PPDA and xG in Esports: How Football Data Is Reshaping the Way We Read a Digital Match

In the last three matches, the high-press index of the team considered a title favorite dropped from 8.1 to 6.4. Nobody in the press room mentioned that number. Headlines only circled a few elegant plays and a controversial moment in the final minute. I sat in the seventh row, my notebook filled with small handwriting, and one question kept repeating in my head: are we measuring the right thing?

I came to esports by a roundabout road. In 2026, I began my career as a player and then a tournament organizer, before moving into media. But it was not until 2026, when I left my Master of Movement Science chair and joined the Miami Herald, that I truly understood that a dry spreadsheet, untouched, is just mud. Raw data is mud; to see the truth, you must put your hands in it.

My debut at the Miami Herald was Miami FC versus Indy Eleven at Riccardo Silva Stadium. I meticulously recorded the passing numbers of midfielder Richie Ryan — 87 touches, 74 passes, 91.9% accuracy. I wrote a piece based entirely on the spreadsheet, listing every stat. My editor dismissed it: 'Dry as toilet paper.' I did not argue. I quietly rewatched the entire match tape, then built a framework I called the Territorial Influence Index — combining receiving positions, pass directions, and controlled space. The second piece was born with the very same numbers, and the editor immediately put it on the front page.

That lesson has followed me for nineteen years. A number only has value when tied to a concrete image the reader can picture — from Richie Ryan's turn away from pressure to the space he created with a forty-meter crossfield pass. My rule since then: data illuminates the story, it does not replace it. And when I moved to covering esports for the American market, I carried that rule with me.

Today I want to tell a different story — about how metrics born in football are quietly reshaping the way we read a digital match.

In 2026, at 27, I worked at The Athletic as a data journalist. Ahead of the World Cup, I built a prediction model based on xG differential and PPDA — a metric measuring an opponent's passes before a team performs a defensive action. The lower the PPDA, the higher the press, the more proactively a team closes down to win the ball back. I publicly predicted France would win even though they were rated below Germany and Spain. Russia 2026 is where I staked my honor on the PPDA model and did not regret it.

In the semifinal against Belgium, I pointed out that France's average PPDA was 7.8 — extremely low — meaning they deliberately gave up possession to counter. Belgium had a PPDA of 11.2 but lacked speed at the back. France won 1-0, and my piece was shared more than three thousand times on Twitter. After the tournament, I received an invitation to write my own tactical analysis column.

What I learned was not that PPDA is always right. It was this: a good theoretical framework turns a spreadsheet into a defensible argument. From then on, my writing was no longer a match summary but a viewpoint backed by data. I also learned to explain PPDA and xG to general readers with concrete examples: low PPDA means your team does not mind letting the opponent pass, as long as they push the ball into their own half.

In 2026, when the pandemic emptied stadiums, I was 29, working as a data editor at ESPN. I followed the MLS is Back Tournament in the Orlando quarantine bubble. No crowd, no home advantage — traditional metrics like possession became distorted. I collected GPS data from 37 matches, measuring the running distance of every player. The result: the average player ran 9% less than the previous season, but sprint counts rose 12%. Matches were more explosive, with longer dead-ball periods. I wrote a 4,200-word internal report arguing that the way we measure performance must change in a crowdless context. The report was later edited into a piece on ESPN's front page, sparking a debate about the 'new-style match.'

In the Orlando bubble, the data was silent, but the silence had an echo. I learned that a crisis does not necessarily break the data — it only breaks the way we look at it. Since then, before analyzing any number, I always ask: what is the background condition of the match?

In 2026, when Euro 2026 was held late because of the pandemic, I was 30, handling data for a European football podcast. In the semifinal between Denmark and England, I noticed midfielder Mikkel Damsgaard — a name nobody mentioned in 'players to watch' lists. I calculated his pressing recovery rate: 4.2 ball recoveries in the opponent's final third per match — the highest among players under 23. Against England, Damsgaard made 5 tackles, all successful, creating 3 chances from high pressing. My piece, 'Damsgaard — the modern midfielder the data is missing,' was shared by more than forty European outlets. Afterwards, three Premier League club scouts emailed me for further consultation.

