Comments/Ratings for a Single Item

This could be considered the well-known divide and conquer technique. In
politics, it's the spoiler phenomena. For this tournament, it will
likely
result in the majority of the games so named being eliminated. I have 2
suggestions:

A) Contact Mr. Duke, Mr. Joyce, and Mr. Paulowich, and ask them which
variant should be the one to be included in the poll. While Mr.
Capablanca
cannot be asked his preference, it could reasonably be assumed that the
game he designed would be the one he'd prefer. 

B) List only 5 games:
     Capablanca's
     Falcon
     Grand Shatranj
     Great Shatranj
     Mir
then let the players decide which version they want to play. In the case
of a tie, go with a predetermined game. While I may be mistaken, I
believe
Mr. Duke has expressed a preference for Falcon 100, Mr Joyce for the
'D'
variants, and Mr. Paulowich for Mir32.

Please note that I have listed 5 games, not 4. I do not find that there
is
a great similarity between Grand and Great Shatranj. I would point out
that
 these 2 games are doubly handicapped, as they are placed in a 4 way
contest between the 2 games. There are, in the D versions, 5 pieces
different of the 10 in the games. Promotion rules are different. The
boards also are of different sizes. What, exactly, is the criteria for
deciding difference? How do these games fall on one side of the line,
while games with apparently far fewer significant differences fall on the
other side? As a ridiculous example, Dimension X uses all the standard
chess pieces, and adds only 3 different pieces and more squares; Templar
Chess adds only 1 piece and a few more squares; Alice Chess only adds
more
squares. As another ridiculous example, the drop games all may be
considered quite similar. If different pieces, different boards and
different promotion rules don't differentiate adequately among games,
what does? One might gain the impression that these games are being
singled out for their theme. Yet in the same way, FRC and CRC are the
same, as is chess 256, chess 960....

I would respectfully ask that method B above be used for the tournament.
If the games really are that similar, it should not make much difference
which version is played in any game, should it?

This is not a divide and conquer technique. Multiple candidates from the same party can split the vote in a plurality election, such as is common in United States politics, but this poll is a ranked ballot condorcet poll. The condorcet method is not subject to the same problems as plurality voting, and in particular, multiple candidates from the same party will not split the vote in a condorcet election. Since the condorcet method is not used in most, if any, political elections, you are probably unfamiliar with it, and I would suggest reading more about it.

The purpose behind giving separate entries to different presets on the same page is to allow the participants of the tournament to democratically decide which of the very similar games they prefer to have in the tournament. There is no reason for anyone to split their votes among very similar games unless they have strong preferences for one over the other. I will not be changing this poll to fall in line with your suggestions.

In response to 'no one''s comments, the Condorcet (MAM) voting method
will not 'split the vote' between the very-similar games pointed out by
Fergus.  The method satisfies the following criterion:

     [from http://alumnus.caltech.edu/~seppley/]
     independence of clone alternatives (ICA, promoted by TN Tideman): 

     If there is a subset of alternatives such that no voter ranks any
     alternative outside the subset between any alternatives in the
     subset, then the election outcome must not change if a strict subset

     of that subset is deleted from the votes and from the set of
     nominees. 

A sample election, where B and B' are the 'clones' (they may be from
the same party, or just be very similar in some way):

40%: A > B > B'
30%: B > B' > A
30%: B' > B > A

In a plurality voting system, such as you might be used to, A would win
over B and B'.  But in any Condorcet method, B would win, because B would
win in an election against any single opponent.

I am both an American and quite familiar with the Condorcet technique. In fact, I have for a few years now advocated it as a more democratic method than any other voting technique that I'm familiar with, especially IRV or 'Instant Runoff Voting' which I believe to be even less democratic than the unfortunate, zero sum, winner takes all method which decides most elections in American politics. Because I have had experience as an elected officer for a third party in America (who was in fact elected by an alternate voting technique), I have had the incentive and opportunity to look into other modes of voting. In my opinion, IRV violates the 'one man, one vote' principle we in America have fought to uphold. So, it was with great pleasure that I saw the MAM Condorcet Voting method being used here on chess variants website.

So my initial criticism which began this aspect of this thread is not of the voting method for which I have the utmost respect.

However, I do persist in thinking that my concern is a good one and it has been left unaddressed. I believe that listing two versions of Great Shatranj and two versions of Grand Shatranj as well as two versions of Mir Chess will unfairly diffuse them. It would be perfectly alright with me if they had been included in the first poll and each qualified separately. But they were not, unlike Falcon Chess and Falcon Chess 100, each included and each separately qualified.

