The Table Nobody Reads: PPDA, xG and the Final-Twenty-Minutes Problem in V.League

Minute 71, Hang Day Stadium. A visiting central midfielder receives the ball on the edge of his own box, his back to goal. He has four seconds before the pressure arrives. In those four seconds, three home players leave their nominal positions: a striker drops to block the lateral pass, a wide midfielder tucks inside, a full-back pushes high to smother the flank. The ball spills loose. Eleven seconds later, the net ripples.

I replayed that passage seventeen times. Not because it was beautiful — it was raw, even a little chaotic. I replayed it because in my tracking database, that passage does not stand alone. It belongs to a cluster of events sharing the same signature: the home side's PPDA, averaged across the season, was 11.3. In the final twenty minutes alone, that figure fell to 8.6.

PPDA — passes allowed per defensive action. The lower the number, the denser the pressure. A figure of 11.3 marks a side pressing at a reasonably good level for this league. A figure of 8.6 marks a different team altogether. Same squad, same coach, same opponent — the only difference is the clock past the 70th minute.

That is the starting point. And like every starting point in my work, it begins with a table nobody bothers to read. Every prophecy begins with a table nobody bothers to read.

I first wrote that line nearly a decade ago, sitting through twenty-six rounds of a season nobody wanted to rewatch. Now, as the annual campaign rolls through its fixtures, I still sit with the same habit: before I say anything, I open the table. Not to find answers. To find the right question.

And the right question of this season, for me, sits in the final twenty minutes.

---

Context: a season longer than any summary table

Vietnamese football has a paradox when it comes to data. The league has plenty to measure — fourteen clubs, twenty-six rounds, more than 180 matches, every match 90 minutes plus stoppage, every minute hundreds of events logged by positional tracking. But most of what gets logged will sit idle on a hard drive, waiting for someone to decide it is worth reading.

In the first four months of 2026, I sat through all twenty-six rounds of a title-winning side, measuring an average PPDA of 9.8 — the highest in the league at the time. My first analytical piece was dismissed by colleagues as academic, bloodless. I did not change my style. I simply added xG comparison tables and squad-length charts to the next three pieces. By the end of that year, several clubs had begun copying that pressing approach, and the article suddenly circulated widely among players.

The lesson was not that I was right. The lesson was that I was patient enough for the numbers to speak for themselves.

The current season poses a different problem. More matches, shorter rest periods, a calendar compressed by national-team windows and regional tournaments. Five substitutions have become standard. Teams no longer play with eleven players across ninety minutes; they play with fifteen or sixteen different humans across a period fractured into segments.

That means a match is no longer a block. It is a sequence. And each segment of that sequence carries its own data signature.

I started by dividing matches into four twenty-minute blocks, plus stoppage. For each block I measured three things: PPDA, passes into the opponent's final third, and total xG created. Placed side by side, these three metrics give me a picture not of one team, but of four different teams wearing the same shirt.

The result made me sit longer than planned.

V.League does not lack numbers; it lacks people who know how to turn numbers into windows. That is why I never present a single figure alone. A metric standing by itself is a sentence with its second half cut off.

---

Core analysis: four matches inside one match

Start with the first block, minutes 1 to 20.

In my tracking sample this season, the league-wide average PPDA in this block is 12.7. This is the phase where legs are fresh, lineups are unbroken by cards or injuries, and — most importantly — both sides are still probing. The low tempo is not from a lack of ideas, but because nobody wants to open the door first.

By the second block, minutes 21 to 40, average PPDA dips slightly to 12.1. This is when teams have finished reading each other. Long passes begin to appear more often among mid-table sides, while stronger teams shift to short possession in the middle third.

The third block, minutes 41 to 60, is the least volatile: average PPDA of 12.4. Many matches reach half-time level, and the first fifteen minutes of the second half tend to be a tactical lull — teams hold their structure, waiting for the opponent to err.

Then comes the fourth block.

From minute 61 to 90 plus stoppage, league-wide average PPDA falls to 10.9. But that average conceals the most important thing. When I isolate the final twenty minutes — minutes 71 to 90 — average PPDA drops to 10.2. And among the top half of the table, it falls to 9.1.

That is a significant gap. To picture it: the difference in PPDA between the opening block and the final twenty minutes among strong teams is equivalent to an average pressing side suddenly becoming an elite pressing side simply by letting the clock run.

My question is not whether this phenomenon is real. It is real, and it repeats often enough for me to trust it. My question is: what creates it?

Hypothesis one: opponent fitness depletion.

