You tagged it. Now what does it tell you?

That line closes a study of 131 Pro League matches, men's and women's, from 2019 (Lord et al. 2023). The authors mapped how the ball moved through every one of them and found that match context had only small effects on how teams moved it. There isn't one winning style. What winners do share is narrower than a style, and that's what this piece is about.

Because here is where most of us sit after part one. We tag. We have circle entries, possession, penalty corners, shots, maybe a GPS file, and a spreadsheet that grows every weekend. The Tuesday evening question is not "what do we have?" It is "which of these is worth my Tuesday evening?"

Territory beats possession. In the one hockey study I found that ranked winners against losers across several stats, circle entries separated the two groups most strongly and possession about half as strongly. Penalty corners convert at roughly one in five at world level, and the four group winners at the 2018 World Cup scored 38% of their group-stage goals from them. Beyond those, the number worth chasing is not the average, it is the anomaly: the figure that does not match what you expected. And on AI, the models that help here read the numbers a human tagged. They do not watch the match.

At the 2018 men's World Cup, across 36 matches, winners came out ahead of losers on all five stats measured: possession, shots, pass accuracy, circle entries and penalty corners (Bagchi et al. 2023). The interesting part is how strongly each one separated the two groups. Circle entries: .663. Possession: .415. Shots: .307. A structure coefficient is simply how strongly one number separates winners from losers, so possession matters at roughly half the weight of getting into the circle.

One catch, and it's the one from part one. This study counts a circle entry as every entry into the circle by the attacking team, a line-crossing count. Secret Analyst's definition from part one throws out the entries that produce nothing, so it would give a smaller number from the same match. I'd guess the ranking holds either way, but don't hold your own entry count up against these figures without knowing which one you're counting.

A 2022 systematic review in football went further. It found match outcome was not related to possession percentage, and that goals came down to shooting quality rather than possession (Wang et al. 2022). Treat it as a nudge rather than a finding for us, because it is football and no hockey study says possession is worthless. The safe line is that territory and entries beat possession, and there is more than one way to be successful.

The older women's data agrees from a different angle. Sunderland and colleagues looked at 130 goals from 70 women's international matches in 2006. Goals arrived faster when the ball was won in the attacking 25, and 68% of all repossessions happened in the attacking half, outside the circle. Win it high, go direct.

The Hudl case study on the Belgian federation describes a four-zone model, with every event tagged by zone and always from the perspective of the team in possession, which turns entries into a picture of where a team actually threatens rather than a single total. Emily Calderon numbers those zones one to four from the team-on-ball perspective: zone one is your own outlet zone, two and three are the middle of the pitch, four is the attacking circle. Her dashboard reads them as a funnel, 25 entries into circle entries by zone into chances or corners, with the method of each entry counted underneath: crash ball, run in, pass in, defender mistake. The Lord et al. study of 131 Pro League matches linked more circle possession and more direct movement to goal from deep with more shot attempts, and concluded that teams should be unpredictable when the opposition are organised.

One professional analyst I asked goes a step further and puts a value on every event rather than ranking a handful of stats. His operation uses its key event data to calculate the goal-scoring probability of all events, ball wins, outlets, circle entries, shots, cards, presses, and for both teams. So the same model that values your attack also prices what each event gives away.

Count entries by zone, not possession by minute. Possession is hard to tag reliably and tells you less. Entries into the 25 and into the circle, left, centre and right, for and against: that is a sheet a parent can fill in from the side of the pitch, and it is the number the World Cup data points at. Write your entry definition at the top of the sheet first (part one has one you can borrow), or two parents will count two different games.

At the 2018 World Cups, the strongest men's teams converted 23.6% of their corners and the strongest women's teams 17.8% (Science and Sport, 2020). Klatt and colleagues, citing Rowe (2019), put the average at the 2018 World Cup at around 20.5%, and note that the four group winners scored 38% of their goals from corners in the group stage. Chris Fry, then analysing for UNC and the USA women, puts it plainly: "Some 40% of goals will come from a penalty corner" (Hudl, 2020).

Opponent-specific corner databases, kept across a season, so the same routine and the same runner can be recognised before the injection. Sebastien Roland reads the picture rather than the strike. "The first thing I check ... can I guess the action they're going to play. And often just by watching the players around the D ... you can guess." He looks at who stands on the castles and who actually blocks (Get the most from video analysis). Emily Calderon labels every corner in the Belgian league with the quarter, the attacking team and the defending team, and keeps the database per team and per part of the season. All six league games are coded by Monday noon, which is what makes that database opponent-specific rather than a pile of clips.

Two counts per match. Corners won and corners conceded, each with its outcome. After ten matches you have your own conversion rate and your opponents'. That's a scouting report that cost you two tally marks a match.