In their first season, Paju Frontier have less of the ball than anyone in the league and allow their opponents the third-most chances. Yet they have conceded 9.5 fewer goals than their opponents' expected goals — fewer than any other team. They are the exact mirror image of Chungbuk Cheongju from episode 23. I put all 212 league matches on one scale and opened the four phases and the last yard in front of goal one by one.
On 23 August, in R23, Paju were pinned back all afternoon away at Seoul E-Land. Non-penalty xG read 0.28 to 2.58. Seoul E-Land took 22 shots without a penalty, and 10 of them were on target. Goalkeeper Wonwoo Ryu saved 9 of them. Paju scored first on 46′, conceded on 59′, and it finished 1-1.
It was not a one-off. In their first season Paju have less of the ball than any team in the league, and have allowed the 3rd-most expected goals against. Yet the goals they actually conceded are 9.5 fewer — the biggest gap in the league. In the September table Paju are 8W 5D 12L, 12th on 29 points.
Episode 23's Chungbuk Cheongju kept the ball, gave little away, and still conceded 11 more than expected. Paju are the opposite: they give up the ball, allow a lot, and concede less than expected. This piece follows the same order — attack, defensive block, transition, set pieces, then the last yard in front of goal. Once again the answer sits in that last yard. Only this time it is on the side of the saves.
I unrolled all 25 matches minute by minute — when Paju led, when it was level, when they trailed, and the moments the goals went in.
There are few goals in Paju matches: 24 scored and 28 conceded in 25. They failed to score in 11 matches and lost by a single goal 10 times (7 of them 0-1). So the first goal weighs a lot. In the 10 matches where Paju scored first they took 22 points (2.20 per match); in the 11 where they conceded first, only 3 (0.27 per match). They never once drew after conceding first, and came from behind to win just once — 2-1 against Seongnam in R22.
Add up the expected goals of every shot opponents took against Paju and you get 37.5. The goals that actually went in: 28. 9.5 fewer than expected.
That is the biggest gap among the 17 clubs. At the other end sits episode 23's Chungbuk Cheongju (+11.0). Paju allowed the 3rd-most expected goals against, yet conceded the 4th-fewest goals. They gave away 6 penalties (2nd in the league).
But the gap was not spread evenly across the season. Without penalties, the first 14 rounds (13 matches) brought 16 goals conceded from 16.0 expected — almost exactly as expected. The 12 matches from R15 brought 8 from 16.8. Nearly all of the gap came late in the season. One thing changed at that boundary: the goalkeeper. Before that story, the four phases show what kind of team Paju are.
Paju do not keep the ball for long. Their pass share of 41.3% is the lowest in the league, and so is their pass completion of 67.6%. An attack that starts with the ball has 3.1 passes on average — again the fewest in the league.
What they have instead is speed. The share of chains reaching the final third within 10 seconds is 60.3%, first in the league. They enter the final third 36.2 times a match, 6th. Getting forward is not the problem.
What comes next is. The share of final-third entries that end in a shot is 20.6%, the lowest in the league, and box entries that end in a shot rank 15th. The ball gets up the pitch quickly and is cut off just before the shot. So open-play xG is 0.71 per 90, 14th. The goals lean on Borja Bastón — 7 of the team's 24.
The less Paju have the ball, the longer opponents attack. Opponents cross 17.4 times a match against Paju, the most in the league, and get into the box 14.0 times, 2nd in the league. Set-piece xG allowed is 0.21 per 90, 2nd, and they have conceded 6 penalties.
They did at least lower the quality of those crosses: opponents' cross success is 20.5%, the 4th-lowest in the league. But so many come in that the total is large — 37.5 expected goals against over the season, 3rd in the league. On the block alone, Paju are a team that should concede a lot.
In transition the defensive side is good. Paju win the ball back 58.1 times a match, 6th in the league, and their central recoveries rank 3rd. The share won back within 5 seconds of losing it is 5th. The xG they allow within 10 seconds of losing the ball is low too, 0.30 per match.
But the balls they win do not turn into goals. xG within 15 seconds of a recovery is 0.17 per match, 15th in the league, and attacks started by a recovery produced 7 goals all season, 16th. It is the same pattern as the attacking phase — forward quickly, cut off at the end.
Set-piece xG created is 0.11 per 90, 14th in the league; set-piece xG allowed is 0.21, 2nd. They won 4 penalties and conceded 6. Open all four phases and Paju are a team that allows more than it creates. Their non-penalty xG difference is −0.38 per match, 15th in the league, and their expected points of 1.07 per match are those of the 15th-placed team. Their actual points are higher.
The four phases say Paju should concede a lot. They conceded less. What is left is what happens once a shot is heading for goal.
Of the 110 shots on target against Paju, 22 went in. A save rate of 78.2% is first in the league (league average 69.6%). Measured by where those shots went in the six cells of the goal, the placement expectation is 32.7 goals; they conceded 22 — −8.7, again the best in the league.
That figure is the sum of two men. The first 13 matches (R1–R14) had Minseung Kim in goal; the 12 from R15 had Wonwoo Ryu. They are two of the three goalkeepers signed together in the winter. In Kim's matches: 52 shots on target, 16 conceded, placement expectation 15.6 — about as expected (+0.4). In Ryu's: 58 shots on target, 8 conceded, placement expectation 17.1 — 9.1 fewer than expected.
