The selection criterion for liver resection (LR) in intermediate-stage (IM) hepatocellular carcinoma (HCC) is still controversial. This study aims to compare LR and transarterial chemoembolization (TACE) in the range of predicted death risk The multivariable Cox regression model (MVR) was estimated to predict mortality at 5 year. The cutoff values were determined by a 2-piece-wise linear regression model, decision curve analysis with MVR model, and hazard ratio curve for treatment plotted against the predicted mortality. 825 IM-hepatocellular carcinoma (IM-HCC) with hepatitis B cirrhosis were included for analysis (TACE, n = 622; LR, n = 203). The 5-year overall survival (OS) rate of LR patients was higher than the TACE group (52.8% vs 20.8%; P < .0001). The line of LR and TACE were crossing with predicted death risk at 100% (P for interaction = .008). The benefit of LR versus TACE decreased progressively as predicted death risk > 0.55 (95%CI: 0.45, 0.62). When predicted death risk over 0.7, decision curve analysis suggested that LR and TACE did not increase net benefit. Patients were then divided into 4 subgroups by the cutoff values (<0.45, 0.45≥/<0.62, 0.62≥/<0.7, ≥0.7). The stratified analysis of treatment in different subgroups, hazard ratios were 0.39 (95%CI: 0.27, 0.56), 0.36 (95%CI: 0.23, 0.56), 0.51 (95%CI: 0.27, 0.98), and 0.46 (95%CI: 0.27, 0.80), respectively. LR reached the maximal relative utility in the interval of 0.45 to 0.62, and both LR and TACE did not increase net benefit at the 5-year death risk over 0.7.Abbreviations: BCLC = Barcelona clinic liver cancer, IM-HCC = intermediate-stage hepatocellular carcinoma, IPTW = inverse probability of treatment weighting, LR = liver resection, OS = overall survival, TACE = transarterial chemoembolization.