Being able to forecast is a big thing for scientists, but not always having understanding of the system leads to better forecasting. In physics we have good understanding of celestial dynamics, thus we can make accurate predictions about the movement of planets. Though we also have a decent understanding of how the weather works, weather forecasts are not that reliable. And most likely weather forecasts won't get much better over the time, because we will never be able to precisely predict large number of minor influences, which get amplified and then the system deviates from the forecast (the so-called butterfly effect).
Situation with forecasts in social sciences is even worse, but we can do more to improve them. In this SciShow video Hank Green tells us about some of the interesting intricacies.
In this video I especially liked one aspect - open-minded non-experts outperformed close-minded experts. The non-experts did a better job, because they more often incorporated evidence contradicting their prior beliefs, while the experts more often rejected such evidence.
So, the takeaway lesson? I guess I should not trust an "expert" who shows off his firm beliefs. Nor should I myself be an "expert" who values his beliefs more than the evidence.