By identifying patients at greatest risk for distant metastasis and, hence, most likely to benefit from adjuvant therapy, the grading of sarcomas has been one of the most important contributions pathologists have made to the treatment of sarcomas. Over the years, many grading schemes have been proposed and validated as efficacious. The three-tier system proposed by the French Federation of Cancer Centres is precisely defined, easy to use, and is the most widely employed. However, no system performs perfectly on all sarcomas. Sarcomas that do not lend themselves well to grading include (i) those in which grade provides no incremental information over histological subtypes (e.g. well-differentiated liposarcoma/atypical lipomatous neoplasm, Ewing's sarcoma); (ii) tumours traditionally considered "ungradable" (e.g. epithelioid sarcoma, clear cell sarcoma, angiosarcoma); and (iii) sarcomas that customarily have been graded but in which grade has recently been shown not to correlate well with outcome (e.g. malignant peripheral nerve sheath tumour). Consequently, several sarcoma-specific risk stratification schemes have been proposed. The future may well witness a synthesis of these two approaches. Nomograms, which incorporate clinical, histological and demographic findings, have proved accurate in predicting disease-specific survival in sarcomas. Diagnosis and grading are increasingly based on tissue obtained by core needle biopsy, which poses new challenges for pathologists, particularly if neoadjuvant therapy is to be given. Grading on needle biopsies may require a two-tier grading system (i.e. low versus high grade) and a close dialogue with clinicians to resolve ambiguities.