Hi Eric

I think you are approaching it the right way. I used to have some great games between engines and dedicated computers through the use of NPS. It gives you the feeling that the best moves that each computers considers are played this way.
I tried SF16-1-2200 ELO the other day against TM Lyon with test score of 2223. SF16-1 playing at 30 secs and TM Lyon 2hrs/40 speeded up to 400% through Mess. Even though the game played out correctly and competitively, you can observe SF16-1 deliberately picking the occasional 2nd, 3rd, 4th best moves as its way of sticking to 2200 ELO. This in my opinion makes the games seem very artificial. SF16-1 is probably not alone with this approach, and I would bet most of the other ones do something very similar. Which means unless they deliberately pick a losing move you are never really going to win and at best draw. It feels so very condescending.

Best regards
Nick