Just looking back at my predictions, I thought I would add a few more numbers. Again, just for fun, and I’m not sure that this either confirms or denies the accuracy of the model. Probably with more data (i.e. as years go by if I can keep it up), it will become more accurate. I will definitely switch from scoring place to actual place as that’s more significant, but I may also move to the cycling measure of time behind leader, as this is probably more useful than placing, and it also standardizes across courses (there would be some differences, but I think it’s more accurate). Also, it would help, time-wise, if courses were accurately measured. I don’t mean that they all have to be 4k or 8k or 10k exactly, I just mean, if it is 5.5k, say that it is 5.5k. It’s cross country, the distance doesn’t matter.
Here are the top ten women and men in terms of improved placement, based on predicted finish place vs actual finish place. Note that these are scoring places not actual places.
The average difference was -0.189 places for 36 predicted athletes who ran both races. 8 people finished exactly as predicted, and 11 were either one better or one worse than predicted. So I was more than half right?
HOWES Concordia -7 (predicted 34, actual 27)
VEZINA UQAM -7 (predicted 45, actual 38)
LAHAIE UQTR -6 (predicted 32, actual 26)
MARANDA Sherbrooke -5 (predicted 25, actual 20)
ROY Concordia -4 (predected 7, actual 3)
PICHETTE UQTR -4 (predicted 44, actual 40)
HEWITT Concordia -3 (predicted 36, actual 33)
PATTOU Laval -2 (predicted 17, actual 15)
CARON UQTR -2 (predicted 37, actual 35)
CASTONGUAY UQTR -2 (predicted 41, actual 39)
For the men the average difference was -1.075 (-0.846 if you take out JS, but since it’s such a small sample anyway, might as well keep him in). Only 4 people were exactly as predicted, and there were 8 people either one up or one down out of 39 people who were in the prediction and who raced. About 1/3 right…good in baseball, maybe not so good in predictions?
LAPOINTE Laval -10 (predicted 11, actual 1)
FRÉCHETTE UQTR -8 (predicted 31, actual 23)
ARCHER Concordia -7 (predicted 44, actual 37)
ROBIDAS UQTR -7 (predicted 33, actual 26)
PELLETIER UQAM -6 (predicted 30, actual 24)
LAFOREST Concordia -6 (predicted 38, actual 32)
THIBEAULT UQAC -6 (predicted 47, actual 41)
VIGER-BERNARD UQAM -5 (predicted 34, actual 29)
GUEND Concordia -4 (predicted 12, actual 8)
LAVOIE-TRUDEAU UQTR -4 (predicted 29, actual 25)
Of note that the prediction model predicted almost exactly (to less than a second, and adjusted assuming the provincials course was 10.5k) the finishing times of the following athletes:
The only prediction that close on the women’s side was LEMIEUX UQAM.