How Not to Lose

Gyeongnam FC’s 2026 season went wrong after six minutes. On 1 March, at home to Jeonnam in R1, Kihyun Lee — the 192 cm goalkeeper signed from Hwaseong over the winter — denied an obvious goal-scoring opportunity outside his box and was sent off. Ten men held out for most of the match and lost 1-4.

From the next match Bumsoo Lee kept goal. From R2 to R14 Gyeongnam went 4 wins, 3 draws, 6 defeats, 15 points — lower mid-table. Kihyun Lee did not start again until R15 against Yongin on 6 June.

The 12 matches since are where this piece begins: 4 wins, 7 draws, 1 defeat. The only loss was R21 against Daegu. In the September table Gyeongnam are 9th on 34 points.

One question, then: how did Gyeongnam stop losing? The data answers in layers — the flow inside matches (they fell behind less often), the last cell in front of goal, matches that drift towards a draw and the away 1-1s, and an attack that shoots from distance yet scores more than expected. And at the end it asks again what “not losing” actually means.

Split the season by goalkeeper and the results clearly change. What is strange is what sits underneath.

After R15 the chances Gyeongnam gave up rose rather than fell: shots against per match from 10.1 to 12.1, non-penalty xG against from 0.96 to 1.25. Yet goals conceded fell from 21 in 13 matches to 9 in 12. Replaying the same shots 20,000 times at their xG, R1–R14 brought 1.7 points fewer than expected and R15–R27 3.9 more.

Cumulatively, the two lines separate around R15. Before it Gyeongnam fell short of expectation; after it they kept beating it.

So where did the missing goals against go? Opponents did not shoot less. Their shots went in less.

To test “not losing”, look at the flow inside matches rather than the results. Each of the 25 matches is drawn minute by minute below: the time Gyeongnam led, were level and trailed, and the moments the goals went in.

Up to R14 Gyeongnam conceded first in 7 of 13 matches and took 3 points from them (a comeback win over Suwon FC in R13). They lost the other 6. They spent 34% of their minutes behind.

From R15 it changes. They conceded first in only 4 of 12, and drew 3 of those (R17 Chungnam Asan · R19 Jeonnam · R20 Gimpo) by pulling level. The only defeat was R21 against Daegu, decided by a stoppage-time goal. Time trailing fell to 12%; time level rose to 73%.

Matches in which they scored first look much the same in both spells — 6 matches for 12 points up to R14, 7 for 15 after. Their ability to hold a lead barely changed. The substance of “not losing” is that they fell behind less often, and recovered when they did. The price was that matches mostly drifted along at level.

The portal records which cell of the goal each opponent shot on target went to — two rows (top, bottom) × three columns (left, centre, right). League-wide conversion by cell tells you how many goals balls into those cells produce on average (the placement expectation).

While Kihyun Lee was in goal he faced 45 non-penalty shots on target. Placement expectation: 13.1 goals. Actual: 8. He stopped around 5 more than an average keeper would have. With Bumsoo Lee in goal: 38 on target, placement expectation 11.7, 14 conceded (+2.3).

Rank the league’s 20 keepers (30+ shots on target) by goals minus placement expectation per 10 shots on target, and Kihyun Lee is 2nd. Bumsoo Lee is 16th.

It mattered most away. In the six away matches since R15: 27 shots on target, 4 conceded, placement expectation 7.6. The six home matches: 18 on target, 4 conceded, expectation 5.4.

It was not only the goalkeeper that changed. Between R15 (6 June) and R16 (5 July) came a month-long break, and afterwards Gyeongnam handled the ball differently.

From R15 Gyeongnam kept the ball longer and more safely. Possession rose from 47.1% to 52.8%; balls lost per match fell from 71.6 to 62.7, and in their own third from 17.8 to 13.7. Possessions from their own third reached the final third 36.5% of the time instead of 23.4%, and box entries ended in a shot 59.3% instead of 43.8%.

There was a cost. Keeping the ball longer left more space behind when it was lost: xG allowed within 10 seconds of losing it rose from 0.25 to 0.37 per match, and the share of balls won back within five seconds fell from 24.8% to 20.5%. That is one strand of the rise in xG against. Gyeongnam chose to keep the ball, and the risk left behind was absorbed by the last cell in front of goal.

