First of all, I really enjoy all the blog posts (and the cartoon in this one). From the reliability diagrams, it looks like the CFSv2 is much better at getting the probabilities right on La Nina than on El Nino. Is that the correct interpretation and does it mean that La Nina is more predictable than El Nino? Or is just a sample size issue with more La Nina cases to make the statistics settle down? Has anyone tried looking at a reliability diagram in three dimensions with lead time on the z-axis. It would make for an interesting "reliability surface".