<p><br />
<br />
---------------------------- Original Message ----------------------------<br />
Subject: [EM] "Condorcet winner" versus "winner of Condorcet's method" (was Re: 2018 Chess Candidates Tournament)<br />
From: "Steve Eppley" <SEppley@alumni.caltech.edu><br />
Date: Wed, March 28, 2018 11:52 am<br />
To: election-methods@lists.electorama.com<br />
--------------------------------------------------------------------------<br />
<br />
> @Ross Hyman: Ding Liren was not a Condorcet<br />
> winner in that chess tournament, because a<br />
> Condorcet winner is an alternative that<br />
> defeats all other alternatives pairwise. <br />
> Ding Liren didn't defeat all other players;<br />
> he won only one game.</p><p>yes, i would not call that the CW.</p><p><br />
> Some people might prefer a weaker,<br />
> non-standard definition of Condorcet winner:<br />
> a candidate that's undefeated pairwise. (Like<br />
> Ding Liren.) In public elections the two<br />
> definitions (if implemented by two voting<br />
> methods) would behave the same with regard to<br />
> the incentives on candidates, potential<br />
> candidates, voters, parties, donors, etc.,<br />
> because ties are rare when there are many<br />
> voters, as there are in public elections.<br />
</p><p>and the reason for that is that with an electorate of decent size (like at least hundreds of voters) the probability of a tie in any pairing of candidates is very low.</p><p><br />
> Don't be misled the way many people have<br />
> been, especially mathematicians not familiar<br />
> with the social choice theory literature. <br />
> They wrongly believe "Condorcet winner" means<br />
> the winner according to Condorcet's method,<br />
> and thus that Condorcet's method simply<br />
> elects the candidate that defeats all others<br />
> pairwise, and is indecisive when no such<br />
> candidate exists. "Condorcet winner" is a<br />
> term of art (a.k.a. jargon). Unlike Borda<br />
> winner, which is not a term of art and merely<br />
> means the winner according to Borda's method,<br />
> and Black's method, which is not a term of<br />
> art and merely means the winner according to<br />
> Black's method, etc.<br />
><br />
> Because sometimes there is no candidate that<br />
> defeats all others pairwise, the confusion<br />
> has caused a number of writers to wrongly<br />
> claim Condorcet's method is often indecisive<br />
> and therefore unsuitable for elections.</p><p>i just read what i see here and what i see in the EM wiki and in Wikipedia. i hadn't thunk there was a "Condorcet's method" but that there are a few decisive methods that are "Condorcet compliant", which means these methods
will elect the CW **if** a CW exists (and i really think that in most public elections with a ranked-order ballot, that a CW will exist virtually all of the time, and most of the time, i'll bet that the IRV method will also elect the CW, but not always).</p><p>> (In simulations with random
voting, the frequency<br />> of scenarios in which no candidate defeats<br />
> all others increases asymptotically to 100%<br />
> as the number of candidates increases to<br />
> infinity, and as the number of voters<br />
> increases.) But the voting method Condorcet<br />
> promoted in his famous 1785 essay is very<br />
> decisive:<br />
><br />
> CONDORCET'S METHOD (copied from page lxviii<br />
> of his 1785 essay):<br />
><br />
> Here's its literal translation to English:<br />
> "The result of all the reflections that we<br />
> have just done,<br />
> is this general rule, for all the times when<br />
> one is forced to elect:<br />
> one must take successively all the<br />
> propositions that have<br />
> the plurality, commencing with those that<br />
> have the largest,<br />
> and pronounce the result that forms from<br />
> these first<br />
> propositions, as soon as they form it,<br />
> without regard<br />
> for the less probable propositions that<br />
> follow them."<br />
><br />
> The phrase "this general rule, for all the<br />
> times when one is forced to elect" meant he<br />
> was referring to a very decisive voting method.<br />
><br />
> A "proposition" is a pairwise statement like<br />
> "x should finish ahead of y." It has the<br />
> plurality if the number of voters who agree<br />
> with it exceeds the number of voters who<br />
> agree with the opposite proposition.<br />
><br />
> "Taking successively commencing with the<br />
> largest" means considering the propositions<br />
> one at a time, from largest to smallest.<br />
> (Like MAM and Tideman's Ranked Pairs do. <br />
> However, MAM and Ranked Pairs measure size in<br />
> different ways: MAM measures the size of the<br />
> majority, whereas Ranked Pairs subtracts the<br />
> size of the opposing minority from the size<br />
> of the majority.<br />
</p><p>that's RP-margins. there is also RP-winningVotes. how does this method from Condorcet differ from RP-winningVotes?</p><p><br />
> The word "plurality" can<br />
> mean either of those: either the larger<br />
> count, or the difference between the larger<br />
> count and the opposing count.)<br />
><br />
> The "result" is an order of finish, like "x<br />
> finishes ahead of y, y finishes ahead of z,<br />
> etc." It's a collection of pairwise results,<br />
> each of which is obtained either directly<br />
> from a proposition that has a plurality, or<br />
> transitively from a combination of pairwise<br />
> results obtained directly. An example of a<br />
> pairwise result obtained transitively is the<br />
> pairwise result "x finishes ahead of z"<br />
> obtained transitively from "x finishes ahead<br />
> of y" and "y finishes ahead of z." By<br />
> definition, an order of finish is an<br />
> ordering, and is thus transitive and acyclic.<br />
><br />
> "Without regard for the less probable<br />
> propositions that follow" means disregarding<br />
> propositions that conflict (cycle) with the<br />
> results already obtained from propositions<br />
> that have larger pluralities.</p><p>I cannot see how that differs from Ranked Pairs.</p><p><br />
> For example,<br />
> disregarding "z should finish ahead of x"<br />
> after having obtained the pairwise results<br />
> that "x finishes ahead of y" and "y finishes<br />
> ahead of z."<br />
><br />
> Note: No language in the definition of<br />
> Condorcet's method refers to an alternative<br />
> that defeats all others pairwise. (Nor to an<br />
> alternative that's undefeated pairwise.) <br />
> Although it can be deduced that Condorcet's<br />
> method will elect an alternative that defeats<br />
> all others, it will also elect an alternative<br />
> even when no alternative defeats all<br />
> others... in other words it's very decisive. </p><p>so this historical "Condorcet's method" always elects a single-winner and, **if** a pairwise champion exists, it will elect that pairwise champion. so "Condorcet's method" is Condorcet-compliant.</p><p><br />
> People who write about "Condorcet completion"<br />
> rules -- first check whether there exists an<br />
> alternative that defeats all others and then,<br />
> if no such alternative exists, proceed in<br />
> some other way to find the winner -- have<br />
> misunderstood Condorcet's method,<br />
</p><p>or, perhaps we haven't heard of Condorcet's "method". but if they apply "Condorcet completion" rules to another Condorcet-compliant method that doesn't need completion rules (such as RP or Schulze), i think that reflects the same
misunderstanding.</p><p><br />> which is<br />
> already "complete" (very decisive when there<br />
> are many voters, because when there are many<br />
> voters it's rare that any two majorities are<br />
> the same size, and rare that any pairings are<br />
> ties).</p><p>yup.</p><p>thanks for the information, Steve.</p><p><br />
--<br />
<br />
r b-j rbj@audioimagination.com<br />
<br />
"Imagination is more important than knowledge."<br />
</p><p> </p><p> </p><p> </p>