This is the most common explanation, and the easiest to verify. In the final twenty minutes, weaker sides — those rotating less, with thinner squads — begin losing their ability to keep the ball. Shorter passes, more lateral passes, and most importantly, longer ball-handling time per player. When handling time rises, the number of passes a team completes in a fixed window also rises, and that is precisely what PPDA measures.

But this hypothesis has a flaw. If fitness were the sole cause, the phenomenon would appear evenly across all teams. It does not. It clusters among sides with better squad depth — sides able to introduce three or four quality substitutes from the 60th minute onward.

Hypothesis two: five substitutions as a strategic weapon.

This is where I spend most of my time.

Five substitutions were introduced as a fitness measure, but they have become a tactical tool. A coach can completely alter his team's pressing structure at minute 60 without sacrificing quality. He brings on two fast players on the flanks, keeps the midfield intact, and suddenly the team's PPDA drops sharply while the opponent is at peak fatigue.

In my sample, sides using at least four substitutions before the 70th minute average a PPDA in the final twenty minutes 1.8 units lower than sides making only two or three changes in the same window. A gap of 1.8 sounds small. It is not. It is equivalent to upgrading a midfield from average to good simply by managing the clock better.

But here is what I want to stress, and I will say it plainly: five substitutions help deep squads, but they also turn the final twenty minutes into a war of attrition. When both sides have five changes, the advantage no longer belongs to the side that substitutes more. It belongs to the side that substitutes better. And "better" is a concept very few data tables in Vietnam can measure, because it depends on whether a club has a clear plan for the last fifteen minutes.

Hypothesis three: match structure shaped by the scoreline.

This is my favourite hypothesis, because it is the most counter-intuitive.

When a team leads, it usually slows the tempo deliberately. When a team trails, it is forced to push up. In both cases, the match structure changes in a way that increases the number of defensive actions within a given window — and therefore lowers PPDA for both sides.

In other words, a significant share of the "fiercer pressing in the final twenty minutes" phenomenon is not a tactical choice. It is a mathematical consequence of the match losing its balance.

I tested this by splitting matches into three groups: those with a two-goal margin or more at minute 70, those with a one-goal margin, and those level. In the two-goal-plus group, average PPDA in the final twenty minutes was 9.4 — the lowest of the three. In the level group, it was 11.6.

This result does not refute the fitness hypothesis. It adds a layer of explanation. And it reminds me that every metric carries half its truth outside itself.

---

xG: the gap between what is created and what is scored

Now to the second part of the picture: chance quality.

xG — expected goals — is a metric estimating the probability that a shot becomes a goal, based on location, angle, shot type, and the preceding context. It is not perfect. No metric is. But it lets me compare two things the league table never lets me compare: the goals a team scored, and the goals a team should have scored.

In my sample this season, one title-contending side has an average xG differential per match only 0.21 higher than the mid-table group. That is a very small margin relative to their points gap. In other words, that side is collecting more points than the quality of its chances justifies.

This does not mean they will collapse. It means they are living on a loan whose repayment date I do not yet know.

At the other end of the spectrum, one side in the lower half of the table has a better xG differential than two teams above them. This is the pattern I always watch, because it is usually the sign of a team playing better than its results — or of a team missing a finisher of sufficient calibre.

Placed side by side, these two cases give me a question more important than any prediction: if chance quality and results are diverging at both ends of the table, which part of the remaining season will be the correction process?

And this is where I must be most careful with myself. Correlation is not causation. A team with higher xG but fewer points is not necessarily unlucky. Perhaps their attack finishes systematically poorly — something xG cannot capture, because xG measures chances, not the people taking them. Perhaps they concede at pivotal moments that aggregate metrics do not reflect.

I write this line in every analysis of mine, and I will write it again: a player speaks in emotion; ten seasons are needed to form a system. One season of xG diverging from results is an observation. Three seasons diverging in the same direction is a pattern. Five seasons is a structural trait.

---

The transfer market: valuing by context, not reputation

Now to the part I am paid to do.

Every time the domestic transfer window opens, I hear the same kind of question: what is this player worth, did that club get a bargain. The question is not wrong. It is simply half of a question.

A transfer fee means nothing on its own. It only means something beside three things: the player's age, actual minutes played over the last two seasons, and the tactical role the buying club intends to give him.

I once built a simple model for the domestic market: transfer fee divided by minutes played in the top division over the last two seasons, then adjusted for age. The result yields an index I call "price per proven minute". It is not perfect, but it does something absolute price cannot: it exposes deals priced by reputation rather than ability.

In my sample across several recent windows, one pattern repeats. The clubs spending the most money are usually not the clubs buying the players with the best metrics. They are the clubs buying the most famous players. The gap between these two groups is small financially, but very large in value per minute.

