With regard to assumptions, the usual distinction between determinism and statistics can confuse rather than clarify an issue. Determinism can simply mean there is the assumption that there is a right or definitive value. Whereas statistics is merely meant to mean that there is no sure right value but only an estimate of value, some estimates being better than others. Roughly speaking, this is the journey that physics has traveled from classical mechanics to quantum mechanics. But it need not apply just to the latter. Engineers are said to be already familiar with the Heisenberg uncertainty principle. (And Paul Dirac, who first related quantum theory to special relativity, was an engineer.) In the case of special relativity, traditionally regarded as a classical theory, it turns out that the “deterministic” variables of the theory are really implicit averages. But the ranges, of which they are averages, remain implicit, except in the realm of high-energy physics. Likewise with a deterministic assumption in election method that there is some definitive value of election results, rather than just a process of arriving at better estimates. The “Condorcet winner” is a case in point. I found it worked at least as well as Borda method for single vacancies,(the two rational methods agreed unlike their unrational counterparts)  even for a very finely balanced vote for the candidates, provided the Condorcet pairings were proportionally weighted to the respective votes for paired candidates. But I guessed the reason for this was that in the case of single vacancies, a single preference is all that is required. Thus, the order of importance, that voters attach to the candidates, is of minimal importance, because only one candidate will be elected anyway. Therefore, it matters more that one try out all the logical alternatives, of single candidate preference, for a Condorcet winner. But in multi-member constituencies, the order of importance, to the voters, becomes more and more important, the more seats there are in the constituency. And as the Condorcet winner does not take into account this range of relative choice, it becomes less and less relevant, to more multi-member constituencies. Regards, Richard Lung.

On 2024-05-24 09:58, Richard Lung wrote: > But in multi-member constituencies, the order of importance, to the > voters, becomes more and more important, the more seats there are in the > constituency. And as the Condorcet winner does not take into account > this range of relative choice, it becomes less and less relevant, to > more multi-member constituencies. I agree that bloc Condorcet is a bad idea for multiwinner, at least if you want proportionality. I would support something like Ranked Pairs loser STV or one of the multiwinner Condorcet generalizations (Schulze STV, CPO-STV) instead. The main problem with Schulze STV and CPO-STV is that if you thought single-winner Condorcet methods were hard to understand, you haven't seen nothing yet. If I remember correctly, Schulze STV's comparison function involves solving a linear program for each comparison. -km