Problem in Schulze_Proportional_Prefilter in Experimental/Schulze-STV branch about condorcet HOT 5 CLOSED

I have decided to give up on Schulze_Proportional_Prefilter. I don't know how to make it as aggressive as it can be without ever removing the wrong candidates.

In a recent commit, I started remaking Schulze-STV, extended from SingleTransferableVote. While it doesn't output the correct results for all the Tideman tests, the process runs very fast. I feel confident enough that if I ever get it to find the correct results, it will be fast enough that we don't need a prefilter.

No more prefilter. Issue closed.

from condorcet.

I just checked inside $election->pairwise->pairwise[6]. These values are also all set to 0.

from condorcet.

It turns out that Schulze_Core::makeStrongestPaths() has to be modified. The problem is that it sets a path from one candidate to another to 0 if they have a pairwise loss to the other candidate.

The modification I made & reverted earlier that uses $M as an argument would work. However, you may also just remove the condition if ($election->getPairwise()[$i]['win'][$j] * $M > $election->getPairwise()[$j]['win'][$i]) and it will probably give the same results to the Schulze Method.

from condorcet.

@LiamM32 Is it still an open issue?

from condorcet.

Is it still an open issue?

Maybe. I haven't solved it. But it might just be a matter of the whole algorithm being flawed. Perhaps we should contact Schulze himself to ask about this.

But I think it's most likely that the algorithm itself is flawed. Based on the pairwise results, it indeed should be eliminating the candidate that it eliminates according to the algorithm I envisioned. But somehow, this candidate 'K' is supposed to be one of two winners in Tideman A22, according to the Schulze program.

Unfortunately, I don't understand Schulze STV enough to understand how K wins in this case.

from condorcet.