Anyang have scored 36 and conceded 46 in 29 matches. The −10 goal difference is almost exactly what their chances say it should be: −9.8 on expected goals. And once the ball reaches the attacking third, they shoot more often than their opponents do. So the gap is not in finishing — it is in how often they arrive. That drop-off sits in one phase of play.
Anyang have 8 wins, 11 draws and 10 defeats from 29 matches — 35 points, ninth. They have scored 36 and conceded 46. A record like that usually gets one of two explanations: bad luck, or poor finishing. The data does not strongly support either.
Expected goals for: 31.9. Expected goals against: 41.7. Expected goal difference −9.8; actual −10. A gap of 0.2 of a goal. They have scored 4.1 more than expected and conceded 4.3 more than expected — two errors of the same size, cancelling out. Finishing has not been unusually good, and the defensive phases have not been unusually bad.
The source of the deficit is equally simple. 29 of the 36 goals and 39 of the 46 concessions came in open play. The set-piece ledger is level: corners 2–2, free kicks 1–1, penalties 4–3. The goal difference was effectively created in open play.
281 shots to 391; 9.7 to 13.5 per match; 155 shots inside the box to 241. Read that far and it looks like a team that cannot create. Change the denominator and a different picture appears.
For every 100 passes played in the final third Anyang take 11.27 shots; their opponents take 10.45. Expected goals per shot also favour Anyang slightly, 0.1135 to 0.1067. While the ball is in the final third, Anyang shoot about 8% more often than the opposition.
Anyang account for 20.9% of their events in the attacking third; their opponents 26.4%. Which narrows the question to one thing — where does the ball stop?
Every validated metric was re-sorted into eight phases by pitch position and possession state. "Opponents" here is not a league average but the 11 clubs Anyang actually played, in their matches against Anyang — two or three games each, so the variance is high and the direction matters more than the absolute value.
Anyang play 5,765 passes in the middle third. 11.0% of them reach the final third; for their opponents it is 12.6%. In the other direction, 11.3% go back into their own third, against 8.5%. Slightly less forward, slightly more backward.
1.6 and 2.8 percentage points. Small, on the face of it. But Anyang play close to 200 middle-third passes a match, so it converts to three fewer forward passes and five or six more backward ones per game — roughly eight opportunities to go forward that turn around instead. And because the ball had already been advanced to the middle third, each of those turnarounds costs more than it looks.
Pass completion in the middle third is also lower, 84.8% to 88.6%. That is closer to an effect than a cause: forward passes are harder and fail more often, so a team that plays more of them should have the lower completion rate. Anyang attempt fewer forward passes and complete fewer passes overall — which makes "there are not many options to play into" a better fit for the 3.8pp than "they collapse under pressure". In their own third (phase 1), completion is 80.4% against 80.0%: identical.
The 1.6pp from phase 2 accumulates along two paths in phase 3. Anyang attempt 18% fewer entries, and those attempts succeed 6.3pp less often. The two multiply.
A second, independent denominator gives the same answer — 11.2 final-third entries per 100 possessions against 15.3 (73%).
The method of entry differs slightly too. Long passes make up 45% of Anyang’s entry attempts and 41% of their opponents’, while the long-pass success rate is 40% for both. Same success rate, higher share, so every long ball chosen lowers the expected return. Their entries are also more central — 20% through the middle lane against 14% — which is the narrowest route in.
Anyang circulate 14.8% of their passes in their own central zone; their opponents 10.6%. In the attacking left and right zones they reach 77% of the opposition share. These two pictures explain every number above in one frame — the ball spends a long time deep and central, and a short time high and wide.
Splitting by position locates it immediately. The share of touches taken in the wide attacking areas is 10.3% for Anyang’s defenders against 10.6% for their opponents’ — level. The gap opens among midfielders, 10.8% against 17.2%. Anyang’s midfielders appear in that zone at 63% of the opposition rate.
Three distinct groups.
Three can be enough. Where a front line has rehearsed patterns, output still comes with few players advancing. Which takes the next question in one direction — what happens above Matheus?
Two hypotheses of my own, both rejected, first. The idea that the forwards do not connect with each other is not true (forward-to-forward links 1.8%, fourth in the sample). And whether their combinations repeat cannot be decided from this data — Anyang have 29 matches against two or three each for the opponents, which contaminates any "unique combination share" comparison with sample size.
The shape of the repeated combinations is clear enough. Kim Dong-jin → Matheus → Kim Dong-jin (9), Ailton → Kim Dong-jin → Ailton (7), Lee Tae-hee → Choi Gun-ju → Lee Tae-hee (6), Matheus → Kang Ji-hoon → Matheus (6). All of them give-and-goes. Wall passes between the full-backs and Matheus, and between the full-backs and the forwards, are this team’s rehearsed pattern. It is not the shape of a side without one.
His 581 final-third possessions, sorted by what he did next and whether a shot followed within 12 seconds.
The two most valuable actions are the completed cross (36.4%) and the forward pass (23.8%), and together they account for 17.6% of the total. Square and backward passes make up 28.7%, incomplete passes 17.4%. Conversion and frequency run in opposite directions.
