I had a few minutes spare at work and was playing with Excel data connections when I pondered whether we can discover anything about the races this season from the accuracy of peoples predictions. My thinking process goes like this.
Statistically, if we had enough predictions, we should find that the average prediction is more or less accurate to the actual race result if the race went as expected. If it was a surprise there would be a mean of the results. A further measure of the degree of this unpredictability would be the standard deviation – There will be a greater range of differences in predictions if the race was unpredictable.
For the purposes of the analysis I have removed all of the zero scores and replaced them with Null values. My belief is that they are more often than not people not taking part in that weeks championship – possibly a flaw but feel free to discuss.
Without further ado…
Aus Mal Chi Tur Spa Mon Can Eur Bri Ger Hun
Mean 6.8 5.9 9.1 7.8 10.0 8.0 4.6 10.9 7.8 9.6 9.9
Sdev 3.2 2.8 4.4 4.5 4.9 4.9 1.9 7.2 5.5 4.6 3.9
The most surprising race of the season to date is Canada – with the lowest average score and most people scoring low (as indicated by standard dev). This one caught most people out, which makes sense given the conditions.
The least surprising race of the season was Spain, with Europe a close second. Spain had a narrower SDev than europe meaning that the average prediction was more consistent. Europe, on the over hand, had high scores but a wider spread, indicating that more people were uncertain of the actual result.
Just out of interest I also used the data to predict how well I would be doing if I had taken part in every round of the predictions competition to date vs the average prediction – and it lead me to believe that I should be around 35th as we stand today ;-) Completely farcical use of stats but it’ll spur me on to get my entries in!