And then we discovered that they were worse than bad: when applied to a really simple record of temperature, they performed worse than a table of random numbers. (cato.org)
As a thought experiment, suppose he used a table of random numbers to classify days into windy and calm ones. (climateaudit.org)
This analysis verified that in fact both models performed no better than a table of random numbers applied to the U.S. Climate Data. (heatisonline.org)