The risk they never bought,
the chances they never got

00How narrow the season was

What this team’s matches actually look like

Twenty-one of Jeju SK’s 27 matches were decided by one goal or none. Nine draws, twelve one-goal games. Only six were settled by two goals or more.

The nine draws read 0-0 four times, 1-1 four times, 3-3 once. Eight of them finished with two goals or fewer on the board.

None of that is chance. The narrowness is chosen. This piece works out what the choice costs.

01Jeju are the exact inverse of Jeonbuk

Same rounds, same portal, same coordinate frame

Episode 04 covered Jeonbuk Hyundai. The datasets line up exactly, so the two can be laid on top of each other.

02And yet they concede an average number

Why is a low-possession, low-press side 9th for goals against?

Eleventh for passes. A PPDA of 6.5. Eleventh for exposure-normalised defensive volume. How many goals should a side with all three concede? Twenty-seven in 27 matches — ninth in the league. Average. With nine clean sheets.

The raw defensive counts do not explain it. Interceptions 5th, blocks 6th, tackles 7th — nothing that dominates the league. This is not a side that defends by covering ground.

So what holds their goals-against at the league average? The answer is not how much they defend but what they defend.

03They absorb slow build better than anyone

And only the counter gets through

Opponent attacks were split by phase type: fast break (three passes or fewer and 15 seconds or less), slow build (eight passes or more), everything else middle.

Deny them time and they break. So how does an opponent come to have no time?
There is only one way — Jeju have to give the ball away.

04Which is where I got it wrong

One definition, and the whole conclusion moves

Here is what the first draft of this piece concluded, verbatim.

Of the 653 times Jeju lost the ball in their own last third, 33 became an opponent shot within 20 seconds. A 5.1% conversion — five times the rate of any other zone. Both sides make the same mistake; only Jeju are punished for it.

A tidy story. It was also wrong.

The problem was the definition of “losing the ball”. I had counted failed passes plus lost ground duels plus lost aerial duels. Losing a header in your own box as a cross comes in is not losing the ball — the opponent already had it. That is a defensive failure, not a turnover, and counting it as one turns the finding into a tautology: “they lost a duel just before conceding, therefore they conceded.”

Splitting Jeju’s 129 duel losses in their own last third by who held the ball in the previous three events: 118 of them (91%) were never turnovers at all — and five of the seven goals hung on those 118.

05The obvious suspect

And the most clearly innocent

Delete the wrong answer and the question reopens. Where do those 11 counter goals come from?

The first suspect was the goalkeeper. 432 of Kim Dong-jun’s 811 passes (53.3%) travel more than 30 metres. 45.1% of them arrive. And of Jeju’s 653 losses in their own last third, 221 — 34% — came off his boot.

On the numbers alone he looks guilty. But those numbers mean nothing without something to compare them to. Is 45% high or low?

So I built a league benchmark. The Jeju, Jeonbuk and Ulsan datasets each contain both teams’ events; deduplicated by match id they merge into 73 matches, 105,271 events, 4,343 goalkeeper passes.

06So where does the counter begin?

Measure rates, not outcomes, and the place is identifiable

Each of the 11 conceded counters was traced backwards to the last thing Jeju did before possession changed hands.

The goalkeeper originates none of the 11. Nor does any pass over 30 metres. The trigger is the moment they try to knit it forward in their own half.

But where exactly, and why? Eleven goals will not support a verdict on any one zone. So this was measured as a rate rather than an outcome — every short pass Jeju attempted, 5,298 of them (opponents 6,728).

6-1 · The flaw sits in exactly one cell

6-2 · Put that cell on the pitch

6-3 · Which closes the goalkeeper’s story

Chapter 05 noted that Kim Dong-jun plays 379 passes under 30 metres and completes 98.9% of them. Each of those 379 was followed for five events to see what became of it.

This is why he kicks half of them long. The short ball does not come back. Chapter 05 acquitted him because his long passes do not lead to goals. Here is the second half of it — when he does play short, his team cannot hold it.

6-4 · Player or place — a verdict

All of the above is about location. Does the man standing there matter at all?

The split was done by indirect standardisation. The pitch is divided into 18 cells (three thirds × three channels × two directions), and for each player the question is “how many passes should he have lost, if his mix of positions and directions had been played at the team’s average rate?” The reference rate excludes the player himself, so nobody sets his own benchmark. Actual failures divided by that expectation is O/E.

