It seems reasonable to expect that new humans joining the system should be no
better, nor worse, on average, than those who came earlier. This is indeed the
case, according to the data on fig. 3.9a: both good and not-so
good people keep joining the system. Tron agents (fig. 3.9b) do
Our attempt for progressively increasing the quality of new agents produced by the novelty engine, by having them train against those best against humans, was partially successful: graph 3.9b shows a marginal improvement on the average strength of new players, first to 2500-th. But noticeable better agents beginning at 2800 come to confirm the previous findings of other researchers [4,130] in the sense that the coevolving population used as fitness yields more robust results than playing against fixed trainers who can be fooled by tricks that have no general application. This point is discussed in detail in section 3.7.