<div> <font face="Arial, Helvetica, sans-serif">Jobst wrote:<br>
> </font> 45: A 100 > C 90 > B,D 0 <br>
<font face="Arial, Helvetica, sans-serif">> </font> 44: B 100 > C 90 > A,D 0 <br>
<font face="Arial, Helvetica, sans-serif">> </font> 11: D 100 > A,B,C 0 <br>
</div>
<div> <font face="Arial, Helvetica, sans-serif"><br>
Yeah, I was thinking that. The problem is that a voter who votes one candidate max and the rest zero is basically saying that they aren't willing to cooperate and have locked their vote into that candidate. There needs to be a way to allow a sub-group to cooperate.<br>
<br>
One option would be to split out any voters who don't agree with the largest compromise and apply the process recursively. Recursion would end when a group converges to a single candidate at 100%.<br>
<br>
In the above example, after "round 1", the 89 voters will all agree to increase the probability of option C at the expense of A, B and D.<br>
<br>
The voters are then split into two groups and the process is applied to each sub-group.<br>
<br>
group 1: (89%) <br>
</font><font face="Arial, Helvetica, sans-serif">> </font> 45: A 100 > C 90 > B,D 0 <br>
<font face="Arial, Helvetica, sans-serif">> </font> 44: B 100 > C 90 > A,D 0 <font face="Arial, Helvetica, sans-serif"><br>
<br>
</font><font face="Arial, Helvetica, sans-serif">starts with 50.56% A and 49.44% B<br>
<br>
will converge to 100% candidate C<br>
<br>
group 2: (11%)<br>
</font><font face="Arial, Helvetica, sans-serif">> </font> 11: D 100 > A,B,C 0 <br>
<font face="Arial, Helvetica, sans-serif"><br>
starts with 100% D<br>
<br>
will converge to 100% candidate D<br>
<br>
Result being C 89% and D 11%<br>
</font><font face="Arial, Helvetica, sans-serif"><br>
This is marginally better than 88%/11%/1%, assuming that they are honest ratings (but perhaps it is unfair to the 'A' voters).<br>
<br>
The range winner is C ofc.<br>
<br>
Also, if this process was used to elect a legislature, then it might be possible to ensure some additional stability. <br>
<br>
For example, if all probabilities are rounded to 1%, you could create 100 possible results such that a candidate who gets an 11% chance of election is in 11 of them. However, 2 members of the same party must be placed in all of the 100 before you can place a 2nd member of any party in a result.<br>
<br>
Ofc, if that was the case, it suddenly becomes a reasonable strategy to rate party candidates at 100% and the rest at zero in order to get your favourite party as many seats as possible.<br>
<br>
It would effectively become a PR system.<br>
<br>
Raphfrk<br>
</font><font face="Arial, Helvetica, sans-serif"></font><!-- end of AOLMsgPart_0_f37066d7-e852-414c-9c6e-e3f4baba5472 -->
</div>
<div class="AOLPromoFooter">
<hr style="margin-top:10px;" />
AOL's new homepage has launched. Take a <a href="http://info.aol.co.uk/homepage/" target=_blank>tour</a> now.
</div>