When voting systems are evaluated, tactical voting is considered a problem. I agree to the extent that an average voter has enough information to successfully tactically vote. For example, in a FPTP election, I can look at any poll and realise that my preferred third-party candidate has no chance of winning and redirect my vote to one of the more popular candidates.
But in any more complex system, a particular tactical vote seems like it's only actually tactical if you have a lot of knowledge which in my experience as a voter isn't typically available.
In a simple example of single transferable vote, let's assume there are three parties, left (L), centre (C), and right (R), and all voters for (L) and (R) give their second preference to (C), and voters for (C) give their preferences equally between (L) and (R). Let us say that the voting proportions are roughly L = 40%, C = 30%, R = 30% according to a poll. If my preferred candidate is (L), I could tactically vote (R - L - C) to "ensure" that (C) is eliminated first, ensuring that (L) receives 15% extra from (C)'s preferences, and defeating (R) by 55% to 45%. This would be better than if I voted (L - C - R) according to my true preference, because then (R) might be eliminated first, allowing (C) to defeat (L) by 60% to 40%. All power to the tactical vote.
But real life is much more complicated. I can't really trust the poll, and maybe the difference between (L) and (R) is a little less, and the voters from (C) are leaning a bit more to (R). For all I know, my vote might well allow (R) to actually defeat (L) and win the election. Especially if I consider that other people might be tactically voting. This is surely relevant: whether to tactically vote depends on how many others will be tactically voting (and which way).
Given this, why is the theoretical of tactical voting under perfect information considered such a big deal? I would be interested in evaluating a voting system by, for example, whether if you could estimate the likely outcome to within 4%, is tactical voting still possible. To me, as a lay person, that would meaningfully illuminate differences between systems, but I can't find evaluations of voting systems according to these criteria. So my secondary question is, do such assessments exist?