I had been wondering how this problem would be addressed. I can say that I believe that either method A or method B proposed by 'no one' are plans that would address my concerns. I am not 'no one' but I happen to know who 'no one' is. Either method seems more fair and more democratic than the one now being implemented.

Ah, I see. To be frankly honest, I wasn't paying sufficient attention to your stipulation and for that I apologize. I feel the way I do when I make a blunder in a chess game. Hm. But I think my position, though it may appear weakened, still has value.

You do say 'Unless the number of players justifies it...' You were surprised by the number of people who voted in the initial poll; you could also be surprised by the number participating in this one. I don't think that's likely, so that doesn't remain as a substantive concern.

I do still have concerns that I don't see you addressing. Perhaps you can.

I still do have the concern that some people will not meticulously list all four as you seem to suggest they should and the votes will diffuse and that caprice could end up diffusing them. With the exception of Falcon Chess and Falcon Chess 100, which I think independently qualified and should both be added if they are both ranked high enough, I am still in favor of 'no one's' A or B. I know someone who strongly dislikes cannons and because of that, he is likely to rank one version of Grand Shatranj much lower; someone else could do the reverse and the result is diffusion. On the other hand, both people who like Grand Shatranj might end up being disappointed to know that neither were included despite the fact that each voted for a version of it. Isn't there an assumption you are making that people who like Great Shatranj will tend to like both versions approximately equally, that people who like Falcon Chess will like Falcon Chess 100 almost just as much? I am someone who prefers less familiar games so I will rank Falcon Chess 100 higher. That's an example of how votes between Falcon Chess and Falcon Chess 100 could get diffused with neither of them qualifying (though I think it fair games since both qualified).

In actual fact, to sum up, I expect diffusion so the chances that any versions of Mir Chess, Grand Shatranj, Great Shatranj and Falcon Chess will show up are lessened and they are playing on an unequal footing. In my last comment, I suggested that we ran the real danger of both versions of both showing up. In fact, the opposite concern seems to be more germane, no?

What I am saying in both comments though is that people are unlikely to rank all versions of each similarly, though if they look at this thread, it may motivate them to do so, but how many will pay attention to this detail?

You're proposing this set of votes:

M36 > A > M32
M32 > A > M36

According to Stephen Eppley's MAM Calc script, which runs through the PHP
on my website at http://www.duniho.com/remote-mamcalc.php, A would win. The
condorcet method favors the least objectionable candidate to all voters. If
only two people were to play one game, and they chose which game from three
games, with votes like these, then A is the choice that is going to
minimize discontent. If one of the voters really preferred either version
of Mir Chess to A, voting to reflect his true preferences would allow one
of the versions of Mir Chess to win over A. The MAM method will reflect
voter preferences so long as people vote their true preferences. It can't
and shouldn't be expected to secondguess what anyone's preferences would
be. I have previously examined several voting methods. I consider the
condorcet method to be the best in general, and I consider MAM to be the
best version of the condorcet method.

Both times it told me that m won, a came in second and M came in third. I confess I do not understand this, as it certainly appears that the votes for M and m were exactly equivalent. Notice the only difference between this vote and the one previously discussed is the addition of a guaranteed loser in b. I am not trying to be a nitpicker. But I still have problems with how the vote turns out, depending on the number of entries, and a random factor. By adding an apparently ignorable loser, the outcome is changed from 'a' winning to one in which apparently either 'm' or 'M' can win, depending on the random factor mentioned in the Eppley calculator.

I would like the wider range of choices available for the tourney, given
there is no overriding reason to restrict them. [As long as there is not a
problem, I'd almost always opt for more.]
Also, I was looking over the poll, and noticed that Berolina Chess is tied
to the Avalanche rules page.

I am very happy with the selection of games we have from which to choose,
and the method for selection.  I do think that like 1.25x selection option
better.  Also, I definitely do not think that Janus Chess should be
included with the other Capablanca variants as 'very similar.' 
Personally I like Janus Chess the best, because it isn't flawed like
Capablanca, and it isn't random.  I would be happy to see CRC included
too, but I am somewhat less interested in seeing it.  I would be happy to
see both, and think both of them are quite different.  I think standard
Capablanca's Chess should not be included for reasons that have already
been well established.

Thanks, Fergus, for your vigilant attention to detail.  I am pleased at
the level of participation that it looks like we will have, I am looking
forward to a great tournament.  I just hope that no one quits and forfeits
all their games as has happened in the last two.

Fergus, thank you for your quick correction of your previous statement.

Not everyone would be that forthcoming. It leaves me with a better understanding of how the system works. And in particular, I see that it is apparently true that truncating makes little or no difference. I will also agree that the Condorcet method will apparently give the least objectionable results.

