I am starting to lean hard on the Mathieu DiffEq sloshing model to see how it can break.
The top graph (and error curve) is an almost 100 year fit of the sloshing model to the Darwin portion of the SOI.
Then I decided to push it a little more and make it at least 100 years by getting the last several months of data, up to July of this year.
Notice how in adding a few more months, the sudden downward trend was predicted.
I admit that this is perilously close to overfitting, but the indications are that this model is getting perilously close to describing the sloshing of the equatorial Pacific waters.
That downward spike is also an indicator of El Nino, btw. The last big ones were 1982, 1987, and 1998.