The share of shots on target that become goals makes the difference sharper: 30.8% in Kim's spell, 13.1% in Ryu's. The league average is 27.1%. Where were the saves made? Over the season, 36 shots from the six-yard box brought 7 goals (xG 10.0), and 110 from the central box brought 9 (xG 14.2). They conceded less from close range, from the shots that ought to go in.
Why this cannot be put down to one man · The share of shots blocked by Paju defenders is 20.3%, 5th in the league. The placement expectation knows nothing of shot speed, screening just before the shot, or deflections. Ryu's sample is 12 matches and 58 shots on target. What this data supports is "he saved far more of the shots that went to the same cells than the league average" — and no further.
I split the season in two at the goalkeeper change. The odd thing is that the performances were worse in the second half.
From R15, the non-penalty xG Paju allowed rose from 1.23 to 1.40 per match. The xG they created barely moved, 0.93 to 0.94. On performance, they were pushed back further in the second spell. Yet goals conceded fell from 1.46 to 0.75 per match, and points rose from 1.08 to 1.25. Re-take the same shots 20,000 times and the 12 matches from R15 are worth 12.1 expected points; they took 15.
The attack actually fell away: 9 goals in the 12 matches from R15, 0.75 a match. So Paju in this spell are a team of 0-0s and 1-0s held by the goalkeeper, and one-goal defeats.
The other face of a team whose goalkeeper keeps them in games is their attack. Once Paju fall behind, they rarely come back.
In the 11 matches where they conceded first Paju took 3 points, 0.27 per match, 15th in the league. Draws after conceding first: zero, the only club in the league. They lost 10 of those 11; the exception was R22 against Seongnam, turned around by two second-half goals from Jeho Lee. In the 10 matches where they scored first, by contrast, they took 2.20 points per match (8th in the league). When trailing, Paju's open-play xG is 0.92 per 90, 11th in the league — not enough to chase a game.
So Paju's season is a season of 0-1s and 1-0s: 7 defeats by 0-1, 11 matches without a goal. Only 5 draws, and 4 of them 0-0. While the goalkeeper holds out, one goal wins it; no goal loses it.
I lined up all 25 performances by non-penalty xG, splitting them into edge, even and outplayed at an xG difference of ±0.3.
Paju were outplayed in 15 matches, more than any other category. Those matches went 2W 5D 8L. The 7 matches in which they were outplayed but not beaten (R10 Chungbuk Cheongju 0-0 · R12 Cheonan 0-0 · R18 Suwon Samsung 0-0 · R20 Jeonnam 2-1 · R23 Seoul E-Land 1-1 · R25 Daegu 0-0 · R27 Gimhae 2-0) held up Paju's points total. With the edge: 4W 0D 3L from 7. Even: 2W 0D 1L from 3.
1–4 are the last yard in front of goal, 5–7 the shape of the phases, 8–10 what came of it.
Representative matches are chosen by this series' four rules — ① the match closest to the season average ② the most characteristic (shots per final-third entry) ③ result furthest ahead of process ④ process furthest ahead of result.
A 0-1 at home. Even exchanges, lost to one goal. One of the 7 defeats by 0-1.
Ryu's first start. Opponent xG 3.10 — held to one goal, but Paju could not score.
0.56 expected points, 3 taken. All 4 opponent shots on target saved (placement expectation 0.9).
Non-penalty xG 1.58 : 0.09 in Paju's favour, lost 0-1. The opponent had one shot on target, and it went in.
③ and ④ are the two sides of this piece. When the goalkeeper keeps it out they win even while pushed back; when one goes in they lose even while on top. Most of Paju's results were decided in the last yard in front of goal.
Opponent crosses are the most in the league, box entries 2nd, expected goals against 3rd. The block did not allow little; what it allowed went in less often.
Only half true. Results improved, but xG allowed per match rose from 1.23 to 1.40. What changed was the goal.
They reach the final third quickly and often. But their entry → shot rate is the lowest in the league. Fast progression does not turn into shots.
Partly. But that luck is concentrated in one goalkeeper's spell from R15, and the same man ranks first among league keepers. Kim's spell, by contrast, conceded as expected.
This data cannot say. Having less of the ball is Paju's way of playing, and the price is a defence that allows a lot. Whether possession causes results would need to be tested separately.
Paju give up the ball and go forward fast. They allow a lot, get forward well, and are cut off at the finish. A team of that shape should concede plenty and score little. In fact they conceded 9.5 fewer than expected, most of it in one goalkeeper's spell from R15. In their first season, Paju have held on through the goals they kept out.
The remaining fixtures: R28 Busan (Home, 10 Oct) · R29 Gimpo (Away, 17 Oct) · R30 Hwaseong (Home, 25 Oct) · R31 Chungbuk Cheongju (Away, 1 Nov) · R32 Chungnam Asan (Home, 8 Nov) · R33 Cheonan (Home, 22 Nov) · R34 Yongin (Away, 29 Nov). R31 takes them to episode 23's Chungbuk Cheongju — the team that conceded most against expectation versus the team that conceded least. Three things to watch.
Limits · The portal's xG model is not published. Expected points treat shots as independent. Placement expectation uses only the six goal cells and knows nothing of shot speed, screening or deflections. Goalkeeper figures are approximations by starting keeper. Besides the goalkeeper, the two spells also differ in opponents and squad. The samples are 13 and 12 matches. The figures were found while scanning many measures (multiple comparisons).