Laparoscopic liver resection for hepatocellular carcinoma (HCC) in Child–Pugh A cirrhosis has been demonstrated as beneficial. However, the role of laparoscopy in Child–Pugh B cirrhosis is undetermined. The aim of this retrospective cohort study was to compare open and laparoscopic resection for HCC with Child–Pugh B cirrhosis. Data on liver resections were gathered from 17 centres. A 1 : 1 propensity score matching was performed according to 17 predefined variables. Of 382 available liver resections, 100 laparoscopic and 100 open resections were matched and analysed. The 90-day postoperative mortality rate was similar in open and laparoscopic groups (4.0 versus 2.0 per cent respectively; P = 0.687). Laparoscopy was associated with lower blood loss (median 110 ml versus 400 ml in the open group; P = 0.004), less morbidity (38.0 versus 51.0 per cent respectively; P = 0.041) and fewer major complications (7.0 versus 21.0 per cent; P = 0.010), and ascites was lower on postoperative days 1, 3 and 5. For laparoscopic resections, patients with portal hypertension developed more complications than those without (26 versus 12 per cent respectively; P = 0.002), and patients with a Child–Pugh B9 score had higher morbidity rates than those with B8 and B7 (7 of 8, 10 of 16 and 21 of 76 respectively; P < 0.001). Median hospital stay was 7.5 (range 2–243) days for laparoscopic liver resection and 18 (3–104) days for the open approach (P = 0.058). The 5-year overall survival rate was 47 per cent for open and 65 per cent for laparoscopic resection (P = 0.142). The 5-year disease-free survival rate was 32 and 37 per cent respectively (P = 0.742). Patients without preoperative portal hypertension and Child–Pugh B7 cirrhosis may benefit most from laparoscopic liver surgery.