Liquefaction　potential　index　(LPI)　is　commonly　used　for　liquefaction　risk　assessment.　Although　several　LPI　classifications　have　been　proposed,　those　classifications　are　categorical　and　descriptive,　which　are　not　useful　for　quantitative　risk　assessments　(risk　=　probability　x　consequence).　In　this　study,　a　total　of　135　soil　liquefaction　data　points　from　Taiwan,　Turkey,　and　Italy　were　collected　and　used　for　developing　a　new　empirical　relationship　between　LPI　and　liquefaction　probability　with　logistic　regression.　In　addition,　we　conducted　10　additional　sensitivity　analyses　to　evaluate　the　possible　impact　of　random　data　arrangement　on　this　study,　showing　that　the　impact　was　very　insignificant　given　the　similar　results　obtained　from　the　analyses.　©　Taiwan　Geotechnical　Society.