The answer: location first, but the player survives it. Three things are settled.

One — Kim Ryun-seong is cleared. His raw failure rate in his own half was 15.4%, among the highest in the squad. But he takes most of his forward passes from the deep left corner, the hardest square on the pitch for this team. Adjust for it and he sits at O/E 0.89 (0.65–1.21) — exactly where he should be.

Two — Celestin and Tobias are genuinely good at it. 0.38 and 0.52. From the same squares they lose it at less than half the team rate. Celestin’s upper bound is 0.59 across 342 attempts.

Three — Negueba’s 2.94 needs an explanation. 219 attempts in the build-up zone, 16 expected failures, 47 actual. Neither location nor position accounts for it.

In episode 04 I found, late, that a finishing failure I had pinned on players was in fact attached to a phase. This time it runs the other way. Adjust for location and one man walks free, four do not.

07The risk they never bought

Does the avoidance come free?

The same phase split, applied to Jeju’s own attacks.

Going long is free. It originates none of the goals conceded on the counter.
It also produces nothing — 0.061 shots per phase.

Playing short is where the goals conceded begin. Four of eleven.
It is also where the shots come from — 0.240 per phase.

Jeju are a team that declined this trade. They never bought the risk, so they never got the chances.

It meshes exactly with episode 04. Jeonbuk bought the chances by playing through, and watched them harden against a low block. Jeju never went to the counter at all.

The result is nine draws. Four 0-0s, four 1-1s. The most common scoreline was 1-2 — five times. That is the scoreline of a team that loses narrowly. The record is the strategy.

08Attack — Jeonbuk’s mirror image

And what the sample will not allow

Episode 04 concluded that Jeonbuk could not finish the chances they made — 36 shots and no goals from slow build. Jeju are the exact opposite.

One thing the sample does support: Jeju’s attacking problem is not finishing but the sheer number of decisive chances. Of 312 shots, 43.3% are worth less than 0.05 xG (Jeonbuk 31%). They are not short of shots; they are short of places to shoot from.

Negueba alone took 82 of them — 26% of the team’s total — at 0.072 xG per shot against a team average of 0.095. Six goals.

And on the seven days they failed to score, it was not for want of trying.

On the days they did not score, Jeju shot more and delivered more. The opponent sat deeper, shot quality fell to 0.075, and big chances dropped fivefold. It is the same picture episode 04 found — against a low block the attempts rise and the quality collapses.

Set pieces, by contrast, are one of the few places this team manufactures anything. Of 107 corners, 19 produced a shot within 15 seconds and 3 produced a goal. Across the merged 73-match sample the league converts 435 corners into 5 goals — 1.1% against Jeju’s 2.8%. Though it is 3 goals; that is not a number to lean on either.

09A team that plays from in front

Score first and they do not lose. Not once in twelve

One record is attached here without an explanation. At home: 17 matches, 1.29 points per match, six blanks. Away: 10 matches, 1.70 points, one blank. They are worse at home. But 17 against 10 is too lopsided to separate opponent quality from a real effect. It is recorded, not claimed.

10So what kind of team is this?

• Less of the ball (11th for passes), a softer press (PPDA 6.5), fewer defensive actions per exposure (11th).
• And yet they absorb organised attacks better than anyone in the league (0.44×).
• What beats them is the counter — 42% of goals conceded, at 1.88× the expected rate.
• Those counters begin not with a long ball but with a failed short pass in their own half. The goalkeeper originates none.
• One cell carries the flaw — short forward passes in the defensive third (+12.1 points), and inside it, both box corners (56.8% and 51.9%).
• Adjust for location and players remain: Kim Ryun-seong is cleared; Celestin 0.38, Tobias 0.52 and Negueba 2.94 are not.
• They finish the chances they make (66.7% of big chances). But there are 15 of them. No verdict.
• What is certain is the volume — 43% of their shots are worth under 0.05 xG.

Jeju are built not to lose, and the design works.
39 points, fifth, nine clean sheets, unbeaten when they score first.

The same design produced nine draws and seven blanks.
Decline the risk and you decline the chances with it. That is the bill for this season.

11Four matches to watch

All 27 scored on this piece’s own axes — slow-build absorption, build-up rupture, output per phase