Take it from a practicing quant - there are way too many observations to draw any conclusions (one way or another) that the simulation is working or not. What this means is that even it appeared that McGwire, for example, had sim performance close to his real-world performance, you could still not conclude that the sim was working correctly.

As I said, it’s all about sample size, and the 7 observations that DaveF took are about 10% of what you need to get a statistically reliable estimate. In other words, the mean and standard deviation derived from his 7 observations are themselves unreliable estimators of their true (i.e., sim) values.

Also, Dave has his facts wrong about the standard deviation: the range around the mean is two-thirds (67%), not three-quarters (75%) as he says.