Split by receiver, each cell holds 10 to 27 events and the confidence intervals are wide. Grouped, it holds up.
And yet Matheus’s most-used outlet is Kim Jung-hyun — 27 passes, 7.4% shot conversion. The forward options are Ailton 11 (36.4%), Choi Gun-ju 13 (30.8%) and Chae Hyun-woo 10 (30.0%). The door that opens most often is the one with the lowest return.
The incoming side is balanced. The 211 passes played into him convert at 19.9% within 12 seconds, against 19.5% for the 221 played out. Going through him is not the cost. The distribution out of him tilts backward.
90 corners won, 147 conceded — they concede 1.63 corners for every one they win. Crosses run at 3.14 per match, 12th in the league. That is not a stylistic choice: crosses per wide pass are 0.24 against 0.27, 89% of the opposition rate, while wide entries themselves are 34% fewer. The territorial gap from phases 2 and 3, reappearing on the flanks.
Squad height is not the issue — it is unremarkable. Mean 181.2cm, fifth in the league; eight players at 185cm or above, five at 188cm or above. What is unusual is that the tall players are goalkeepers, defenders and midfielders.
The forwards who actually play are Choi Gun-ju 175cm (2,040 min), Kim Woon 181cm (1,491 min) and Ailton 179cm (1,313 min), with aerial-duel win rates of 33.3%, 43.9% and 41.8%. The squad as a whole wins 48.0% — it is the front line specifically that sits in the 30s and low 40s.
The players who win most first contacts at Anyang corners are Matheus 177cm (6), Choi Gun-ju 175cm (5) and Lee Chang-yong 180cm (4). Kwon Kyung-won (188cm) and Kim Young-chan (189cm) have two each. The delivery is not finding the tall players. The three set-piece goals point the same way — Ailton 179cm scored two, Kim Young-chan 189cm one. Kim Young-chan is a defender with two career goals, not a designed standing aerial threat.
In January 2026 Anyang signed the 189cm Brazilian striker Elcurano on loan. He replaced the 193cm Mota, 14 goals the previous season, who left for Jeonbuk (where he was folded into their so-called triple-tower front line). The club’s stated reason for the signing was "strength in the air and in duels, finishing inside the box, and holding play". They had read their own squad accurately and recruited to the gap.
The sample is small but the direction is consistent. 15.3% of his receptions come inside the box (second in the squad), 12.9 final-third touches per 90, 21.1% of final-third touches turn into a shot. All three are above Choi Gun-ju (13.5% · 9.4 · 19.8%) and Kim Woon (6.8% · 11.2 · 14.0%).
Less a player who went unused than one who started and was withdrawn relatively early. He started ten matches and came off between 53 and 69 minutes in most of them. Korean reporting credits him with "189cm, 89kg and a good leap, capable of scoring plenty of headers" while noting that "judging the flight of the ball and adapting to the tempo have kept him from matching his predecessor in the air". The reason for his absence from R24 could not be established — the pattern reads like an injury, but no report confirming it was found, and the data shows only the absence.
The shape is not unique to Anyang. Gimcheon’s Jung Jae-min (192cm) was on the pitch for 31% of his own side’s corners. The same pattern appeared independently at two different clubs. The character differs — Jung was used as an impact substitute, Elcurano started and was withdrawn early.
The other end matters too. Choi Gun-ju has played 2,040 minutes at 9.4 final-third touches per 90, the lowest of the regular forwards, and Kim Woon 1,491 minutes at 14.0% shot conversion, the lowest of any. Minutes and these particular rankings do not line up.
The argument so far sits on 29-match aggregates; here it is on individual matches. The four were picked by number, not impression — ① the match closest to the season average across composite indicators, ② the match where the phase 2–3 gap was widest, ③ the match with the season’s highest output once in the final third, ④ the match that moved the season table most. They map onto the report’s four pillars.
Standardising six indicators against the season average, this is the closest match of the 29 (z-distance 0.38). Fewer shots, fewer entries, behind on expected goals — and a one-goal win. Seven of their eight wins came by one goal, and 24 of the 29 matches finished within one goal. This team’s default is not a match it controls; it is a narrow one.
If the report compresses into one match, it is this one. Final-third possession is a third of the opposition’s and entries run at 37%, yet shooting output inside that small territory is 1.75× theirs. They lost the shot count 12–21, but not for want of chance creation — for want of time in the zone. It finished 1–1.
Below-average final-third possession produced a season-high 21 shots — the phase 4 advantage at its most extreme. But 21 shots returned 1.56 xG, below the 2.16 Daejeon took from 16. This is what the 55%-inside-the-box qualifier looks like in practice.
One match moving a season table. Remove it and home goals conceded per match falls from 1.86 to 1.46 (the points total stays at 14). Worth noting: Anyang made more final-third entries, 32 to 19 — a consequence of holding the ball while well behind, and a counter-example to reading entry counts as threat.
That is as far as the data carries the argument. Two things sit outside it. Rather than force a narrative onto them, they are left as they are.
21 points away, 14 at home. Narrowing to the 18 pairs of opponents faced both home and away, away is still +8 points, so this is not an artefact of schedule strength. The defensive numbers point the other way.
