Comparing Predictions with Reality, European Championships 2015
Just before the European Championships we published our predictions at http://globaldressageanalytics.com/who-will-qualify-for-rio-in-aachen-this-week
These championships saw some very big surprises, but how big were the differences with the team score predictions really? At every actual event the judges can have a small average difference compared to the predictions of course. To evaluate this we calculate this average difference removing from the sample the big “mistakes”, we define that as differences of more than -3%. (Fluctuations down of 3% or more are usually due to mistakes in the ride, our predictions take into account the “usual” behaviour of each couple by simply tracking their recent scores.) In the Grand Prix in Aachen this fluctuation down was 0.5%, this is very typical particularly at championship events. So to compare our predictions with reality we have subtracted this 0.5% from our predictions, otherwise they are left unchanged. Below we plot the same graph as before, but now including the actual Team Scores achieved
We sure got the medal places wrong! But how exactly?
The German team had three rides well under prediction, so much so that they even dropped below the score we predicted if their top ride had been removed. The British team score was a little lower than predicted and the Dutch team was right at the top of the predicted range of scores. The Swedish team also scored a bit lower than expected (Actually I think they rode safe to lock in that Olympic place) and the Danish significantly lower. The Spanish pulled out all the stops and like Holland they hit their best expectations. France came out right on prediction but all their competitors for the coveted 6th place and Rio qualification spot did a little worse than expected. The overall pattern however of team scores was pretty accurate.
Predictions of scores are based on past behaviour; they take into account the typical strong and weak parts of each test. But they won’t predict the sorts of problems that some of the top riders had this past week.
All 4 German rides were under expectation, as the world has already noted in exhaustive analysis. Fiona Bigwood produced a score towards the top of her predicated range, Carl came in on the money and we all saw that Charlotte and Valegro were not on top form in the Grand Prix. Edward, Hans-Peter and Diederik all came in with scores at the top of their predicted range and that was enough to carry them to Gold.