<div dir="auto">An example of a lottery L that could be better (in some contexts) than random ballot favorite: let L(Y) be the percentage of ballots B on which Y is the highest ranked alternative that has a beatpath to each of the other alternatives that are ranked on ballot B.<div dir="auto"><br></div><div dir="auto">This choice of lottery L would (1)make compromising less of a temptation, and (2) make winning ties less likely.</div><div dir="auto"><br></div><div dir="auto">Another example: Let L(Y) be the percentage of ballots B on which Y is the alternative with the worst rank on ballot B that is not pairwise defeated by any alternative with a better rank on ballot B.</div><div dir="auto"><br></div><div dir="auto">In general, L(Y) should be the percentage of ballots B on which Y would be an optimal compromise alternative for the ballot B voter in a Plurality election.</div><div dir="auto"><br></div><div dir="auto">One more: Bubble sort pairwise the alternatives ranked on ballot B, giving rectification priority to the out of order pairs closest to the bottom (worst end) of the ballot. Then let L(Y) be the percentage of the ballots on which Y ends up on top.</div><div dir="auto"><br></div><div dir="auto">Finally: For each ballot B, while the alternative h currently ranked highest on B is covered by some alternative y ranked lower on B, swap out t for the highest such y. Then let L(Y) be the percentage of ballots on which alternativeY ends up top ranked.</div><div dir="auto"><br></div><div dir="auto">More ideas for lottery L?</div><div dir="auto"><br></div><div dir="auto">-Forest</div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Tue, Oct 11, 2022, 5:44 PM Forest Simmons <<a href="mailto:forest.simmons21@gmail.com">forest.simmons21@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="auto">In the quoted text below I gave a slight generalization of Friendly Voting (FV) in a formulation that will be more convenient for the voting method criteria proofs offered in this message (EM List posting).<div dir="auto"><br></div><div dir="auto"><div dir="auto" style="font-family:sans-serif">"Let L be any proportional lottery on the alternatives.</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">Elect argmin S(X), given by</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">Sum over Y of d(X,Y)*L(Y),</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">Where d(X,Y) is the number of steps in the shortest beatpath from X to Y.</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">When L is the random ballot favorite lottery, the above method description becomes an equivalent formulation of Friendly Voting."</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">First, FV is Landau efficient:</div><div dir="auto" style="font-family:sans-serif">Suppose that X is the FV winner and X' covers X. Then if there is a beatpath from X to Y of length d(X, Y), then replacing X with X' in that beatpath will give a beatpath of the same length from X' to Y. If X' directly defeats any later member of that beatpath, then d(X',Y) will be strictly less than d(X,Y). because of the shortcut ... ETC</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">Next, Clone Independence:</div><div dir="auto" style="font-family:sans-serif">As Kristofer pointed out to me, cloning a member Z of the shortest beatpath from X to Y doesn't change the length of the shortest beatpath, because you can just replace Z with any of its clones.</div><div dir="auto" style="font-family:sans-serif">So it was Kristofer who gave us the courage to use the number of steps in the shortest beatpath, rather than the customary "strength of the weakest link" metric used in the (Markus Schulz) CSSD Beatpath method.</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">Monotonicity:</div><div dir="auto" style="font-family:sans-serif">Suppose the winner X moves up in rank on one or more ballots, while the other alternatives maintain there ranks relative to each other. It is obvious that this change cannot increase the value of S(X). But could it decrease the value of some S(Z) more than it decreases the value of S(X)?</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">Well, if S(Z) decreases, it has to be because, for some Y, d(Z,Y) has decreased, i.e. the shortest beatpath from Z to Y just got shorter due to a shortcut newly created by a defeat of some alternative A that was not defeated before X got raised in the ranks.</div><div dir="auto" style="font-family:sans-serif">Let's compare the changes:</div><div dir="auto" style="font-family:sans-serif">Before the change the minimal beatpath from X to Y had a length of d(X,Y), and after the change it had a length of d(X,A) plus d(A,Y). Meanwhile the length of the shortest beatpath from Z to Y changed from d(Z,X)+d(X,Y) to</div><div dir="auto" style="font-family:sans-serif">d(Z,X)+d(X,A)+d(A,Y).</div><div dir="auto" style="font-family:sans-serif">In both cases the net effect on the Y term in the S sum of the distances is d(X,A)+d(A,Y)-d(X,Y).</div><div dir="auto" style="font-family:sans-serif">In summary, S(Z) cannot decrease more than S(X) does when X is raised on a ballot.</div><div dir="auto" style="font-family:sans-serif">[The other factor, L(Y), in the corresponding terms of the sum S does not depend on X or Z]</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">So, barring some subtle, but fatal error in my reasoning, it appears that the method is (1) Clone independent, (2) Landau efficient, and (3) Monotonic.