This is good. Using the known tidal and Chandler wobble terms, and only training data from 1910 to 1950, here is a validated Mathieu wave equation fit.
Notice the almost excessive overfitting applied to the training interval, yet the validated fit beyond 1950 shows a still very reasonable agreement with the data.
The intriguing aspect to this is that if the Chandler wobble terms were known as of 1950, the El Nino of 1998 could have been predicted 48 years in advance. Apparently the Southern Oscillation was [discovered in the 1920's](https://ams.confex.com/ams/annual2003/techprogram/paper_58909.htm)