Weather Prediction

I was looking at my electricity generation chart just now - the one that averages the amount of generation over the last 5 years and highlights the chances of sunshine on a particular day - and noticed something strange; while there are almost no incidences of any particular day having a higher than average chance of sunshine, there are multiple days when the chances are that there will be below average sunshine.

Click on the image to enlarge it and you'll see that the peaks in the blue line, which are high levels of confidence, coincide with below average electricity generation (i.e. cloudy days). So, rather than being able to predict, on average, days of fine weather, the chart is more suited to predicting days of poor weather.

There's a huge anomaly on the right - I think it's a high confidence prediction of a General Election or a 2nd referendum...