the 85th minute against opponents who had pushed up. The outside-the-box group converted at three times their own previous-season rate. That is the signature of an outlier run, not of a skill that appeared in eight weeks.

Goals are the output; shots are the process. The V.League transfer market prices the output and has almost no instrument for pricing the process.

The consequence shows up on the price sheet. A striker with 12 goals from 15 xG gets paid at the top of the market; a striker with 9 goals from 9.4 xG is marked down and may lose his place. The first has spent three seasons of luck in one. The second is holding his production rate and will overtake once his shot volume rises.

Euro 2026 gave me a cleaner sample. Patrik Schick scored five goals from a total xG of 2.6. His strike from 49.7 metres against Scotland on 14 June 2026 carried an xG of 0.03. The press called it a moment of genius. The model called it a low-probability event that happened. A Premier League club rang me before deciding whether to spend 40 million euros. They did not spend it.

The physical layer behaves the same way. In 2026, with stadiums shut, I rebuilt five seasons of V.League data across 240 players, measuring acceleration and distance covered. COVID closed the pitches; I reopened the V-League directory. No league is meaningless. Nguyễn Trọng Hùng was still scoring, but his acceleration had fallen 38 per cent year on year, the signature of a curve bending down rather than form rising. I recommended against a contract extension. Club management replied that I was sitting in Nha Trang talking about a pitch I could not see. He moved to Bình Dương, played 11 matches and lost his starting place.

Here is where I have to stop myself. Correlation is not causation, and regression to the mean is not a law anyone is obliged to obey. Some strikers hold a positive xG overperformance for a whole career because they read positions better than the model's baseline can capture. That skill repeats, and basic xG does not measure it.

My guard is to fix the threshold before I look at the numbers: at least three seasons or 300 shots, a pre-set error band, and a separate test for the shots-inside-the-box group. Setting the threshold after you have seen the result is the most efficient way of fooling yourself that I know.

I have been wrong. In 2026 I marked down a striker because his sample held only 41 shots, under the threshold. He kept scoring for two more seasons. The lesson was not that the model is weak; it was that I used a small sample to make a large claim.

Data never lies, but it knows how to hide. Luck is something I do not have. I have probability and enough data.

The human layer sits outside the model's reach. A player with trouble in the dressing room, a goalkeeper who has just lost a family member, a squad that changed coach midweek, none of that sits in a column. Every analysis I publish ends with a qualitative check: actual fitness, psychology, internal relationships. Drop that section and I am a spreadsheet in a jersey.

People look at the valuation column; I look at the curve. Plenty of transfers die before they are announced.

In the next window I will track three signals in the V.League: shots inside the box per touch for the strikers being quoted at premium prices; the acceleration slope of players over 30 after the 70th minute; and the gap between expected and actual goals for teams in the top half. The team with a large positive gap whose shots still come from outside the box will be the first to pay when the run-in arrives.

Teams do not collapse in one night. They collapse when the indicators stop connecting to each other. The question I keep for myself after every round is simple: if you remove the goals column, what is left of this team?