Smallpox holds a unique position in the history of medicine. It was the first disease for which a vaccine was developed and remains the only human disease eradicated by vaccination. Although there have been claims of smallpox in Egypt, India, and China dating back millennia [1-4], the timescale of emergence of the causative agent, variola virus (VARV), and how it evolved in the context of increasingly widespread immunization, have proven controversial [4-9]. In particular, some molecular-clock-based studies have suggested that key events in VARV evolution only occurred during the last two centuries [4-6] and hence in apparent conflict with anecdotal historical reports, although it is difficult to distinguish smallpox from other pustular rashes by description alone. To address these issues, we captured, sequenced, and reconstructed a draft genome of an ancient strain of VARV, sampled from a Lithuanian child mummy dating between 1643 and 1665 and close to the time of several documented European epidemics [1, 2, 10]. When compared to vaccinia virus, this archival strain contained the same pattern of gene degradation as 20$^{th}$ century VARVs, indicating that such loss of gene function had occurred before ca. 1650. Strikingly, the mummy sequence fell basal to all currently sequenced strains of VARV on phylogenetic trees. Molecular-clock analyses revealed a strong clock-like structure and that the timescale of smallpox evolution is more recent than often supposed, with the diversification of major viral lineages only occurring within the 18$^{th}$ and 19$^{th}$ centuries, concomitant with the development of modern vaccination. ; Other ; M.F.P. and K.H. are funded by the Helsinki University Hospital Research and Education Fund, the Finnish-Norwegian Medical Foundation, the Academy of Finland (grant no. 1257964 ), the Medical Society of Finland, the Sigrid Jusélius Foundation, and the Jane and Aatos Erkko Foundation. D.P.-M. has been supported by the Education Exchange Support Foundation, Ministry of Education and Science, Republic of Lithuania. G.L.S. is a Wellcome Trust Principal Research Fellow. E.C.H. is funded by an NHMRC Australia Fellowship (grant no. GNT1037231). A.T.D., E.C.H., and H.P. are supported by NHMRC grant GNT1065106 . H.P. is supported by a Canada Research Chair, NSERC, SSHRC, CIFAR, and McMaster University. We thank current and former members of the McMaster Ancient DNA Centre, C. Pepperell, and I.H. for their input. We are especially grateful to Agnius Urbanavičius, Justina Kozakaitė, and Daumantas Liekis for their precious support during this research. A.D. and J.-M.R. are both employed at MYcroarray and provided the bait set used here. We thank the Michael G. DeGroote Institute for Infectious Disease Research (IIDR) for generous seed funding for this work.