The current metric involves that some “truth” model was used to form groundtruth. Otherwise if only binary answers were used it turns out that “truth” probabiliy equals 0 or 1.
Can anyone comment how groundtruth was obtained and what factors caused for choosing of such metric.
Thanks.
look up logloss (the metric used to score this competition) on wikipedia.or, better yet, see this: http://www.exegetic.biz/blog/2015/12/making-sense-logarithmic-loss/