For fun, I decided to see how well an inverted ENSO graph fitted with itself and find out whether this fit was optimised at any specific lag. One pattern (writing shown correctly) is dragged back in time to find whether the current period is matched by previous years.
If there is a match between a current El Nino and a previous La Nina it will be shown as a match in the upper half of the graph. If there is a match between a current La Nina and a previous El nino, it will be shown by a match in the bottom half of the graph at the lag.
I think there is a match in the lower half at around 1.2 years – and to highlight this I’ve paused the graphic temporarily. Around 70% of strong El Ninos are a match for a corresponding La Nina 1.2 years laters. This lag suggests a 70% chance a La Nina will follow 1.2 years after a (strong) El Nino.
There is also a match in the top half at around 2 years suggesting an El Nino tends to follow about 2years after a (strong) La Nina. Mayby 80% of strong La Ninas are followed by a El Nino 2 years later. Finally I tried to match the current pattern of El Ninos with previous ones. There might be a very weak match at 5.5years.
Given the El Nina peaked around December 2016, this suggests that there’s around a 70% chance of another La Nina peaking around February 2018. Note, this may sound high, but if La Nino and La Nina are equally probable, there’s a 50% (1 in 2) chance of having a La Nina at any point. So, the predictability of 2 in 3 – or that La Nina is twice as likely as El Nino – is only fractionally better than pure chance. The pattern is too chaotic to predict with any confidence over longer periods.