Fewer shots faced and a lower expected concession at home, but more actual goals. And removing the 0–7 against Seoul takes home concessions from 1.86 to 1.46 (points unchanged at 14). A 14-match sample containing one seven-goal defeat. On current evidence it is too early to call this a home problem. Flagged for retesting once the season closes.
67% of the goals conceded (31) come after the break. That shape is usually explained by fitness or substitutions — except that 64% of the goals scored (23) arrive in the same window. Less one side collapsing than the game opening up. This data takes it no further.
Six hypotheses raised during this analysis failed against the data. Five of them were mine. They are kept rather than deleted — a rejected hypothesis draws the boundary around the conclusions that survived, and stops the same mistake recurring.
Rejected. Tracing each of the 391 opposition shots back 25 seconds to the last change of possession, 48 (12.3%) began with an Anyang misplaced pass, and only 8 (2%) with one in the central zone. This data does not support a counter-attack narrative. The real gap was sustained-possession shots, 66 to 24, which is what led into the territory argument.
Rejected. Anyang’s 90 corners produced one opposition shot within 30 seconds (1.1%). Corners are close to free for this side. There is no basis for treating avoidance as a rational choice.
Partly rejected. The points reversal survives controlling for opposition, but home is lower on both shots faced and xGA. Removing one 0–7 takes home concessions from 1.86 to 1.46. Closer to "fewer chances conceded, more goals conceded at home" than to a defensive collapse.
Rejected. None of the top 14 final-third combination pairs was forward-to-forward, and I read that as an absence. But top-N lists skew toward players with the most minutes. Recounted at club level, forward-to-forward links are 1.8%, fourth in the sample. Absence from a top-N list is not absence.
Downgraded to undecidable. The basis — "unique/total combinations 0.78, lowest in the league" — was sample-size bias: 29 matches for Anyang against two or three for each opponent. Subsampling Anyang to three matches 2,000 times lifts the median to 0.919 [0.868, 0.975], and a matched within-match comparison gives Anyang 0.914 against 0.879 — less repetitive, if anything — at p=0.508. Whether the patterns exist cannot be decided from this data.
Rejected. This was an item I had proposed as worth a closer look, and it was not. His failure rate is mid-squad, 46% of the losses are crosses and set-piece deliveries, and the 101 produced three opposition shots and no goals.
The gap sits in phases 2 and 3, so the interventions start there. They are ordered by cost — not recruitment or a change of shape, but the positions and choices of players already in the squad.
No central midfielder other than Matheus exceeds a 10% share of touches in the wide attacking areas (Kim Jung-hyun 5.1%, Thomas 8.3%, Knezevic 3.1%). The opposition midfielder average is 17.2%.
It is the most direct lever on the 11.0% forward rate in phase 2, and the cheapest thing in this squad to change. The residual risk behind the ball is far smaller than pushing a centre-back forward.
His most-used outlet today is Kim Jung-hyun (27 passes, 7.4% shot conversion). The three forward options convert at 30–36% but received only 10–13 passes each.
This is workable on the training ground. It changes neither the personnel nor the system — only where he looks first when he receives in the final third.
Long passes make up 45% of Anyang’s entry attempts against 41% for their opponents, while the long-pass success rate is 40% for both. Same success rate, higher share, lower expected return.
Entries are also more central, 20% against 14% — the narrowest lane in.
31% first contact means the delivery is not reaching the tall players. Kwon Kyung-won (188cm) and Kim Young-chan (189cm) have won two first contacts each; the leader is Matheus at 177cm with six.
Since corners themselves are close to free (1.1% shots faced within 30 seconds), committing more bodies is also cheap.
Blaise (16.4% box receptions, 369 min) and Elcurano (15.3%, 860 min) take more final-third touches per 90 than the regulars. Choi Gun-ju is at 9.4 per 90 across 2,040 minutes.
This is a proposed test, not a conclusion. The sample is small and a selection effect has not been ruled out. Giving them minutes is itself the test.
Phase 1 is level with the opposition (pass completion 80.4% vs 80.0%, own-third exit rate 22.6% vs 22.8%, long-ball share 20.7% vs 19.8%). There is no basis for saying it is breaking under pressure.
Phase 4 already favours Anyang (11.27 shots per 100 passes against 10.45; 0.1135 xG per shot against 0.1067). Little to gain from touching an area that is already ahead.
In phase 6, 30 of 2,421 losses led to a shot within 15 seconds (1.2%); in the final third, 8 of 778 within 20 seconds (1.0%). Asking for "safety" would push an already low 11.0% forward rate lower still.
Filling the wide shortfall from the back. Goals conceded are already second-highest in the league, and concessions while leading match expectation (1.75 vs 1.76). There is no case for emptying more space behind. The gap forms in front of the defensive line, not in it.
Whether the front line has rehearsed patterns (undecidable on sample size) · how much Matheus’s absence actually costs (three of the five matches he missed were against the top three) · why the home points return is lower (not explained by the defensive numbers) · why Elcurano was absent from R24 (outside the data, unconfirmed in reporting) · why concessions cluster after the break (goals scored rise with them).