No side in the league allows fewer open-play chances than Incheon. Rate their matches on expected goals and only Seoul are better. Yet they sit seventh, and have conceded more than six goals above expectation. To separate luck from structure, I put all 168 league matches on the same scale and checked one step at a time.
Incheon have 10 wins, 8 draws and 11 defeats from 29 matches — 38 points, seventh in the table of 13 September. Re-rate every match by the expected-goal value of each shot, though, and the story changes.
Across all 168 league matches, Incheon’s expected goal difference per match is +0.39, second only to Seoul. Their actual goal difference is +0.14. Without penalties it is +0.25, 3rd in the league.
Replay every shot at its xG probability and simulate each match, and Incheon’s expected points come to 43.8. They have 38 — 5.8 points short.
It looks like a lot. Set against the rest of the league, it is not.
| Club | Matches | Expected pts | Actual | Diff. | P(actual or lower) |
|---|---|---|---|---|---|
| Gwangju | 28 | 23.5 | 14 | −9.5 | 0.04 |
| Gimcheon | 29 | 36.4 | 30 | −6.4 | 0.15 |
| Pohang | 28 | 42.3 | 36 | −6.3 | 0.16 |
| Incheon | 28 | 43.8 | 38 | −5.8 | 0.19 |
| Daejeon | 28 | 41.1 | 36 | −5.1 | 0.24 |
| Gangwon | 28 | 45.0 | 42 | −3.0 | 0.32 |
| Jeonbuk | 28 | 44.8 | 43 | −1.8 | 0.42 |
| Jeju | 28 | 39.8 | 40 | +0.2 | 0.56 |
| Ulsan | 28 | 41.9 | 44 | +2.1 | 0.67 |
| Anyang | 29 | 32.2 | 35 | +2.8 | 0.72 |
| Bucheon | 28 | 26.9 | 32 | +5.1 | 0.83 |
| Seoul | 26 | 43.2 | 56 | +12.8 | 0.99 |
Conceding more than expected does not make this a weak defence. The opposite.
Opponents rarely get deep into Incheon’s half, rarely produce a good shot when they do, and rarely hurt them straight after Incheon lose the ball. On suppression alone, this is the best defence in the league. So where did the extra goals come from?
Excluding penalties and own goals, Incheon allowed 260 shots. Using every shot in the 168 matches, I worked out how many of them should have gone in if they had been stopped at the league rate.
A shot becomes a goal in three steps — where and what kind of shot was allowed, whether it reached the target (unblocked), and whether a shot on target went in.
In total, +6.9 goals — second only to Gwangju. Fewer shots were blocked than distance would predict (43 against 49.3), and more reached the target (108 against 96.1).
I then controlled for one more thing: whether a defender got to the shooter. Unchallenged shots go in more often against anyone, so the question is whether Incheon’s excess survives once that is in the model.
The chance of an excess this large by luck alone is about 8% (20,000 simulations with the adjusted model, P = 0.083). Too high to call it structure, too low to call it luck. So I looked at which shots the excess sits in.
Split the shots allowed by whether an Incheon defender acted in the three seconds before, and the excess is all on one side.
| Situation | Shots | xG | Expected (league-adj.) | Conceded | Blocked | On target |
|---|---|---|---|---|---|---|
| Within 15 s of the opponent winning the ball | 135 | 13.3 | 12.0 | 18 | 20.0% | 43.0% |
| Settled attack | 99 | 8.4 | 7.4 | 9 | 13.1% | 39.4% |
| Set piece | 26 | 2.7 | 2.7 | 2 | 11.5% | 42.3% |
More than half the shots allowed (51.9%, second in the league) came within 15 seconds of the opponent winning the ball. Volume and quality are both low — the xG Incheon allow after losing the ball is 0.48 per match, the league’s lowest. Yet 18 goals went in from this situation (expected 12.0). The shot is taken before a defence that is still turning can reach the shooter, and it goes in.
| Goalkeeper | Matches | Shots on target | Expected conceded | Conceded | Diff. | Save rate |
|---|---|---|---|---|---|---|
| Kim Dong-heon | 23 | 87 | 20.6 | 25 | +4.4 | 71.3% |
| Lee Tae-hee | 5 | 21 | 3.8 | 4 | +0.2 | 81.0% |
Most of the excess at the “shot on target goes in” step (+4.6) came in the 23 matches Kim Dong-heon played (+4.4). But this data has no shot placement or speed. A hard shot into the corner and a soft one straight at the keeper are both simply “on target”. So I do not read this as a verdict on the goalkeeper — it needs checking on video.
