Makes you wonder if they can not predict the next day how they can predict 50 years.

Indeed.

Actually, predictions are easy, accuracy is hard.

Actually being a weather forecaster is the only job that only has to get things right 10-20% of the time. No one ever expects them to get it right more than that.

They still don't know enough about how weather works. They know a lot more about climate, which is the long term averages, than they do about the short term changes in conditions. They have multiple computer systems that use different data inputs and take different approaches to modeling the fluid dynamics of the atmosphere. The science for each model is reasonably sound, but the different models still disagree in a lot of cases. Each of the models only works with a subset of the possible variables and applies only a subset of the possible equations that can be used to track and predict changes. And each of the models can only be as good as the resolution and accuracy of the inputs that they use. The science simply isn't understood well enough to combine the models into one system that always works, so when the models disagree the forecasters end up using their experience and intuition to try and figure out which one is closer to being right.

In any case, I find the forecasting in my area to be much more accurate now than it was 20 years ago.

In the UK I remember my Maths teacher (so not yesterday then.....) claim that the most accurate Weather forecast was to say that tomorrow's weather would be the same as today's. No idea how accurate that was though.

Interestingly one of the Sunday papers was claiming yesterday that the UK Met Office's weather predictions were a lot more inaccurate than what forecasters in America were saying our (UK) weather would be like. Which begs the question why are we paying the Met Office so much, perhaps we should just tune into an American Weather station. :-)

I heard the same thing about the weather forecast when I was in school. We were told the same thing about the stock market, too. I think it's a fundamental principle of systems with stochastic and/or chaotic behavior, similar to the idea of regression to the mean.