Classicists League Table

[InterMPC]

Unfortunately the change you have suggested is not easy. I can work on it if you want but bear in mind;

Your system won't be 0 sum anymore, It won't be true to the Elo system, There will probably be a snowball effect of unintended consequences that will need adjustments.

Basically in the scenario you've described, player 1 and 2 draw against each other. What happens in chess when 2 differently ranked players draw? The player with the lower ranking goes up and the person with the higher ranking goes down. That is what is being emulated here.

The previous Elo emulation in the table doesn't calculate a score for each player against the others individually, but each player against the average of the participants, therefore sidestepping this issue.

[valent]

Consider using multiple scoring systems:
Original scale = use the current point system for the site
Soloist scale = plus 12 points for every solo win, minus points for every way draw win or survival with the number of points equal to the number of nations in the draw
Never surrender scale = plus one point for every game completed minus 5 points for every surrender
Balance of power scale = plus points for every solo or draw win with the number of points equal to the year of play (e.g., 1911 would be 11) multiplied by the number of winners
Consistency scale = average consistency percentage averaged over games with fading
Italy specialist scale = plus 12 points for every solo win as Italy, plus 6 for every two-way draw as Italy, plus 4 points for every three-way draw as Italy, etc., and minus 3 points for every loss as Italy if eliminated or minus 2 points for every loss as Italy as a survivor.
Russia specialist scale = plus 12 points for every solo win as Russia, plus 6 for every two-way draw as Russia, plus 4 points for every three-way draw as Russia, etc., and minus 3 points for every loss as Russia if eliminated or minus 2 points for every loss as Russia as a survivor.
And others . . .
To address the problem of different metrics, all of these could also be reported as a percentile comparing the player to all other players. Scores would communicate something about the individual’s philosophy, values, and behaviour.

[fellowes]

@Valent

I'm not sure how useful these many rankings would be. Aside from the confusion and work they would bring, would you really try to rank a diplomacy player on how long their games lasted, and how many drew, with higher being better???
It's far simpler to have a single ranking that measures how good you are at winning the game.

Unfortunately the change you have suggested is not easy. I can work on it if you want but bear in mind;

Your system won't be 0 sum anymore, It won't be true to the Elo system, There will probably be a snowball effect of unintended consequences that will need adjustments.

Basically in the scenario you've described, player 1 and 2 draw against each other. What happens in chess when 2 differently ranked players draw? The player with the lower ranking goes up and the person with the higher ranking goes down. That is what is being emulated here.

The previous Elo emulation in the table doesn't calculate a score for each player against the others individually, but each player against the average of the participants, therefore sidestepping this issue.

No. The system still is 0 sum. Each loser loses 1/N (N is number of "winnners") times to each winner. That is all. The winners get the points lost to them. 0 sum.

The reason I'm looking at this differently than chess is that it is a 7-player game, and if you want to draw with someone, it is relatively easy to do so. What is hard is beating someone else. So players 1 and 2 aren't really drawing against one another, they are just beating players 3,4,5,6 and 7.

This sort of thinking, I believe, is reinforced by this passage in Allan Calhammer's (inventor of the game) discussion of objectives (http://www.diplomacy-archive.com/resour ... ctives.htm). He focuses on the fact that members of a 3 way draw beat the 4 other players - not that they scored some half win/half loss as is considered in chess, where a draw must mean exactly that due to the fact there are only 2 playes. A draw is simply a (often far) lesser victory, due to a reduction in vanquished parties, and an increase in the number of parties sharing in that triumph.

Nevertheless, it is not wholly clear why the draw is not an adequate secondary objective, inasmuch as the game is probably a draw with best play from the overwhelming majority of positions actually encountered. One of the difficulties may be that the draw is reputed to be inconclusive, because it is so reckoned in chess or checkers. However, a draw in Diplomacy may be more conclusive than victories among an equivalent number of chess players. If seven players play Diplomacy, and three draw, those three have scored above the four others. If six players play in three chess games, and all the games are wins, those three have scored above only three others, rather than four. Yet in the Diplomacy game, there is still the possibility of one player winning it all.

[echotwo]

The reason I'm looking at this differently than chess is that it is a 7-player game, and if you want to draw with someone, it is relatively easy to do so. What is hard is beating someone else. So players 1 and 2 aren't really drawing against one another, they are just beating players 3,4,5,6 and 7.

fellowes, that's a mistake you keep making. what's hard in diplomacy isn't 'beating someone else', it's beating six other players. 'beating' 3 or 4 or 5 other players isn't hard at all.

