Per,
Your work doesn't pass the first sanity check. All you have to do is backcast it to years before 1980. There are 100 years of ENSO data prior to 1980 to validate against. If you can do this and your results show a good correlation, I will be pleasantly surprised.

Otherwise, I believe you may be a victim of a massive over-fitting exercise. Beginners always go through the overfitting stage, but usually get straightened out when they extrapolate to the out-of-band intervals.

The other curious statement in your claims is that you say that you can't get the amplitude right on projection tests, yet your amplitude looks very good on your test interval, 2005-2014. That makes no sense unless you were actually fitting against your test interval. That's no good and it makes it look like you are cooking the books.

My strong recommendation is don't wait for the future when you can learn from the past. I am going all the way back to 1880 and unless you do that too, I remain suspicious of your results.