Background: The outcome of the patients with BCLC stage B hepatocellular carcinoma (HCC) varied. Albumin-Bilirubin (ALBI) grade, as a surrogate of the Child-Pugh (CP) grade, was evaluated to be a simple tool for the assessing of the liver function and prognosis. However, it appears to be arbitrary and crude to eliminate the ascites variable from the ALBI grade. We aimed to develop a predictive model constituted with the ALBI grade, the ascites and tumor burden related parameters in patients with BCLC stage B HCC. Methods: Patients diagnosed as the BCLC stage B HCC were collected from a retrospective database. Construction and validation of the predictive model were performed based on multivariate Cox regression analysis. Predictive accuracy, discrimination and fitness performance of the model were compared with the other eight models. The decision curve analysis (DCA) was used to evaluate the clinical utility.Results: A total of 1773 patients diagnosed as BCLC stage B HCC between 2007 to 2016 were included in the present study. As the methods for the assessing of the liver function, the ALBI grade and ascites showed their independent prognostic value, and then the two parameters were combined into one, the ALBI-AS grade. Subsequently, the ALBI-AS grade (hazard ratio (HR)=1.26, p=0.008) along with two tumor burden related parameters (the AFP level and the 8-and-14 grade, p<0.001) were used for the development of a prognostic prediction model after multivariate analysis. The area under the receiver operator characteristic curve (AUROC) for overall survival at 1, 2 and 3 years predicted by the present model were 0.73, 0.69 and 0.67 in the training cohort. The concordance index (c-index) and the Aiken information criterion (AIC) were 0.68 and 6216.3, respectively. In the internal and external validation cohorts, the present model still revealed excellent predictive accuracy, discrimination and fitness performance. Then the ALBI-AS based model was evaluated to be superior to other prognostic models with highest AUROC, c-index and lowest AIC values. Moreover, DCA also demonstrated that the present model was clinical beneficial. Additionally, participants could be classified into three distinct risk groups by the model.Conclusion: The ALBI-AS grade, as a pragmatic alternative of the ALBI grade, is a novel predictor of survival for patients with BCLC stage B HCC. The ALBI-AS based model was evaluated to be an accurate prognostic tool for individual prognostication, and performed well in terms of discrimination and fitness against other prognostic models. And it is appropriate to validate our findings on a larger prospective cohort.