[bb82]

The reason I'm looking at this differently than chess is that it is a 7-player game, and if you want to draw with someone, it is relatively easy to do so. What is hard is beating someone else. So players 1 and 2 aren't really drawing against one another, they are just beating players 3,4,5,6 and 7.

fellowes, that's a mistake you keep making. what's hard in diplomacy isn't 'beating someone else', it's beating six other players. 'beating' 3 or 4 or 5 other players isn't hard at all.

And I don't know why he keeps going back to that article. Imo he makes assumptions that that paper simply doesn't imply.

You either beat all 6 or you survived. Thinking that players 1 and 2 beat 3,4,5,6, and 7 is a distortion of the facts. They got a draw, nothing more than that. This game isn't about beating as many as you can.

[InterMPC]

I personally agree with bb82 and echo.
Moreso, IMO if you are much stronger than your opponents, getting a 4 way draw should be considered a failure....
Opinions aside, It would be good to get fellowes' model working as he intended so that we can compare how the results shift with each different model.

fellowes,

I'll try to iterate what I think your method is in my own words and you can tell where I'm going wrong.

There are seven players and 3 drawers,lets call them winners for the sake of argument (is that a bit cheeky?).
So 4 losers each losing 1/N = 1/3 (N = 3 = number of winners)
so thats four people losing 1/3 (0.333) each meaning we are down to -4/3 (-1.333) in the balance sheet. This means we then divide +4/3 (1.333) by the winners. 4/3 divided by 3 = 4/9 (0.444)
So the four losers go down 0.333 each while the winners go up 0.444 each. Net balance = 0.

This is 0 sum, however how does player rankings and the Elo method come into it? or is that excluded from your proposal?
If I have this wrong, can you give me some more hypotheticals, fellowes, so that I can implement it properly?

[valent]

@Valent

I'm not sure how useful these many rankings would be. Aside from the confusion and work they would bring, would you really try to rank a diplomacy player on how long their games lasted, and how many drew, with higher being better???
It's far simpler to have a single ranking that measures how good you are at winning the game.

The zero sum proposals seem to me to be the most meaningful ranking system proposal. I am not taking issue with that, or attempting to replace that, only to suggest that adding more information would be useful.

Would it be hard to also include some "scores" or other useful indices, such as those I described earlier in this thread? Perhaps think of these as indices for interpretation rather than rankings. I suppose I can simply look up a player's archived games and discern for myself how he/she behaves, whether he/she never accepts draws, always accepts draws, never stabs, always stabs, etc. I find the consistency and surrender stats to be a useful pieces of information, and I wonder if other useful pieces of information can be automated. For example, it might be useful to know someone's soloist index, or their balance of power index. Or does knowing these things beg for metagaming? Most of those scores I proposed earlier are not really rankings in the usual sense, but might communicate something useful about the players.

[VGhost]

While I'm 100% behind adopting ELO for our ranking, I just wanted to ask a question. Although I'm hesitant to open the can of worms... are we counting a draw for our purposes as "included in draw" (site rules) or "survived" (original rules)? I favor the latter, but at all events I'd like to have this clarified.

[InterMPC]

In my head i've been working on the assumption that the draw result would be the same as how the site determines it. This way it is easier for mapmaker to cross reference when the table needs to be updated. Otherwise I don't have a preference either way.

I think it's important to get agreement on all of these matters soon, there are games in progress and I'm sure people would prefer to know what they are playing for. It may influence how they play.

[bb82]

No system will be close to perfect, so why make it complicated?

If I was to make a change at the site, it would be the following:

For ranking points I would allow 5 player draws to be worth 1, 6-7 player draws worth nothing, and 2-4 player draws worth 3 points. A solo would be worth 10, with the standard fading echoes of 50 games. AND DIAS would be the only option for a draw, where points are awarded based on a winner or how many survived.

I might also add/decrease points based on the middle ranked person of the game. For someone ranked below the middle ranked person, I would give 1 extra point for 2-4 player draws and 3 extra for a solo. For someone ranked above the middle ranked person, I would take away 2 points for a solo, one for a 2-4 player draw, and give zero points for a 5 player draw.

For example:
Ranks(at end of game) in order (for surrendered nations ranked is based on last person that sat there).
15
19
30
50 (middle rank)
100
300
350

So if '19' got a solo, he would get 8 points. If '100' got a 2-4 draw, he would get 4 points. If '30' won or got a draw, he would be awarded based on the 'normal' system above.