Now lay the phases over the match flow. The same Gyeongnam behaves like a different team when leading, level and trailing.

Since R15 Gyeongnam have spent 73% of their minutes level. In that time open-play xG per 90 has gone against them, 0.62 : 0.95, and opponents turn final-third entries into shots more often (32% against 29%). Level, Gyeongnam are the team conceding more — and the last cell in front of goal is what has held that time together.

Across the season, xG per final-third chain when level is 0.021, the lowest in the league, and 0.018 when trailing, also the lowest. When leading they hold opponents’ entry → shot rate to 21.0% (the 2nd-lowest). Gyeongnam are not a team that breaks the balance; they are a team that keeps it.

“Not losing” carries a premise — that these were matches they would otherwise have lost. I re-read Gyeongnam’s 10 draws through two layers: the flow of the score and the balance of play.

By the flow of the score, they let wins slip more often than they rescued defeats. Of 10 draws, 6 came after scoring first and being pegged back, 3 after falling behind and levelling, and 1 finished 0-0. Narrow it to the 7 away 1-1s and it is 4 after scoring first, 3 after conceding first. 6 draws after scoring first is the 2nd-most in the league (Chungbuk Cheongju have 7).

In the draws where they scored first, Gyeongnam were outplayed while ahead — open-play xG 0.99 : 1.15 over 156 minutes. With a lead they dropped off, and conceded the equaliser from that deeper position. R12 at Seongnam was scored in the 97th minute and equalised in the 99th; at Chungbuk Cheongju in R23 they led for 23 minutes with 0 shot(s) of their own while the opponent built 0.53 xG, and the equaliser came.

The 3 draws after falling behind (R17 Chungnam Asan · R19 Jeonnam · R20 Gimpo) were all away matches since R15. Over 132 minutes behind, xG was 0.50 : 0.55 — they did not pile on pressure; they stopped conceding and took one chance.

By performance another layer appears. Draws in which they were outplayed by 0.3 non-penalty xG or more were R20 Gimpo (conceded first) · R22 Gimhae (scored first) · R23 Chungbuk Cheongju (scored first) — whatever the order of goals, these were matches they might have lost. There were also draws in which they had the edge (R5 Ansan · R17 Chungnam Asan).

It still stands that their 12 away matches went 1–8–3, with 7 1-1s — the most in the league (next best 3). All six away matches since R15 were drawn; their expected points add up to 5.9, and they took 6. Replay the same shots and those six matches produce 1.7 draws on average; all six being drawn comes up 0.02% of the time. Of those six, three were rescued after going behind, two were leads let go and one was 0-0 — they saved as many matches as they let slip, and the sum is exactly the expected points.

Across the season they have drawn 10 times, 3.3 more than expected — the 4th-largest excess in the league.

The other half of the draws lies in attack. Gyeongnam do not create many chances to win.

Their average shot is worth 0.082 xG, the lowest in the league. Average shot distance is 18.6 m, the 2nd-longest; 47.4% of their non-penalty shots come from outside the box (2nd), and only 4.4% from the 6-yard box (16th). xG per match is 1.02 (15th), and attacks reaching the box number 10.2 per match (16th).

Both routes to chances are blocked. xG from settled attacks is 0.21 per match — the lowest in the league — and xG within 15 seconds of a regain is also the lowest (0.16 per match). They do get forward quickly — 58.0% of final-third entries come within 10 seconds of the chain starting (2nd) — but the quick ball forward does not carry into the box, and the shot comes from where it stops.

Yet they score more than expected: 5.4 goals above their shots’ xG, 2nd in the league. The size of their chances is the smallest; their finishing is near the top.

Where does the surplus come from? Outside the box, where they shoot most, 130 shots produced 5 goals from 4.7 xG — exactly as expected. Long shots did not fly in unusually often. The surplus is inside the box: 12 shots from the 6-yard box for 4 goals (xG 2.5), 56 from the wide parts of the box for 6 (xG 4.1 — the 3rd-largest excess in the league). They scored all 4 penalties. Goals from cross-assisted shots are 3.0 above xG (2nd).

Hyunoh Kim leads the team with 8 goals from 38 shots and 5.2 xG. Sangjun Cho has 4 from 2.3 xG; Danrlei 5 from 3.7 xG (2 penalties).

