Two Weeks In: What FairEleven Has Learned So Far

FairEleven is only two matchweeks old, so this is not the moment for grand conclusions.

But it is enough football to start learning.

That applies to the Eredivisie teams we are following, and it applies to FairEleven itself.

Our aim from the beginning has been to form a view before a match, test that view against what actually happens, and then decide whether the evidence is strong enough to change our understanding.

The first two weeks have shown that this approach can be useful. They have also exposed areas where we are not yet as consistent as we want to be.

The most important lesson so far is simple: the result is only the beginning of the analysis.

PSV 2–2 Fortuna Sittard: a draw that was not an even performance

PSV’s opening league match against Fortuna Sittard was an immediate example.

The game finished 2–2, but the underlying numbers told a very different story. PSV had around 2.65 expected goals to Fortuna’s 0.71, controlled roughly 76% of possession, took 21 shots and produced 16 attempts from inside the box.

The result was level. The balance of performance was not.

That does not mean the correct conclusion was simply that PSV were unlucky and should be upgraded. Football does not work that neatly. Finishing is part of performance, and one match tells us very little about whether poor conversion will persist.

But it does mean a 2–2 draw cannot be treated as evidence that the teams performed equally well.

For FairEleven, this is exactly the distinction we want to preserve. Results matter, but we also want to understand the process that produced them.

And even then, one match is only one observation.

PSV repeated a similar attacking-volume pattern in Matchweek 2 against Excelsior: 20 shots, again 16 from inside the box, with just over 2.0 xG. That makes the attacking volume more interesting than it was after the opening weekend, but still does not make two matches a permanent description of PSV.

The evidence has become stronger. It has not become certainty.

Fortuna 3–1 Cambuur: getting the result right is not enough

Fortuna’s second match gave us a different lesson.

They beat Cambuur 3–1.

It would be easy to take that scoreline, add it to the opening draw at PSV and conclude that Fortuna had produced another convincing performance.

The underlying game was much less comfortable.

Fortuna generated around 1.88 xG, while Cambuur were slightly above 2.0 xG. Cambuur had 64% of possession and created four big chances to Fortuna’s one, missing three of them.

Fortuna won by two goals, but the match did not look like a two-goal superiority underneath.

This mattered to us because part of our pre-match reasoning was supported by the outcome, while part of the explanation was not.

We got part of the call right and part of the explanation wrong.

That distinction matters.

If we only record whether the final prediction looked successful, we learn very little. If we ask why the match produced that result, we can separate a useful piece of reasoning from one that needs to be weakened or discarded.

Fortuna’s opening two matches also illustrate another danger. They have scored five league goals from two matches, but that does not automatically make them an established high-output attacking side. Some of the early finishing has run ahead of the underlying chance quality, and the sample remains tiny.

The right response is neither to ignore the goals nor to overreact to them.

It is to keep watching.

FC Groningen: when competitive evidence overtakes pre-season assumptions

FC Groningen have provided perhaps our clearest example of why a team model has to keep moving.

Before the league started, some of the available evidence pointed towards a front two.

Then Groningen played their first two league matches in a 4-2-3-1, with ten of the eleven starters repeated and the same general structure appearing again.

That is meaningful new evidence.

But even here, the conclusion needs a qualifier. Oskar Zawada was unavailable for both matches, so two observations of the same shape do not yet prove that 4-2-3-1 is Groningen’s permanent preferred system. It may be. It may also partly reflect the personnel available.

The performances themselves were different too.

Against Utrecht, Groningen won 2–1 in a relatively balanced underlying contest, producing around 1.6 xG.

A week later at ADO Den Haag, the attacking picture changed dramatically: Groningen created eight big chances, around 2.21 xG and more than 4.0 xGoT in a 4–1 victory.

That second match is strong evidence that Groningen can create high-quality chances.

It is not yet proof that they will do it every week.

This is the type of distinction FairEleven needs to make better: what have we observed, what does it suggest, and what has actually been established?

Where FairEleven has not been good enough

Groningen also exposed one of our own weaknesses.

