Somewhere in that stream they said that IRV is good for third parties because the 2PD is more “fluid” (I saw no evidence) and that Approval leads to 2PD worse than IRV (which WDS did say was a thing to watch for, but Score Voting and STAR Voting help eliminate the problem of being forced to rate everyone you approve of equally.
Their argument was this: 51% vote for D and 49% for R. Now a Green Party enters, and if just three Green voters bullet-vote for the Green then the Republican wins. But what the IRV propagandists forget is that in that case bullet voting would be a bad strategy and in such a close election the democrats would not want to waste their power in the D vs R race. And besides, real voters are more complicated. Some may vote for the Repub and Green but not the Dem. And perhaps, if the Green is too liberal, maybe the Green does not deserve to win because he or she has a narrow appeal. A centrist would have better success.
But IRV does lead to 2PD:
1 G>R>D (could represent an environmentalist conservative)
10 D>R>G (could represent centrists who think G is too extreme)
We imagine that N will increase gradually over time as the Green gets more recognition. The scores are G=N+1, D = 55-N, R = 44. As long as N < 28, we have the usual “spoiler prevention” and D wins 55-45. But if N > 28, the Democrat is eliminated, and because of the centrists, R wins. And the mere fear that this can happen will leave G perpetually low on the rankings, unable to move up.
(With approval, G can just move up the scale without affecting the D vs R race at all.)