The same breakdown, split by game state.
| Incheon state | Minutes | Open-play xG allowed/90 | League rank (fewest) | Shots allowed | Expected (adj.) | Conceded |
|---|---|---|---|---|---|---|
| Leading | 619 | 0.74 | 1st | 75 | 5.6 | 8 |
| Level | 1627 | 0.70 | 2nd | 141 | 12.7 | 17 |
| Trailing | 621 | 0.59 | 1st | 44 | 3.8 | 4 |
When ahead, Incheon allow less open-play xG than anyone in the league. They do not sit back to protect it — their own open-play xG while leading stays at 0.99 per 90, 6th in the league. Yet in this state they concede 8 against an expected 5.6, the widest overshoot of the three.
Of the 16 matches in which they took the lead, 7 saw it wiped out. Only once did they go back in front.
| Round | Opponent | Final score | Result |
|---|---|---|---|
| R3 | Pohang | 1–1 | D |
| R8 | Bucheon | 2–2 | D |
| R19 | Ulsan | 2–1 | W |
| R21 | Jeju | 3–3 | D |
| R23 | Gimcheon | 1–1 | D |
| R25 | Anyang | 1–2 | L |
| R30 | Daejeon | 1–1 | D |
The other half of that second-best expected goal difference is the attack. But its xG carries the largest penalty share in the league.
Incheon have won 8 penalties and drawn 8 fouls inside the box — both league highs. Penalties are 17% of their xG; 7 of their 36 goals (19%) are penalties.
Incheon reach the final third at the league’s 5th-highest rate (42.3 times per 90). But the share of those attacks that end in a shot is the lowest in the league.
| How final-third chains end | Incheon | League average |
|---|---|---|
| Shot | 18.7% | 21.3% |
| Lost after a completed pass | 17.8% | 17.1% |
| Other failed pass | 18.0% | 16.6% |
| Failed cross | 11.2% | 12.1% |
| Failed pass into the box | 6.2% | 6.6% |
| Failed dribble | 3.4% | 3.1% |
That does not make them poor at creating chances. xG per final-third chain is 0.022, almost the league average (0.023). They shoot less often mainly because they rarely shoot from outside the box (3.8 per match, 11th). It is shot selection, not a shortage of chances. The real shortfalls are elsewhere — attacking straight after winning the ball (shot within 15 s: 4.4%, 10th) and set pieces without penalties (xG 0.070 per 90, 10th).
Mugosa has 12 of Incheon’s 36 goals: 5 from 6 penalties and 7 from open play and set pieces (non-penalty xG 6.3). To see how much the team leans on him, I split their minutes into time with him on the pitch and without.
What Incheon rely on him for is finishing, not creating. Without him a similar volume of chances still comes; what changes is how many go in.
The phase measures by game state in one table. Brackets are league ranks — highest first for attack, fewest first for what is allowed.
| Phase | Leading | Level | Trailing |
|---|---|---|---|
| Incheon minutes | 619 | 1627 | 621 |
| Attack · final-third arrivals/90 | 34.3 (6) | 39.8 (7) | 56.9 (3) |
| Attack · arrival → shot | 20.3% (9) | 18.4% (11) | 17.3% (12) |
| Attack · open-play xG/90 | 0.99 (6) | 0.84 (7) | 1.04 (9) |
| Attack · penalty goals/90 | 0.29 (2) | 0.11 (5) | 0.43 (1) |
| Allowed · open-play xG/90 | 0.74 (1) | 0.70 (2) | 0.59 (1) |
| Allowed · open-play goals/90 | 1.02 (6) | 0.83 (7) | 0.58 (1) |
Leading — they do not retreat, they allow the least, and they concede furthest above expectation. Level — 57% of the time. Upper-middle of the league both ways. Trailing — they press opponents back harder than anyone: the fewest chances allowed, and 3rd for final-third arrivals. But arrival-to-shot is the lowest in the league, and the goals are propped up by penalties (penalty goals per 90, 1st).