Our match analysis had already picked up much of the competitive evidence above: the repeated 4-2-3-1, the personnel context, the underlying numbers and the uncertainty around whether the shape would survive a change in availability.

But the public TeamCard did not reflect all of that quickly enough.

Parts of it still carried pre-season assumptions after competitive football had already provided better evidence.

We have found similar cases elsewhere. A TeamCard might contain the latest score and squad developments while an older hypothesis, tactical description or unanswered question remains unchanged.

That is not good enough.

TeamCards are intended to be living descriptions of Eredivisie teams, not historical documents with recent results added to the top.

If new matches tell us something meaningful about a team’s shape, strengths, vulnerabilities, player roles or unanswered questions, that evidence needs to influence the whole card.

And if the evidence contradicts something we previously believed, that should be visible too.

The reader should not have to know what happens behind the scenes

This is ultimately a question of trust.

A reader should not have to wonder whether one part of FairEleven is more up to date than another.

If an analysis reflects what happened last weekend but the TeamCard beside it still looks like it was written before the season started, the reason does not matter much from the outside. The experience is inconsistent.

That is something we are now working to improve.

The positive side is that some of our deeper match analysis has been using more current evidence than the public TeamCards suggest. We are already combining match results with expected goals, shot profiles, individual contributions, team selection, tactical observations and availability.

But that knowledge also has to reach the parts of FairEleven that readers use to understand a team outside a single match.

We are encouraged by how some of the analysis is developing. We are less satisfied with how consistently that understanding has reached every part of the site.

Both have to improve together.

More data does not always mean more knowledge

The first two weeks have also reinforced another principle: having a metric does not automatically mean we should use it to make a claim.

Pressure data is one example.

We have access to Pressure Index readings from some matches, and they may become useful. But at this stage we are not satisfied that the methodology, scale and cross-match comparability are sufficiently understood to translate those numbers into claims about pressing quality or tactical behaviour.

So we do not.

That may sound cautious, but we think the alternative is worse.

A precise-looking number can create false confidence just as easily as a vague opinion can.

The purpose of adding more data to FairEleven is not to make every conclusion sound more scientific. It is to improve the conclusions that the evidence can genuinely support.

What we are changing

The biggest practical lesson is that updating a team cannot simply mean updating its latest result.

After each match, we need to revisit the existing view.

Did something we considered a strength actually appear again?

Was a vulnerability tested?

Did a tactical hypothesis become more plausible, less plausible or simply remain unresolved?

Did a player establish a new role?

Has an open question already been answered by the football?

And what looked important in one match but disappeared completely in the next?

This does not mean rewriting every TeamCard after every 90 minutes. Two games remain a very small sample.

It means making sure that new competitive evidence is considered rather than merely stored.

What comes next

The immediate priority is to make that process more consistent across FairEleven.

TeamCards need to reflect competitive football more quickly and more completely. Our first match analyses should be available early enough to be useful, then sharpen as new information arrives closer to kick-off rather than appearing only when almost everything is already known.

We are also continuing to develop FairEleven CORE.

CORE is the analytical framework we want to sit underneath more of FairEleven over time.

Its purpose is not to make predictions sound more certain. The ambition is to become better at separating signal from noise, connecting team performance with tactical behaviour and player roles, and recognising when new evidence really deserves to change our view.

That also means recognising when it does not.

CORE is still in development, and there is much we are not ready to say about it yet. But the direction is clear: we want FairEleven to become more consistent at learning across matches without becoming more eager to overreact to them.

Two weeks is enough to start learning

There is nowhere near enough evidence yet to make definitive judgments about the 2026/27 Eredivisie season.

That is part of what makes these opening weeks interesting.

PSV can dominate the underlying numbers and draw. Fortuna can win 3–1 while allowing the better chances. Groningen can produce two victories that look very different beneath the scoreline.

Each match gives us new information.

The difficult part is deciding how much of it should change what we believed before.

That is the job FairEleven is trying to get better at.

And after two weeks, we have also learned that the same standard has to apply to ourselves: keep what the evidence supports, change what it challenges, and be clear about what we still do not know.