This is why I say: the transfer market is not a game of sentiment, it is a game of maps being redrawn. The old map was drawn by name. The new map is drawn by minutes, by role, by age, by the squad structure the player will step into.

And there is a variable very few people account for: stylistic fit. A player with a low personal PPDA — meaning he presses aggressively — will be useless at a side playing a low block. A player with a high volume of progressive passes will be wasted at a side playing long balls. The transfer fee does not reflect this. Only tactical context does.

I once spent four months rewatching every match of a title-winning side, just to answer a single question: does their system work because of the people or because of the structure? The answer was both, but in a ratio nobody wants to hear — roughly sixty per cent structure, forty per cent people. That means if you buy the right players for the right structure, you can replicate much of the effect at far lower cost.

That is the message the Vietnamese transfer market has not yet wanted to read.

---

VAR: controversy does not disappear, it relocates

There is a subject I always approach with maximum caution, because it concerns things xG cannot measure.

VAR arrived with a promise: reduce controversy. After several seasons of observation, I can say that promise has not been fully kept — but not because VAR failed. Rather, because controversy did not disappear. It relocated.

Before VAR, controversy lived on the pitch: the assistant raised his flag, the referee blew his whistle, and the stands reacted within three seconds. After VAR, controversy lives in the review room: an offside line drawn half a boot wide, a frame selected at one-hundredth of a second, a definition of "clear and obvious error" nobody can define.

What interests me, as a data professional, is not whether VAR is right or wrong. What interests me is that VAR has shifted the zone of controversy from an observable space to an interpretable one. And an interpretable space has no clear boundaries.

An example I often use with colleagues: when a goal is disallowed for offside by a few centimetres, the question is no longer "was that player offside". It becomes "which frame was chosen as the moment of contact, and who chose it". That is a question about method, not about law. Yet it is presented as a question about law.

This is why I believe VAR does not reduce controversy. It merely moves it from the pitch into the review room and the grey zone of the law. And when controversy moves into a closed space, it becomes harder to resolve, not easier — because spectators no longer see the decision-making process, only the outcome.

This has a concrete data consequence: the number of cards and stoppages rises, disrupting match tempo. And disrupted tempo is the enemy of every pressing metric. A match fractured by three VAR reviews in the second half will show a higher PPDA than a continuous match, simply because there is less live ball time to accumulate defensive actions.

This is one reason I always check actual ball-in-play minutes before comparing PPDA across matches. Without that step, every comparison is meaningless.

---

Set pieces: the most neglected data zone

If there is one area where V.League sides forfeit the most value, it is dead-ball situations.

In modern football at Europe's top leagues, set pieces account for roughly twenty-five to thirty per cent of all goals. In Vietnam the share fluctuates but is usually comparable, sometimes higher. Yet the training hours devoted to it are disproportionately low.

I measure this with a simple metric: xG created per corner. In my sample, the gap between the best and worst sides on this metric is larger than the gap between them on xG from open play. In other words, V.League teams differ less in organised attacking ability than in their ability to exploit dead balls.

This means: in a league where the gap in squad quality is not large, set pieces are the cheapest tactical edge to buy. They do not require expensive players. They require time on the training ground and a proper designer.

And here I see signs of a shift. A few clubs have begun hiring dedicated set-piece analysts. The number is small. But the direction is clear.

---

Home advantage, travel, and the price of geography

There is a variable that foreign models never get right for Vietnam: geographic distance.

A trip from Da Nang to Hanoi, or from Can Tho up to Pleiku, is not a three-hour bus ride in Europe. It is a day of travel, a hotel night, a familiarisation session, and a body that must adapt to a different humidity.

In my sample, sides with the highest average travel distance per away match have a PPDA in the final twenty minutes 0.7 units higher than sides with the shortest travel. A gap of 0.7 is smaller than the gap caused by rotation, but it is real and it repeats.

This leads to a conclusion I consider important for management: the fixture list is not merely a matter of sporting fairness. It is a performance variable. A side scheduled for three consecutive away matches with long travel will show a lower pressing metric in the second half of the third — and that can be predicted in advance, not after the match ends.

I once published a prediction based on this kind of analysis, with a ninety-five per cent confidence interval and an explicit statement of the model's assumptions. When the prediction failed, I wrote a retrospective piece to trace the data gap. Mistakes, recorded properly, are public learning documents. They only become disasters when hidden.

---

The contrarian angle: when causation runs the other way

At this point I must say what I always say to myself before concluding anything.

This entire analysis rests on a chain of correlational observations. Better pressing teams get better results. Teams substituting more show lower PPDA in the final twenty minutes. Teams with higher xG usually have more points.

But correlation is not causation. And in football, the direction of causation often runs against intuition.

