[EM] RE: Election-methods digest, Vol 1 #388 - 7 msgs

[MIKE OSSIPOFF]

Someone wrote:

One of the side effect of the Markus and Mike chat is that I understand
Mike is programing something

I reply:

No. I'm not programming anything. Yes, quite some time ago I posted a Python 
program to implement BeatpathWinner. And yes, I re-posted the BeatpathWinner 
algorithm a few days ago, this time not in any particular programming 
language.

But I'm not now programming anything.

What started this discussion was when Markus said that my BeatpathWinner 
algorithm wouldn't work, because it isn't the Floyd algorithm, whatever that 
is.

The algorithm that Markus posted as the Floyd algorithm differs from mine 
(actually Steve Eppley's) by only making one pass through the 3-candidate 
permutations.

As I say in my reply to Ernie, my algorithm works. It looks at every 
permutation of 3 candidates, i, j, and k. If the beatpath from i to j, and 
the beatpath from j to k, are both stronger than the beatpath from i to k, 
then the value of the minimum of B(i,j) and B(j,k), which is the strength of 
the beatpath made by concatenating the ij and jk beatpaths, becomes the new 
value of B(i,k)--the strongest beatpath from i to j found as-yet at that 
time.

It repeatedly makes passes through the permutations until doing so doesn't 
make any changes. Then its task is completed. Each pass finds one or more 
new, longer beatpaths that replace a previous one that wasn't as strong.

The beatpaths that it initially looks at are single-step beatpaths, pairwise 
defeats. But, via the process described above, the algorithm eventually 
finds the strongest beatpath from each candidate to each other candidate.

If Markus believes that it doesn't work, I'd be curioius how he justifies 
that claim.But sometimes he says it merely takes longer to execute than the 
Floyd algorithm, and not that it doesn't work.

It was irresponsible for Markus to say that that BeatpathWinner algorithm 
wouldn't work, unless he can justify his claim, tell why he thinks it 
wouldn't work.

False statements, statements that he can't justify,  have always been 
Markus's stock-in-trade. But this time he's doing the disservice of 
misinforming people about a practical matter. But Markus doesn't care, he 
just enjoys being on the attack.

Now, Markus says that it's possible to find all the strongest beatpaths by 
making just one pass. He says the Floyd algorithm did that. According to 
websites that I looked at, the Floyd algorithm doesn't find strongest 
beatpaths, it finds shortest paths. I guess what Markus is saying is that it 
can be modified to find strongest beatpaths. He says that, by changing the 
order of the indexes in the line that tests and changes B(i,j) values, it 
can complete its job in one pass. But, if that's true when its job is 
finding the shortest path between each pair of graph-nodes, that may or may 
not mean that it's true when its job is finding the strongest beatpath from 
each candidate to each other candidate.

I'm not debating whether that's so. I don't know, and I don't reallly care. 
It doesn't matter because, though Markus says that the one-pass procedure is 
faster, both procedures are fast enough for all practical purposes.

As I said, the only reason why I said anything was because of Markus's 
mistaken statement that the BeatpathWinner algorithm wouldn't work.

Mike Ossipoff



That person continued:

... The more I think about it, the more I
believe it might be this:
http://www.fairvote.org/ChoicePlus/
http://votingsolutions.com/

I have nothing to do with the fairvote website. The other one, I'd have to 
check to find out what it is. But I'm not programming anything there. 
There's an interactive BeatpathWinner counting website, and it uses an 
algorithm similar to the one that I've posted here. But the website owner 
didn't get the algorithm from me, and I'm not working on it.

That person continued:

If it is... I will come with more on that...

I reply:

It isn't. I'm not programming anything there, and I'm not programming the 
BeatpathWinner algorithm. I did that a long time ago.
Mike Ossipoff

_________________________________________________________________
Grab our best dial-up Internet access offer: 6 months @$9.95/month.  
http://join.msn.com/?page=dept/dialup