If you take Samus with 60% and instead Mario wins with 60%, well, that was clearly a foolish pick, and I think you deserve to be scored like a fool.

Not a good example because no one was taking anybody with 60% in that match. Clearly taking Samus in that match was not foolish in the last bit. Only 3 people out of 68 had either character with over 55%. Taking Mario with a high percentage was also not really related to skill - as is easily evident if you look at the final ranks of the people who had the top 10 best picks in that match: 57, 37, 75, 11, 33, 4, 34, 63, 51, 41

Let's not get too hung up on this single Samus/Mario match anyway. The important thing to understand is that in such toss-up matches, success is mainly a noise process in relation to skill (not entirely but primarily). i.e. mainly luck. The main skill involved really would be choosing the most middling pick possible, a risk-aversion skill, due to the exaggerated penalty as has already been discussed.

If we're using Samus/Mario as some sort of classic example then this is readily seen by doing some quick math. The correlation between the percentage people picked for Mario in that match against the amount of points people ended up with excluding the points from that match itself: 0.0112. Clearly almost a pure noise process. Now if you correlate the points from that actual match (instead of the prediction percentage) vs. the points from all other matches, it rises to 0.066; extremely low and possibly random but the slight correlation is probably due to risk-aversion being rewarded (since points gained are rules-related, whereas pick percentages are not).
---
www.SC2K5.com
www.sc2k5.com/gallery