I do think that the change in the way games are voted on in this second poll does allow for the possibility of gaming the vote a bit. Specifically, I think some of the variants that are called clones are sufficiently different that they should get an appreciable difference in ranking. My personal preferences tend more toward the outre, so I would like to play CRC, Falcon 100, Mir36, and the D versions of Great and Grand Shatranj, rather than the other versions. I would have been inclined to rate what I would call the lesser versions below some other games, but I want to play some version of these games, so I am now inclined to rate the lesser versions directly below the preferred versions, and other games below them, to assist in any tiebreaks.

My question to you becomes then: why should not this change in my voting pattern benefit these games? For example, I would like to play Alice. But some of the variants I would previously have rated below Alice I would now rate above Alice. Does, or better, how does the Condorcet method alleviate what I see as a skew in the results, that I have deliberately introduced to counterbalance what I see as Mr. Good's valid objection, that running these games against each other in the second poll dilutes their votes?

I don't really understand the question, perhaps because I don't consider
Jeremy's objection a valid one. His concerns seem to presuppose an
incoherence in someone's preferences. On the one hand, someone has one
set of preferences that he expresses in his votes, and on the other hand,
he has a different set of preferences that leads him to be disappointed
with the results. You can't have it both ways. If someone accurately
votes his preferences, the only cause for disappointment in the results
should be that other people had different preferences.

But let's move on to your specific example. The most accurate way to rank
the games in question, i.e. those designated as very similar to other
games, in relation to the rest of the games is to rank each one as though
the other games it is very similar to were not in the poll at all. So when
you rank Mir32, for example, imagine that Mir36 is not in the poll, and
rank Mir32 accordingly.

as usual, thanks for organizing this, fergus.

i also vote for as many games as possible.

I have the general background to understand and agree with what Mr. Good is saying, but I do not have the math to demonstrate the truth (or falsity, if that should be the case) of the argument. In words, it would be this: the conditions for voting have changed between the first and second ballots.

Specifically, ten games compete not only against all the other games, but also have a 'to the death' competiton with another member of the ten. By changing the conditions under which some but not all of the games are judged, there is an unavoidable bias introduced. A subset of the whole is being judged by different and more stringent standards. Thus the playing field is no longer level. The only question becomes how the two different groups are affected. I believe it is apparent that the result is to lessen the chances of the ten relative to the other games.

I am forced to predict that a lesser percentage of the singled out games will get into tournaments under these conditions. But this again may be misleading, because a closer examination of the ten shows some of them both represented and voted on in the first poll, CC/CRC and FC/FC100, and some that were represented by a single entry which was then split into two, GrandSD/GrandSR, GreatSD/GreatSR, and Mir36/Mir32. This also must skew the results. I believe I must predict that the seven initial entries that became ten will be underrepresented statistically, although one contest does not give an adequate sample.

On the specific question of Mir36 and Mir32, if there were no Mir36 in the tournament, I would put Mir32 in the exact place that I will put Mir36. However, because there are two Mirs in the tournament, one of them must be rated above the other, unless there are provisions for giving 2 different games the exact same rating. I cannot help but believe the lower-rated Mir has lost a little. Am I actually wrong in believing this?

I have added two new sets of similar games. These involve some of my games.
One set is Storm the Ivory Tower and Yáng Qí. They are no more similar than
almost any two Chess variants, but they are both based on Chinese Chess and
were created by the same inventor. When I was voting, I was feeling that
although I would like either game in the tournament, I didn't really feel
the need to have both in the tournament. In particular, I really want Storm
the Ivory Tower in the tournament, and I would want Yáng Qí in only if
Storm the Ivory Tower didn't make it.

The other set is Crazyhouse and Shatranji. Both are similar to but better
than Chessgi, but overall, I think Shatranji is the superior game, because
the less powerful pieces of Shatranj are better suited to a game with drops
than the more powerful pieces of Chess are.

I will also recommend rejoining Great Shatranj and Grand Shatranj into one
group, and adding Modern Shatranj to the Mir Chess group. I would be
interested in having either Great or Grand Shatranj in the tournament but
don't feel much need to include both. I think the chance of one making it
into the tournament is better if they are grouped together. Likewise,
Modern Shatranj seems to have much the same appeal as Mir Chess. But I
will let their inventors weigh in on this before making these changes.

