Generally, if a game goes past 10 generations, players who engine-build and play VP will tend to win.
At 2p, this is basically always, so 2p games are a bit boring. (Last 2p game was 189 to 85, mostly because I ended the game early to get to bed). For 3p+, if you have enough players who will switch to terraforming if it looks like there is a player with an engine building lead, the game gets more interesting. If you have enough players with that attitude, in theory, pure terraforming strategies probably dominate. And again, the game gets a bit boring.
So, assign players a score of 1 or 0.5, depending on their likelihood of terraforming when they don't have the top engine. Random BGA players are about 0.75. (honestly, maybe a bit lower).
I prefer games with a total score from 2.5 to 3.5. So, for random selection on BGA, 3-4p, with 5p sort of acceptable and 2p ==no. 2p might be better if the global selection phase happened twice in every generation. (Might be a nice option). For >400 elo players, I'd prefer 3p to 4p, but that seems rare enough that it is interesting anyways.