Chosen by the numbers, not by impression — ① the largest gap between expected and actual points, ② the most goals conceded per shot allowed, ③ the fewest shots per final-third arrival, ④ the widest xG margin.
Eight shots allowed, three goals conceded. Behind early (9 min), level through a headed corner — then two more conceded, in the 85th minute and in stoppage time. Expected points of 2.02 turned into zero, the largest gap of the season.
Gwangju had 7 shots; without the penalty their xG was 0.59. Those shots produced three goals, penalty included. Incheon shot 17 times, more than twice as often, and lost.
They reached the final third 46 times; 2 of those ended in a shot. Three shots, xG 0.43. And they won 1–0 — a single shot from Ferrier in the 70th minute. It is also a win for a defence that allowed only poor shots.
xG 3.52 to 0.64, 73 final-third arrivals. Ferrier put them ahead in the 40th minute and the equaliser came two minutes later. The winner was a Mugosa penalty in stoppage time. Even this team’s best day ended with a penalty.
Five hypotheses and readings from this analysis did not survive the data. All five were mine. They stay in — a rejected hypothesis marks the edges of the conclusions that did survive.
Corrected. Their rank for the share of unchallenged shots moves with the window (2, 3 or 5 seconds before the shot): 2nd, 1st, 3rd. The gap to the league average is 2–3 points, and the goals that difference cost is +0.5. The core of the excess is not the volume but how often unchallenged shots went in.
Rejected. In the shot-level simulation, the chance of falling this short or further is 0.19. Pohang, Gimcheon and Daejeon are in the same range. The only trace of structure is the shape on the conceding side (unchallenged shots).
Rejected. My first scan read them as “1st for set-piece goals per 90”, but that included 7 penalty goals. Without penalties their set-piece xG is 0.070 per 90, 10th in the league. What is strong is winning penalties (8 fouls drawn in the box, 1st).
Corrected. In the first scan I estimated penalties from the shot’s start coordinates (near the spot) and counted 18 — mixing in open-play shots from the same area. Counted by the event’s penalty tag, it is 8. Event type should not be inferred from coordinates.
Rejected. Non-penalty xG per 90 is 0.98 with him and 0.91 without. Chance creation barely moves; goals do (1.14 vs 0.72).
The diagnosis has narrowed to “they allow little, but of what they allow, unchallenged shots go in more often”. The order below is lowest cost first.
51.9% of the shots allowed came within 15 seconds of the opponent winning the ball (2nd in the league), and 18 of them went in (expected 12.0). The volume is the league’s lowest — the problem is that on those few shots, nobody got there.
On video: once the counter-press is beaten, who is first to turn back, and how far is he from the shooter?
The excess at the “shot on target goes in” step is +4.6 goals, mostly in Kim Dong-heon’s matches. With no placement or speed in this data, the goalkeeper’s share and the shooter’s share cannot be separated.
Split the 29 goals by placement (corner / central), screening and reaction distance, and it becomes clear whether this is goalkeeper coaching or defensive positioning.
18.4% of final-third attacks end in a shot, the league’s lowest, and it is lowest when trailing too. xG per chain is average, so this is not a call for more long shots.
The share of final-third passes played into the box is 27.5%, 11th in the league. That is where to add.
The share of regains that become a shot within 15 seconds is 4.4%, 10th in the league; the xG from them is 9th. The defence wins the ball as well as anyone, and the moment does not turn into attack.
Open-play xG allowed, xG allowed per final-third chain and xG allowed after losing the ball are all the league’s lowest. Dropping the line because of the excess would give up this strength first.
While leading they allow the least xG in the league and keep their own attacking xG at the 6th-highest level. The wiped-out leads did not come from refusing to sit back.
8 fouls drawn in the box, the most in the league. 17% of their xG comes from here. No reason to change it.
Whether the excess reflects the goalkeeper’s ability (no shot placement · needs video) · whether “unchallenged” means an actual absence of pressure (an event proxy — body contact and blocked angles are not logged) · what happened in R27, 0–1 away at Seoul (the portal has no events) · whether this shape will hold for the rest of the season (P = 0.083 is not a verdict).