> > It tells you how often you have to be thinking about "playing the > strategy game" to improve the outcome (or how often you may regret if > you don't). In a low manipulability method, you don't have to start > thinking about whether you should tailor your response to poll data, > etc. as much. > Chances of being able to vote honestly, with no strategic burden to bear. > That's what the manipulability numbers are about. @Filip Ejlak <tersander@gmail.com> @Kristofer Munsterhjelm <km_elmet@t-online.de> I think this paper may interest you <https://www.journals.uchicago.edu/doi/pdf/10.1086/726943>, because they measure something closer to what both of you were talking about: what are the chances that I, a typical voter, will have no strategic burden to bear? The results have the opposite conclusion from the manipulability experiments by KM. The chances you'll have to "play the strategy game" are higher in IRV *than in FPP*! As a statistician, my strongest piece of advice for interpreting probabilities: "never calculate a probability without rigorously defining your sample space first". If you care about the probability that a *person* will have to vote strategically, you can't substitute that for the probability that an *election* involves strategy. Corporations elections are not people! If what you care about is the amount of strategic burden on a voter, you need to calculate *that *number, not a *different* number. Which is a big part of why approval/score/etc. are some of my favorite methods, by the way. None of this messing around with 12-d chess to work out the best ranking. I just place my approval cutoff between the frontrunners, who I can find by Googling "538 polling average". Then I go about my day as usual, knowing I've done my part to guarantee majority-rule will prevail (because this strategy ends with the Condorcet winner being elected).