Take the first example. Do good pressing teams win because they press? Or do they win because they have better players, and better players allow them to press better? If the second holds, then a mid-table side trying to copy a strong side's pressing will not get results — because it lacks the human material to run that system.

This is a trap many clubs fall into. They look at a champion, copy the tactical shape, and ignore that the shape was designed for a specific set of players. A shape is not a system. A shape is a blueprint. A system is a blueprint plus builders.

I have seen this at league level. After a season in which a champion won by intense pressing, the next season saw three or four sides attempt a similar style. Most of them finished in the lower half, with low PPDA and high goals conceded. They pressed better, and they lost more.

Why? Because pressing is not an individual action. It is a collective structure. If one player presses while his teammates do not, space opens immediately behind him. If the whole team presses without cover, a single ball over the top wipes out the entire block. Pressing only works when executed as a system, and a system needs time — usually more than one season.

That is why I always reserve a closing paragraph for self-refutation: what would make me change this view?

My answer, this time, is three things. First, if next season's data shows sides copying the pressing style without changing their squads still sustaining results, my "people decide" hypothesis weakens. Second, if actual ball-in-play minutes rise significantly due to VAR reform, every cross-season PPDA comparison must be renormalised. Third, if clubs begin investing seriously in set pieces, the gap between sides on xG from dead balls will narrow, and the cheapest edge on the market will vanish.

I do not know which of these three will happen first. But I know I will track all three, because they are observable, measurable, verifiable signals.

And this is what I want to stress to the reader: a prediction without a falsifying condition is not a prediction. It is a belief dressed up in numbers.

---

On spectators and empty seats

There is a subject I always return to, even though it sits in no data table.

Vietnamese football has full stands and empty stands. The difference between them is usually explained by results. Winning teams draw crowds, losing teams empty out. That explanation is partly right and partly, importantly, wrong.

In my observation, empty stands are usually not a consequence of defeat. They are a consequence of having no story. A losing team with a clear project — a recognisable style, a few emerging young players, a coach with a philosophy — still keeps its audience. A winning team with nothing to tell beyond the scoreline loses its audience the moment the winning run stops.

I write this not as a comment on emotion. I write it as an observation about attendance data. Average attendance per match is a lagging metric. It responds slowly to results and quickly to identity.

Spectators may leave the stands, but the numbers stay seated. And when they come back, they usually come back for a story, not for a league table.

---

Signals for the next cycle

If I had to distil a set of signals to track over the coming phase of the season, I would choose four.

The first is the convergence between xG and results. Sides currently with an xG differential better than their position should be tracked as candidates for an acceleration. Sides with results better than their xG should be tracked as candidates for a correction. This is not a prediction about a specific outcome. It is a prediction about direction, with a confidence interval I cannot narrow below its current width because the sample is small.

The second is how clubs use five substitutions. More specifically, I will measure average substitution timing and the number of changes made before minute 70. If the pattern I observed holds, early-substituting sides will continue to hold a PPDA edge in the final twenty minutes. If the pattern disappears, that signals coaches have adjusted to neutralise the edge — which is itself valuable information.

The third is the frequency and duration of VAR reviews. If actual ball-in-play minutes fall, every PPDA comparison across matches will need adjustment. I will track this as a control variable, not a performance metric.

The fourth is transfer value per proven minute. If the market begins pricing on this index rather than reputation, that is a structural change. If not, it is an arbitrage opportunity for clubs doing good data work.

These four signals do not answer who will win the title. They answer a different question, and for me a more important one: which systems are operating, and which are merely pretending to.

---

Conclusion: what I will do next

I will not end this piece with a prediction about the championship. Football does not work that way, and neither do I.

What I will do next is reopen my database, add another column for the window from minute 75 onward, and see whether the final-twenty-minutes pattern I observed still holds when I narrow the window. If it holds, I will have a stronger signal. If it disappears, I will have a lesson.

Both outcomes are valuable. That is what I love about this work.

We search for the future of football while it already sits in pasts that have not yet been encoded. Every season leaves behind hundreds of matches, thousands of passages, millions of data rows. Most of them will never be read. But a few — read at the right moment, in the right way, by the right person — can change how a club understands itself.

That is my job. Not to say I know what will happen. But to draw the map clearly enough for others to find their own way — with warning markers placed along the road.

And if there is one thing I want the reader to carry away, it is this: V.League does not lack numbers; it lacks people who know how to turn numbers into windows. That window is waiting for someone to open it.

I have opened it. Now the question for the rest of the season is not who will look in. It is who will be the first to look in and change how they work because of what they see.

An empty stadium is not football without singing, it is football without an echo.