# 2014 medal projections: Some excruciating details

How did the Olympic results compare with all the things we could use to predict them? Glad you asked:

And this:

Basically, I’m comparing three different types of results:

1. Majors: How skiers fared in the last Olympics and the last two World Championships. The number in “Majors” is a median — it ignores any null results, and the 2013 World Championships are counted twice so that they’ll be weighted more heavily.

2. Cup: Median of the last four World Cup seasons, with the last two counted twice so they’ll be weighted more heavily.

3. 13-14: The 2013 and 14 World Cups and the 2013 World Championships. A simple median this time, with no extra weighting.

Then for each skier, I calculated the difference between those numbers and his Olympic finish. Then I took the top 10 from the Olympics and calculated the absolute value of each difference. (In other words — I just want to know how far away from reality it was, so finishing four places higher than projected would be the same as finishing four places lower.)

So at bottom right, I took the median of each of the groups of differences. And that gave me a way of comparing which group of numbers was better for projecting medal results.

For the downhill, the 2013-14 numbers were better than the World Cup results, but the Cup results were much better than the majors. For the super-G, the majors were better, but I think that’s skewed by what I will refer to by a name I hope will catch on in statistics — the Weibrecht factor. That’s Andrew Weibrecht, who took bronze in 2010 and did little else in the intervening years before taking silver in 2014.

This is really too much to do for every event, but I think this exercise has pointed me toward a points system I’ll use for predictions going forward. I may do a few more winter events to refine the points system — it’ll have to be adapted for sports that don’t do World Cups and World Championships on the typical winter sports schedule, anyway.

But the next step, starting in a month or so — 2016. And we’re going to have easy-to-read charts of each athletes’ past performances, all leading to a predictive index.