[EM] VoteFair Ranking software version 6.0 in C++ with MIT license

[VoteFair]

On 1/5/2020 5:48 AM, Kristofer Munsterhjelm wrote:
 > Do you agree with my point about the use of the term "mathematically
 > equivalent"? If the method appears to be equivalent, that might give the
 > impression that you don't need to care about cycles at all.

VoteFair popularity ranking IS mathematically equivalent to the 
Condorcet-Kemeny method. The difference is that John Kemeny described 
the version that counts disapprovals (and finds the smallest sequence 
score), whereas VoteFair popularity ranking counts approvals (and finds 
the largest sequence score).

The VoteFair Ranking software does calculate the full version (of 
VoteFair popularity ranking) to the extent that the 
"global_check_all_scores_choice_limit" value is as large as desired.

I choose to leave that value at 6, even though it could be set to 20 or 
higher with no execution errors.  Of course in that case it would be 
helpful to insert code that checks for time delays and shows progress in 
case a Condorcet cycle is slowing it down.

I'm not sure what you mean by:
 > ... that might give the
 > impression that you don't need to care about cycles at all.

Neither the method nor the software implies that cycles are of no 
importance.

Richard Fobes


On 1/5/2020 5:48 AM, Kristofer Munsterhjelm wrote:
> On 04/01/2020 01.58, VoteFair wrote:
>> On 1/2/2020 3:00 PM, Kristofer Munsterhjelm wrote:
>>> ...
>>> The VoteFair ranking is F>C>E>A>G>D>B with Kemeny score 118, but
>>> C>A>F>E>G>D>B has Kemeny score 119 and is the unique maximum ranking.
>>
>> Yes, the default value for variable
>> "global_check_all_scores_choice_limit" is 6 and it gives a result that
>> differs from the (full) Condorcet-Kemeny method.
>>
>> Increasing this variable to 7 causes the full Kemeny calculation process
>> to be done across all 7 choices, and it gets the same result as doing
>> the full Condorcet-Kemeny calculation method.
>>
>> This is consistent with what I said earlier.
>>
>> Of course the calculation time for the case in which there is a
>> Condorcet (rock-paper-scissors) cycle across 50 choices is going to be
>> way too time-consuming for this software.
>>
>> As I've also said before, the odds of such a cycle (involving either 7
>> or 50 choices) happening in a real election is lower than the odds of a
>> conventional tie.  Either kind of tie can be resolved using a judicial
>> decision -- like the Bush versus Gore decision from the US Supreme
>> Court, which was not even an exact tie.
>
> Do you agree with my point about the use of the term "mathematically
> equivalent"? If the method appears to be equivalent, that might give the
> impression that you don't need to care about cycles at all.
>