Just a suggestion: I think it would be helpful in visualizing the performance difference between the benchmark and an algorithm if we could get log scales on the dollar axis. In particular, it would help determine relative performance difference after a gap event. For example, if algorithm moves to cash before late 2008 and the benchmark doesn't, there's now a gap. Thereafter, in linear space, if both the benchmark and algorithm perform the same on a percent basis, the gap will naturally grow. In a log scale, this is prevented since percentage gains are addition.