The Day 1 gave a start for marathon finals. Read my previous post to get an information about the beginning of the match.
The provisional results are available here.
There is an interesting fact about chokudai’s strategy: he did not use any local testing tool, even batch testing. Instead he submitted every minor change to his solution to the testing system. In my opinion, this was a huge disadvantage for him.
Each participant was to use the same Topcoder computer during the marathon finals. On the other hand, algorithm track finalists may use their own laptops.
It never happened before, and now again Milanin did not submit until the last hour of the competition. The day before the finals I talked to him, and he said that he would not like to see in the finals a game task, or something that would be dependent on the parameter tuning. Games are not his favorite because it is necessary to spend a considerable amount of time to prepare infrastructure for the actual algorithm. As a result, the time left to prepare the actual strategy is not enough. He expressed an idea of creating some machine learning tool that could split marathon tasks’ input parameters into smaller groups, so that it is possible to solve each group independently. Another fact is that his father, mi5 also successfully participates in marathon matches.
wleite did his main strategy as well. He tried some different ideas and compared them, combining successful ones.
It was not enough to solve the 250-point task this time. -Xray- did solve it and made +1/-2 challenges, but did not qualify. His sad story is that he was to challenge scott_wu’s easy, but instead of the challenge result he got “Competition ended” verdict.