Identification of an unrelated HLA allele-matched hematopoietic stem cell (HSC) donor is a costly and time-consuming procedure. To improve search logistics, we have limited the search period to 6 months and have introduced a probability estimate of the chances of identifying a 10/10 HLA allele-matched donor. Probabilities were classified as high (>95%), intermediate (50%) and low (<5% chance) based on allele and haplotype frequencies. By analyzing 350 consecutive searches between 2002 and 2005 (1719 donors tested), the probability estimates turned out to be correct for 96% (high), 88% (low) and 56% (intermediate) patients. For searches with a high probability of success, at least one of the 10 most frequent haplotypes in Caucasoids was found in 69% of the patients, but in only 11% of the patients with a low-probability estimate (P<0.00001). Survival probability at 3 years was significantly higher for HSCT patients classified with a high-probability estimate when compared to patients in the intermediate/low-probability groups (74 vs 51 and 54% respectively, P=0.01). The same difference in survival probabilities was observed when only 10/10 matched unrelated HSCT patients were analyzed. In the intermediate-/low-probability groups, patients with alternative (haploidentical, autologous) or mismatched unrelated donors had similar survival estimates. Probability prediction is therefore feasible in the search process for unrelated donors and can guide the therapeutic strategy.