The image that did as much as any clinical opinion to fix Lucy Letby in the public mind was not a scan or a laboratory printout. It was a grid. Across a run of collapses and deaths at the Countess of Chester between 2015 and 2016, her row was shaded every time and almost nobody else’s was. Prosecutors treated it as the constant: she was there. Statisticians who later examined the grid have called it something else. Jane Hutton, professor of medical statistics at the University of Warwick, has described the material as a pile of broken crockery that cannot hold water however high it is stacked. John O’Quigley, an honorary professor of statistical science at University College London, has compared the reasoning to the lottery fallacy: the chance of a particular ticket winning is tiny, so the winner must have cheated. Peter Green, emeritus professor at the University of Bristol and a former president of the Royal Statistical Society, has said the chart is a classic misuse, and has noted the unsettling echo of Sally Clark, whose convictions for killing her sons were later quashed after flawed statistics were exposed. The technical objection is selection, not arithmetic.

The events on the chart were not every death or collapse on the unit. They were the events that became the case. Collapses when Letby was not on duty were left off. Once the rows are chosen because she was present, her column fills by construction. The chart then proves only that she was present at the events selected because she was present. Private Eye’s medical columnist Phil Hammond has reported that dozens of other deaths or non-fatal collapses in the relevant period never appeared, including deaths for which she was not on duty and was never charged. A separate modelling exercise by Peter Elston, a fellow of the Royal Statistical Society, suggested that a small shift toward smaller and more premature babies could produce a spike in expected deaths that looks, on a crude plot, like foul play. A Royal College review had already noted that the unit was taking more premature and low-weight infants in those years. Police and prosecutors have answered that the grid was one brick in a wall, not the wall.

They say clinical evidence, notes, insulin results and colleague testimony carried the case, and that a statistician was considered and not relied upon. Hutton’s reply is that doctors with no statistical training were allowed to give statistical opinions, and that the failure to test the roster properly is exactly how medical-murder cases go wrong. The Royal Statistical Society has stopped short of declaring the verdict unsafe. It has said the form of argument used is the kind its own guidance, written after Sally Clark, warns courts against. That distinction matters. A bad chart does not free a defendant. It does mean a picture that looked like certainty to a jury may have been a tautology. The Criminal Cases Review Commission is the body that now has to decide whether that objection, joined to the medical disputes, creates a real possibility that the Court of Appeal would not uphold the convictions. The families of the babies have lived for years with a verdict. The statisticians are asking whether the verdict was partly a picture.