Twenty years ago, following a football match meant knowing the score. Maybe possession percentage, if the broadcast bothered to show it. Today a casual fan on their phone can pull up expected goals, pressing intensity, and a win probability curve that updates every thirty seconds.
Something changed. Not the data — that existed all along, locked inside clubs and broadcasters. What changed is who can read it.
The unbundling of sports statistics
The supply side opened first
When Opta and StatsBomb started licensing event-level data, and when public projects like FBref made large chunks of it free, the raw material stopped being scarce. Anyone with a spreadsheet could look at the same numbers an analyst was paid to look at.
But supply alone doesn’t create literacy
Plenty of open data sits unread. The second shift mattered more, and it happened somewhere nobody planned it: in forums.
I’ve spent a while looking at how fan communities actually process numbers, and the pattern is consistent across languages. One person posts a claim. Someone else asks for the source. A third person points out that the sample is 40 matches and the difference isn’t meaningful. Nobody in that exchange is a statistician. Collectively, they’re doing peer review.
Why communities teach better than articles
A well-written explainer about expected goals reaches you once. A forum thread where six people argue about whether xG undervalues counterattacks reaches you repeatedly, from angles the explainer never considered — and you remember the argument.
This is roughly what happened in Korean sports communities, which developed unusually dense analytical cultures partly because the domestic sports calendar is compressed and fans follow multiple foreign leagues simultaneously. On a community like HappyToto, you’ll find match threads where the top-voted reply is frequently a correction. Someone will post a confident read on a baseball matchup, and the response with the most agreement is the one pointing out that the pitcher’s last three starts were against bottom-table teams.
That’s a data literacy engine. It just doesn’t look like one, because it’s wrapped in arguing.
The uncomfortable part
What crowds catch quickly
If you post a wrong number, you get corrected in minutes. Arithmetic errors, mislabeled axes, a stat quoted from the wrong season — the correction arrives fast, because there are simply more eyes than any editorial desk can afford.
What crowds miss entirely
If you post a correct number that implies a misleading conclusion, it often sails through. Selection effects, survivorship bias, the whole family of errors where every individual figure is accurate and the story is still wrong — crowds are surprisingly weak there.
Some communities have tried to address this by writing things down. Rather than re-litigating the same statistical mistakes every season, they build reference material — their community-written guides are essentially institutional memory, an attempt to stop the same argument from restarting every March.
Does it work? Partially. New members still make old mistakes. But the correction now takes one link instead of forty replies.
What this means for anyone building data products
If you’re shipping an analytics feature, the lesson isn’t “add more metrics.” Your users’ statistical literacy was probably built socially, not from documentation. They learned by being wrong in public and getting corrected.
Which suggests something slightly awkward for product teams: the comment section might be doing more educational work than your onboarding flow. Most companies treat it as a moderation cost.
I don’t think most of them have noticed yet.
Frequently asked questions
Do you need a statistics background to follow modern sports analysis?
No, and that’s more or less the argument of this piece. Most people I’ve watched become fluent got there by being wrong in a thread and getting corrected. Slower than a course. Sticks better.
Is expected goals actually reliable?
It’s reliable at describing what happened across a decent sample. It’s much weaker at predicting a single match — which is exactly where most arguments about it take place.
Where does free sports data come from?
Mostly from commercial providers licensing event-level data, with public projects making subsets available. FBref is where most casual analysts start.
Are fan communities better than professional analysts?
At catching arithmetic errors, often yes — more eyes. At avoiding motivated reasoning, no. A crowd that wants a conclusion will find one.