Those three stories — Richie Ryan, Russia 2026, the Orlando bubble — form the methodological framework I bring into esports. I do not bring conclusions. I bring a way of asking questions.

So what do those metrics mean for esports?

I began experimenting in the recent regular season, following an international tournament that gathered the strongest teams. In esports, people usually measure KDA — kills, deaths, assists — or total damage dealt. These are raw metrics, the equivalent of possession in football. They tell you what happened, but not how much it mattered.

I wanted a PPDA-like metric: measuring a team's proactive pressure before the opponent can execute its intent. In a MOBA match, I define a proactive defensive action as any jungle invasion, any ward placed in the opponent's half, any ability trade that forces the opponent back. Counting such actions per opponent possession gives a high-press index.

The result for a team I will call Team A was striking. In the group stage, the index hit 6.2 — among the highest-press teams in the tournament. In the knockout stage it rose to 5.4, then 4.1 in the semifinal. Team A pressed harder the deeper they went. But when I rewatched the tape, I noticed a paradox: the pressure index rose, but the quality of pressure fell. They invaded the jungle more, yet those invasions increasingly failed to create real advantage — no forced tower loss, no major objective secured.

This is the moment to put your hands in the mud. The spreadsheet said one thing; my eyes saw another. I had to go back to the definition.

I split the pressure index into two parts: nominal pressure (number of actions) and effective pressure (actions leading to measurable advantage — lost towers, lost objectives, forced defensive cooldowns). Recalculated, Team A's nominal pressure rose but effective pressure fell from 2.8 to 1.9. They were running more without being more effective. That is the gap between the spreadsheet and the reality of play that I always try to expose.

Alongside that, I built an xG-equivalent for esports — the expected value of a teamfight. In football, xG measures the probability a shot becomes a goal based on position, angle, and pressure. In a MOBA, I measure the probability a teamfight yields major advantage (a key kill, an objective), based on gold differential, formation position, remaining ability resources, and game timing. A teamfight at minute 8 with a 2,000 gold lead has high xG; the same fight at minute 28 against a fully built enemy lineup has far lower xG.

Applying this to Team A, I saw they won through high-xG fights at low frequency — a durable kind of winning, not dependent on luck. Another team, Team B, won through a large volume of low-xG fights — a kind of winning that collapses against disciplined opponents.

Before going deeper, I need to be clear about how I build the metric, because a number without method behind it is just decoration. For the MOBA high-press index, I log every in-game event by timestamp, then classify them into two groups: actions initiated proactively by the team and reactive actions. The final index is the ratio of the first group to the opponent's total possessions.

For the teamfight expected-value metric, I need a large dataset to calibrate probabilities. I take thousands of teamfights from matches across different tournaments, label each: this fight led to major advantage, that one did not. Then I find variables correlated with the outcome: gold differential, number of players involved, map position, game timing, ability resources. The final model produces a probability from 0 to 1 for each teamfight — exactly how xG works in football.

What matters is that I do not use the model to predict match results. I use it to answer a different question: does this team win through control, or through luck? A team that wins by luck will win less over the long run; a team that wins by control will win more. That is the difference between a single win and a winning foundation.

There is one phenomenon I observed this regular season that I want to record. When a team takes the lead, its pressure index usually drops — which makes sense, since it no longer needs to take risks. But there is a group of teams whose pressure index does not drop, or even rises, when they lead. These are teams that understand that in esports, an advantage can vanish after a single teamfight. They keep pressing to deny the opponent any room to breathe. The difference between a champion and a runner-up often lies in their attitude while leading, not while trailing.

I verified this with tape. A team leading by 5,000 gold at minute 20 usually has two choices: push towers and close the game, or retreat to control a major objective. The team choosing the second option often wins more safely, but the team choosing the first — if it has a high effective-pressure index — often closes earlier and reduces the risk of a throw. This is an example of data and observation needing to go together: the spreadsheet tells you who is ahead, but only the human eye sees who knows how to finish.

In first-person shooter titles, the equivalent metric differs. Map control is decisive, and I measure it with a territorial index — the share of time a team holds key areas. A team can win many rounds yet lose on territorial index, meaning its wins rest on individual moments rather than systemic control. That kind of winning does not last. Against a team with better territorial control, it collapses because there is no space left for individual brilliance.

