The spoiler effect is real in FPTP when voting for a third party instead of your preferred major party. When you include not voting in the model, you’re effectively including the option to not vote as an option on the hypothetical ballot. The problem with this is that it changes how votes are counted, making one of the “candidate”’s votes all count for nothing. It’s now a different system with different properties. So the spoiler effect for that candidate disappears unless you get into counting opportunity costs, but accounting for opportunity cost brings into the realm of just arguing against FPTP, which I agree with.

I think it’s worth mentioning that voting is fundamentally a last resort for when consensus building is too slow or won’t scale to the desired population. I think it’s also worth mentioning that liberal democracy is essentially democracy by and for the bourgeoisie, and was certainly not designed by people who had access to Arrow’s Theorem or anything of the sort. So the mathematical issues of which voting systems are good for which types of consensus are really separate from the social and economic issues of political shells and class war. And all in all, I think FPTP has done wonders for maintaining the two party system in the US, which itself has done wonders for maintaining bourgeois control.