Whether we say Alice Chess is in second or third place is going to make no
difference to how MAM operates, because it does not make any use of
absolute ordinal values. What it compares are individual pairs of
candidates. For each pair of candidates, it notes which one has more
frequently been ranked above the other. From this it creates a list of
majorities. A majority is an ordered pair of candidates, which includes
information on how many voters favored each candidate. If there are no
cycles in the list of majorities, then it establishes the final ranking of
the candidates, and the top ranked candidate is the condorcet winner. When
there are cycles in the list, such as (A, B), (B, C), and (C, A), it
derives a subset of the majorities that are consistent with each other,
and it uses this subset to establish the final ranking. It does this by
sorting them, then it goes down the list affirming each pair that is
consistent with all previously affirmed pairs, and affirming additional
pairs that can be logically derived from sets of previously affirmed
pairs. By this method, it maximizes the number of affirmed majorities,
hence the name of the method. The sorting function compares two majorities
only with each other, and when they are equal to each other in all relevant
ways, it sorts them according to the results of the strict tie-break
ranking, which is previously established by randomly picking ballots until
enough preferences are collected to establish a strict ordering of the
candidates. The order in which the majorities go will make a difference
only when there are cycles in the list. When there are no cycles, all
majorities will be affirmed.

In general, placing Alice Chess below both versions of Mir Chess should no
more hurt Alice Chess's chance than placing it below only one game would.
If both versions of Mir Chess come out ahead of Alice Chess, one will be
thrown out, and everything below it will be pulled up in the rankings.
When there are cycles in the votes, this will have no effect on the
ordering of most of the majorities involving Alice Chess. The main effect
it will have will be on the ordering of the majority involving both Alice
Chess and Mir Chess, and, if yours is the deciding vote, it will make it a
majority for Mir Chess rather than Alice Chess. This will increase the
chances of affirming the majority of Mir Chess over Alice Chess.

You have answered my questions well and in the process given what I
believe
is a good explanation of the basics of the Condorcet method. For this I
thank you. Your explanations and my followups have given me good reason
to
believe that the method is all you say it is. Its flaws you have not
hidden
or minimized; I see them as two. One is the random nature of a tiebreak,
although that is a lesser flaw than many other methods. The other is that
it  gives the least objectionable results which is something different
than most elections, and I can live with this too. So I leave you with
one
comment. 

The method you use is certainly appropriate for its purpose and has many
nice features. I think those who followed this thread will generally
agree
that it is a worthy method and will be happy to see it used in the
future.
Many might wish you had stuck with the original configuration of the
method for this tournament's second round of voting. This is the one
area
where I still have substantial disagreement with your decisions.

My purpose was not to harrass or exasperate you, but to gain what I could
for all involved and to register disapproval of those things I believed
were wrong without creating any personal animosity. By maintaining
anonymity I hope I have achieved the latter. I oppose the creation of
factions and do not like what I have seen of flame wars regardless of
provocation. I hope I have been reasonably cordial generally; enough so
that you did not feel that attacks were directed at you personally rather
than some of your actions and decisions. For those times I have gone
across the line, I apologize. 

While I still disagree with some of your decisions I believe it is time
to
leave and allow the individuals involved or possibly the voting body to
continue this argument if they so desire. You have been an honorable if
unpredictable and somewhat inconsistant and arbitrary opponent but that
is
often the nature of genius. I am certainly not all I wish to be, not that
I
claim to be anything special. Your involvement in the creation and
maintenance of this site is something special. I look forward to meeting
you over a chessboard in the future. Goodbye.

In regard to Switching Chess, if played in the tournament will it be up to
the players (of their specific game)to determine which switching rules apply?  For example:

1. King in Check can switch
2. King in Check Cannot Switch
3. Pinned piece (pinned to King) can switch
4. Pinned piece (pinned to King) cannot switch

Thank you

[Modern Shatranj] is my second choice. Joe took the opposite design course of starting with Shatranj and moving forward. All things considered, I am willing to go along with the proposal by Fergus to group this game with my Mir Chess variants.

[Mir Chess 36] is my third choice - originally an attempt to squeeze most of the Shako pieces onto an 8x8 board.

[Shako] therefore becomes my fourth choice. Judging by the latest poll results, Shako may well be the only one of these four games to make it into the tournament. While I still have some misgivings about playing with short range pieces like elephants on a 10x10 board, I also believe that it would be an interesting game to play.

Hi Jeremy!
 Please call me-I am offline, and your phone # is not working.  I am at 
(813) 654-4165.
  If anyone else has a way to reach Jeremy, please pass this message on to
him. Or if anyone wants to contact me about any of my Variants, leave a
message w/ name and #. If i am home, i will answer, otherwise i will call
back.


    Eric V. Greenwood

  P.S. thanks to my friend Chris for allowing me to get this message out! 
:)