I think back to Richie Ryan. He touched the ball 87 times and passed at 91.9% accuracy, but looking only at the numbers, you miss what he actually did: he adjusted the tempo, stretched the opponent's shape, created space for teammates. In esports, there are players whose KDA says nothing. A support player may finish with 1 kill and 3 deaths, yet it is precisely their movement that opens the win. The spreadsheet calls them ordinary. My eyes call them the architect of the match.

Here I remember Damsgaard. In 2026, I discovered him through a predictive metric the rankings overlooked. In esports, the same thing happens. I followed a young player I will call V — average KDA, unremarkable on the stat sheet. But V's pressing recovery rate as a support reached 4.6 resource recoveries in the opponent's half per match — the highest among players under 20 in the tournament. They did not score many kills, but every movement created space for teammates.

That is the kind of player the formula forgets — someone who creates value through things that never appear on the scoreboard. In football, I used to call it reading the silence of the data. In esports, that silence is even larger, because traditional metrics are far cruder than in football.

And I have not forgotten the Orlando lesson. This regular season unfolds against constantly shifting patches. Each patch is a new background condition. A team strong on the previous patch can collapse on the next, not because it got weaker, but because the playing environment changed. I always note which version was used in each match, because comparing metrics across versions is comparing apples to oranges. A patch does not just change character strength — it changes the definition of playing well.

One example made me think a lot. At a tournament, the highest-rated team in the group stage lost in the semifinal. Commentators blamed individual form. But when I reviewed the data, their effective-pressure index had declined steadily across three matches, in parallel with opponents learning to break their press structure. That team did not get weaker — the opponents got better. This is what raw metrics never show: sometimes you lose not because you played worse, but because others played smarter.

But this is where I must doubt myself.

Importing football models into esports has a major trap: correlation is not causation. A team with a high pressure index usually wins, but that does not mean high pressure causes the win. Both may be consequences of a third factor — superior individual skill, or a weaker opponent. If I advise a team to press more only because the champion had a high pressure index, I am committing the most basic logical error of a data analyst.

Second, the football model assumes a continuous space, where the ball moves in straight lines and probabilities shift smoothly. Esports is a discrete space: a teamfight can reverse the game in three seconds, a single kill on a key player can erase twenty minutes of accumulation. My model must relearn how to describe those discrete shocks. This is why I never publish a prediction based purely on an imported model.

Third — and this is where I was wrong and must publicly admit it — I once underestimated psychology and tournament culture. At one tournament, I predicted a team would win based on a superior index. They lost. The cause was not in the data but in an internal roster crisis no spreadsheet recorded. My broken assumption was: complete data means correct prediction. The truth is: complete data is only correct when the human context is stable. I revised the model, added a qualitative variable for roster stability, and reduced prediction weight in tournaments with heavy transfer volatility.

This leads to a larger paradox. The esports analytics industry is entering the stage football went through twenty years ago: data becomes a commodity. Organizations trade models, hire data experts, build analytics departments. But when data becomes a commodity, it also becomes a marketing tool. The biggest risk is not wrong data — it is right data used to tell a wrong story. A team can pick its prettiest metric to persuade sponsors while hiding the ugly numbers. The reader must learn to ask: was this metric chosen because it is true, or because it is beautiful?

And here is what I want to stress about the counterintuitive angle. For years, the esports industry chased 'superstars' — high KDA, beautiful highlights, big followings. The transfer market pushed young talent to absurd prices, just like the young-player price bubble in football: one hundred million euros for a player with fewer than fifty top-flight matches is a naked gamble. In esports, the same is happening. Teams pay high for names inflated by media, while players who create value through invisible things — pressing recovery, territorial control, tempo adjustment — are undervalued. The bubble is not in the true value of talent, but in the gap between true value and the inflated name.

So what are the signals for the next round?

I will watch three things. One, the effective-pressure index — not the number of pressing actions, but the number leading to measurable advantage — of teams advancing deep in the knockout stage. Two, the pressing recovery rate of young players under 20, because that is where overlooked gems lie. Three, roster stability before each major patch, because the playing environment is changing faster than many teams can adapt.

Esports is at exactly the stage football passed through twenty years ago: the data exists, but the people who can read it are still scarce. Whoever learns to put their hands in the mud first will see the match before it ends.