Speaking as a high level quant guy, you can get into a lot of mischief when you attempt to move from individual-level analysis to more aggregated forms of analysis (see Simpson's paradox for an interesting example).
Moreover, there are quite a few potential confounding variables to think about, primarily related to quality of competition. For example, if your trend line is moving upward for quality of competition during the pre-post period, you would need to control for that (so as to understand the degree to which people are losing because of the pre/post issue rather than just facing tougher guys). Ditto with understanding where fighters are on a career arc...if the pre/post period covers a meaningful period of time, the pre-period might be associated with pre-(natural) decline, and the post-(natural) decline, even ignoring PED use/abandonment.
I haven't dug into any of this, so perhaps someone has already done this.