</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">If the simulations of Kevin and Kristofer (and anybody else who wants to experiment) continue to confirm these claims, we can be increasingly confident that FV is worth proposing for public elections.</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">Kristofer and Kevin have done most of the heavy lifting ... I'm just the lucky guy who got to help out a little with the window dressing!</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">Thanks!</div><div dir="auto" style="font-family:sans-serif"><br></div><div dir="auto" style="font-family:sans-serif">Forest</div><div dir="auto" style="font-family:sans-serif"><br></div></div><br><br><div class="gmail_quote" dir="auto"><div dir="ltr" class="gmail_attr">---------- Forwarded message ---------<br>From: <strong class="gmail_sendername" dir="auto">Forest Simmons</strong> <span dir="auto"><<a href="mailto:forest.simmons21@gmail.com" target="_blank" rel="noreferrer">forest.simmons21@gmail.com</a>></span><br>Date: Tue, Oct 11, 2022, 10:07 AM<br>Subject: Re: Friendly Voting<br>To: Jobst Heitzig <<a href="mailto:heitzig@pik-potsdam.de" target="_blank" rel="noreferrer">heitzig@pik-potsdam.de</a>><br>Cc: Kristofer Munsterhjelm <<a href="mailto:km_elmet@t-online.de" target="_blank" rel="noreferrer">km_elmet@t-online.de</a>><br></div><br><br><div dir="auto">Jobst,<div dir="auto"><br></div><div dir="auto">Your remark connecting Friendly Voting to the benchmark lottery suggested the following generalization of FV:</div><div dir="auto"><br></div><div dir="auto">Let L be any proportional lottery on the alternatives.</div><div dir="auto"><br></div><div dir="auto">Elect argmin S(X), given by</div><div dir="auto"><br></div><div dir="auto">Sum over Y of d(X,Y)*L(Y),</div><div dir="auto"><br></div><div dir="auto">Where d(X,Y) is the number of steps in the shortest beatpath from X to Y.</div><div dir="auto"><br></div><div dir="auto">If L is the benchmark lottery, we get Friendly Voting.</div><div dir="auto"><br></div><div dir="auto">Kristofer invented the key concept of the important role of "friendliness:" The fewer beatpath steps from X to Y the friendlier Y is to X.</div><div dir="auto"><br></div><div dir="auto">All I did was make the connection to the generalized median concept. It was Andy Jennings and Rob LeGrand (and you) who got me thinking about the median as a way of lowering incentive for insincere strategy.</div><div dir="auto"><br></div><div dir="auto">All My Best,</div><div dir="auto"><br></div><div dir="auto">Forest</div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, Oct 10, 2022, 11:36 AM Jobst Heitzig <<a href="mailto:heitzig@pik-potsdam.de" rel="noreferrer noreferrer" target="_blank">heitzig@pik-potsdam.de</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div>Hi guys!<br><br>I like this a lot since it makes really smart use of the actual ballots rather than just the defeat matrix, and it treats voters' favourites seriously and as a kind of benchmark, almost like in MaxParC :-)<br><br>Best regards from the workshop on Algorithmic Technology for Democracy,<br>Jobst<br><br><br><div class="gmail_quote">Am 10. Oktober 2022 17:03:52 MESZ schrieb Forest Simmons <<a href="mailto:forest.simmons21@gmail.com" rel="noreferrer noreferrer noreferrer" target="_blank">forest.simmons21@gmail.com</a>>:<blockquote class="gmail_quote" style="margin:0pt 0pt 0pt 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<div dir="auto">Friendly Voting is a form of Generalized Median Voting (GMV) adapted to Ranked Choice Ballots.<div dir="auto"><br></div><div dir="auto">GMV methods elect the candidate whose total distance to the ballots is minimal. Friendly Voting gauges the distance from a candidate X to a ballot B as the number of steps in the shortest beatpath from X to ballot B's first place favorite f(B).</div><div dir="auto"><br></div><div dir="auto">Example: Consider the ballot set profile...</div><div dir="auto"><br></div><div dir="auto">x ABC</div><div dir="auto">y BCA</div><div dir="auto">Z CAB</div><div dir="auto"><br></div><div dir="auto">The cyclic beat order is ABCA.</div><div dir="auto"><br></div><div dir="auto">So the total distance from A to the ballot set is ...</div><div dir="auto">x×d(A,A)+y×d(A,B)+z×d(A,C),</div><div dir="auto">which simplifies to</div><div dir="auto">0+y+2z.</div><div dir="auto">Similarly, the total distance from B to the ballot set is</div><div dir="auto">2x+0+z,</div><div dir="auto">and the total distance from C to the ballot set is</div><div dir="auto">x+2y+0</div><div dir="auto"><br></div><div dir="auto">If we subtract x+y+z from each of these totals, the respective scores become ...</div><div dir="auto">z-x, x-y, and y-z.</div><div dir="auto"><br></div><div dir="auto">Candidate A wins if z-x is the smallest of these, i.e. if fpC-fpA is smallest, i.e. if ...</div><div dir="auto">fpA-fpC is largest, i.e. larger than both fpB-fpA and foC-fpB.</div><div dir="auto"><br></div><div dir="auto">So this seems to be the appropriate generalization of the fpA-fpC method that Kristofer has persisted in pestering Kevin and me about for about half a decade!</div><div dir="auto"><br></div><div dir="auto">Thanks, Kristofer !!!! Bullseye🎯</div><div dir="auto"><br></div><div dir="auto">-Forest</div><div dir="auto"><br></div><div dir="auto"><br></div></div>
</blockquote></div><br>-- <br>Diese Nachricht wurde von meinem Android-Gerät mit K-9 Mail gesendet.</div></blockquote></div>
</div></div>
</blockquote></div>