A 15-Game Denominator: The Empty Data Zone in Vietnam's Basketball Season

In July, a coach sent me his team's stat sheet after four games and asked: "Has my defense improved?" It took me forty minutes to answer in three words: Not enough data.

He was not pleased. Four games, roughly 72 possessions each, under 300 possessions in total. At that sample size, a basketball team's defensive rating can swing by nearly 10 points per 100 possessions simply because opponents made or missed a few three-pointers. His team's DefRtg improved by 6 points across those four games. It looked like a transformation. Read closely, it was noise.

Data is a monastery: the less noise, the more clearly you hear something trying to speak. But every monastery has empty rooms. In Vietnamese basketball, the largest empty room sits right in the middle of the regular season.

A season shorter than a school semester

The VBA launched in 2026. Since then, the regular season has always been compressed into roughly three months, tipping off mid-year and finishing before autumn. Each team plays a double round-robin. Regular-season games per team land somewhere between 14 and 18, depending on how many clubs take part. Add the playoffs, and a championship team rarely reaches 25 official games in a full year.

Put them side by side to see the gap: the NBA plays 82 regular-season games, EuroLeague 34. The Philippine Basketball Association runs three conferences a year, partly because its organizers understand that one short conference cannot separate the strong from the lucky. Vietnamese basketball packs everything into one season.

Based on my experience tracking domestic games over 13 years, the problem is not the low number of games. The problem is that we still read this league using the vocabulary of leagues that play 82.

Even the faces that define the league — Nguyễn Huỳnh Phú Vinh, Đinh Thanh Tâm, Justin Young, Chris Dierker — accumulate far fewer possessions per season than an NBA bench player. The denominator is small, and it is small systematically.

How much data does one VBA season produce?

Fifteen games at roughly 72 possessions each means under 1,100 possessions per team across the entire regular season. Numbers do not lie, but they do not tell stories either. At 1,100 possessions, the standard error on net rating sits around 3 to 4 points per 100. A team winning by 5 points per 100 and a team winning by 1 point per 100 can look identical on a stat sheet.

Not every metric needs the same sample size. This is the table I use when advising clubs:

| Metric group | Games needed to stabilize | Readable in one VBA season? | |---|---|---| | Pace, eFG%, turnover rate | 8-10 | Yes | | Team offensive/defensive rating per 100 | 20-25 | No | | Lineup-specific net rating | 40+ | No | | Clutch metrics, close-game net rating | 60+ | No |

This table explains most of the arguing online after every round. Pace and eFG% stabilize early, so observations about style of play — this team runs, that team shoots a lot of threes — have a basis. Judgments like "the defense has improved" or "this team has more character" rest on the group that needs 20 games or more to be trustworthy. A VBA season is not long enough to reach that threshold.

What actually changed?

Last season, a bottom-tier team improved its DefRtg by nearly 8 points over its final five games. The press called it a defensive transformation. I reopened the data and found a different story.

Three of the opponents in that five-game stretch had nothing left to play for and gave their bench players an average of 22 minutes each. Their three-point shooting was 7 percentage points below their own season average. Meanwhile, the team's own three-point rate did not change — meaning its shot profile was never actually adjusted.

That team did not defend better. Its opponents shot worse. People look at the score to remember a game; I look at expected margins to understand the game that never happened.

This is why I always split data into two layers. The process layer covers shot locations, shot types, and genuinely contested defensive possessions — the things a coach controls. The outcome layer covers shooting percentages and points — the things that depend on whether the ball goes in. In a 15-game season, the outcome layer dominates every story that gets told.

Winning by variance

The championship is decided by something unmeasurable. That is what bothers me most.

With 15 regular-season games, the gap between first place and fourth place is usually a handful of points, inside the margin of error. Then the playoffs compress everything into a few games. At that sample size, one player hitting 5 of 7 three-pointers in a single night can change an organization's entire season. Teams call that character. I call it variance, and variance has no plan.

Every coach talks about feel. I do not have feel; I have standard deviation. And the standard deviation of a 15-game season is large enough to swallow almost all the quality difference between teams.

The phrase "load management" also deserves a straight answer. It was imported from the NBA, where the season is five times longer and injury is the number-one variable. In a 15-game season, resting a star for one game throws away nearly 7 percent of the year's sample. Long mid-season breaks, commercial friendlies and national-team windows are what actually break the continuity of the data. When a team loses two weeks to the schedule, every before-and-after comparison becomes meaningless — yet almost nobody notes that in their analysis.

What to watch in the next cycle

I am not proposing more games. The calendar is constrained by arenas, costs and players' academic schedules. What can be done immediately is to extend the time axis instead of the game axis.

A club that keeps three to five seasons of data will hold roughly 3,000 to 5,000 possessions per core player. At that sample size, the question "has this player improved" finally has an answer. At a 15-game sample, that question only has emotion.

Next season, three signals I will track: each team's rate of genuinely contested defensive possessions, the minutes domestic young players log in the final five minutes of games decided by fewer than 6 points, and whether coaching staffs keep their system intact when results turn against them. All three are process metrics; none depends on whether the ball goes in.

In 2026, the whole world mourned Germany. I quietly reopened my model's log file. The lesson that year was not that the model was right. The lesson was that when data says one thing and the majority says the opposite, most people choose the better story.

Vietnamese basketball sits in exactly that position, only at a smaller scale. A short season, a small denominator, and a media market that needs a new story after every round. If next season is still 15 games, noise will again be called character — unless someone is willing to open the data file first.