In short, Gyeongnam’s attack makes little, from far out, and finishes the few chances that do reach the box better than expected. Enough to make a 1-1. Not enough to make a 2-1.

Gyeongnam break up play more than anyone. 13.6 fouls and 2.48 yellow cards per match — both the most in the league. They are also fouled the most, 13.7 per match. Their matches stop a lot.

Three red cards, all three straight reds for denying an obvious goal-scoring opportunity: R1 Kihyun Lee (6th minute at 0-0, lost 1-4), R16 Lucão (59th minute at 0-0, won 2-1 with ten men), R23 Jungwon Choi (61st minute at 1-1, finished 1-1). A defence that will stop the move with a foul at the last moment.

What they lack is the reflex to win the ball straight back. They regain it within five seconds of losing it 22.9% of the time (16th); after losing it in the opponent’s half, 20.4% — the lowest in the league. Rather than pressing to win it back, they drop off and, if needed, foul. So xG allowed within 15 seconds of losing the ball stays around average, 0.42 per match (9th).

Set pieces balance out: 5 scored, 5 conceded.

Gyeongnam lean on foreign players less than any club in K League 2: they account for 10.9% of minutes, the lowest in the league, and the lowest shares of shots and expected goals too.

Instead they used a lot of players — 35 in 25 matches, the most in the league — and their minutes-weighted average age of 26.6 is the 2nd-youngest.

Hyunoh Kim is the focal point of the attack. Per 90, the team scored 1.11 goals in his 1628 minutes on the pitch and 0.60 in the 899 without him; it conceded 0.61 and 1.40. By non-penalty xG, though, the two are similar (0.75 : 0.95 with him, 0.82 : 0.96 without), so most of that gap comes from finishing and saves.

All 25 matches, one row each, by flow (who scored first) and phases (final-third entries, shots after entry, open-play xG, set pieces). Performance is classed by non-penalty xG difference: ±0.3 separates Edge, Even and Outplayed.

They were outplayed in 10 matches, and went 2–3–5 in them. Even matches: 7, 1–5–1. Edge: 8, 5–2–1. Since R15 they were outplayed in 6 matches (R20 Gimpo 1-1 · R21 Daegu 0-1 · R22 Gimhae 1-1 · R23 Chungbuk Cheongju 1-1 · R24 Paju 1-0 · R26 Seongnam 1-0) and went 2–3–1. Compared with 0–0–4 in such matches up to R14, they clearly stopped losing the matches in which they were outplayed — in that narrow sense, “drawing matches they would have lost” is true. All three of those draws were away.

1–5 are how they stopped losing, 6–8 why they do not win, 9–12 the shape of the team.

Representative matches are chosen by the series’ four rules — ① the match closest to the season average ② the most Gyeongnam-like match (shots per entry) ③ the match where the result most outran the process ④ the match where the process most outran the result.

③ and ④ are this piece in miniature. The rules picked two home matches decided by the last cell in front of goal — one with Kihyun Lee in goal, one with Bumsoo Lee.

Gyeongnam found a way not to lose. Fall behind less, and level when they do. Keep the ball more and leave the risk behind to the last cell in front of goal. The result: one defeat in 12 since R15, and matches that drift along at level. But it is not a way to win — of 10 draws, 6 came after they had scored first.

The last seven: R28 Cheonan (away) · R29 Chungnam Asan (home) · R30 Busan (away) · R31 Hwaseong (home) · R32 Suwon FC (away) · R33 Suwon Samsung (away) · R34 Seoul E-Land (home). R29 brings part 20’s Chungnam Asan to Gyeongnam — a team that cannot score when it keeps the ball away from home, against a team that keeps the ball and does not lose. Three things to watch:

• Does the last cell keep holding? — 45 shots on target, 8 conceded: Kihyun Lee’s saving is among the best in the league. Numbers like that tend to drift back towards average over time, and when they do, the higher xG against (1.25 per match) turns into goals.
• The half hour after taking the lead — 6 draws after scoring first, 2nd. Can they go for a second goal rather than dropping off? With the lowest xG per shot and box entries 16th, it will not be easy.
• The cost of stopping play — most fouls and yellows, three red cards for denying goal-scoring chances. One more could decide a match.