I like the Elo-rating system. But I'm open for suggestions.
(http://http://en.wikipedia.org/wiki/Elo_rating_system)

I'll fiddle with it this evening after work and see what I can come up with. Always happy to help :)

And I really prefer an automated solution. Taking all games of one month and recalculate a ladder does not sound too reasonable to maintain. (Like the Ghost-rating over at webdip).

It would need to take into account draws as well... Draws should count for a lot more in diplomacy than in Chess. Also, different variants would have to be worth more... A solo in a Haven game is worth a lot more than a solo in a 1v1, for example.

Elo is good for 1v1 games...

not sure how, but you'd need to modify it for a 35 player game drawn by 10 people with a wide range of SC's, like 80 to 2.

I would probably recommend the Ghostratings from Webdip, since they already have the nessacary spreadsheets and calculations for diplomacy it would probably be the easiest to impliment.

+1 Captainmeme

There are many ways to use elo for multiplayer games too.
http://www.tomshardware.co.uk/forum/75151-25-realize-multiplayer-game-anyway#t491297
For example.

@fasces. It looks like ghost-ratings have to be calculated by hand each month.

On first glance (a very brief first glance), I think the primary problem other than actually making the basic math work is forcing the equation not to alter ratings at the end of a game when the player is in other games. For instance, a player with a rating of 500 plays in games A, B, and C. He finishes game B, moving his rating up to 525. I believe that the rating should stay the same for the other two games in the calculations, but I think that any computer would change the ratings to 525 in the other games.

I wonder if they use Elo on Chess.com, because they certainly have to deal with that over there, and I think they found a way to get around it, which I am sure can be developed for our many multi-player variants.

Are there any other games that have a worldwide system but multiple players per game? Bridge has this weird system that I doubt would work with us. It seems like all the biggest games are two-player affairs.

Oli, check this out. You can have multiple players (I tried 7).

http://elo.divergentinformatics.com/

Not sure how the exact mechanics work of course, but certainly the tables will be needed. But at least for a single match, one could learn the desired outcome.

It would be necessary to take into consideration maps such as Rinascimento and the new Imperial 2 map where some powers start off greatly disadvantaged compared to the leaders (Pisa/Siena etc on Rinascimento and basically any 3 sc power on Imperial).

How much does it affect in play though? I ask because one would think that by joining the game you accept the risk of getting a 3 SC country (using the imperial variant for the moment). If we are going to take that into account then only players who have reached a certain rank should play it, therefore meaning that you know a "better player" (for lack of a better phrase) is playing the country, therefore meaning it is sort of not completely anon.

Making a special circumstance for them only encourages ass holes (pardon the french) to play them to get the extra boost.

@Drano, I wasn't meaning any of this in a "call you out" fashion. Simply giving my opinion on what you are saying. :)

The Ghost Rating system is hand calculated, but it doesn't have to be. Essentially, it is a system where each "bet" is a fixed percentage of the players current rating, and can be in fractions. The "bet" is not selected by the players. It could operate on the fly, automatically work for different variants, ppsc & wta, and account for draws.

Basically it is this:

Instead of the game creator dictating how many points for the game, the skill rating system takes 1/22.5 of the players skill-points for each game. This makes it more "expensive" for a skilled player to play, than a noob.

So, if a noob (with 100 skill-points) wants to join a game, it would cost him 100/22.5= 4.44444 skill-points to play. If a skilled player (with 1000 skill-points) wants to join the same game, it would cost 1000/22.5=44.44444 skill-points.

Once the game starts, there is a pool of skill-points which would be distributed the same as the current point system works today. Draws would be split equally. Wta games go to the winner, ppsc games get distributed according to ppsc rules. (I am not sure how replacements are factored in.)

The net effect is that more skilled players will gain very little from defeating noobs, but noobs will gain a lot from defeating a skilled player.

This system would be able to work in parallel with the current point system, and would have a different purpose. The current system limits the number of games because a player can run out of points, but does a poor job at measuring skill, but the skill-point system (more) accurately measures skill, but doesn't limit the number of games, since you can never run out of points.

P.S. There is a long mathematical explanation why 1/22.5 factor was used. But basically it prevents wild swings in skill-points, and allows a reasonable opportunity for noobs to gain skill points. A lower number might cause radical drops in skill-points for skilled players, and a higher number might cause too slow of an increase for noobs. My explanation is a very simplified description of their mathematical explanation.

drano raises a good question, and I can already hear the follow on question, that all of our variants on this site have various balance issues and drawing Russia in WWIV is much more advantageous than everyone's favorite runt, Egypt.

One obvious 'solution' to this balance question (but not necessarily a 'good' one) is to normalize the resultant score results for players involved in the game by the stats rating of the country they played.

The downside is that without recalculating everyone's score over all past games each time the stats rating for a country changes, is that the 'difficulty' of a country would then be time varying and a player's performance and resultant score change would be a function of when they played the game. Not sure whether this would be a real, noticeable problem, but it just feels funny to me.

Thoughts?

@imagonnalose,

That calculator is very similar to what I just described. Try it with one player with 100 elo playing against a player with 1000 elo. When the 1000 elo player wins, he gains very little. But if the 100 elo player wins, he gains a lot.

Leif & Drano,

There is no way to balance diplomacy. Even on the standard map, some powers are more advantaged than others. The only maps that are balanced are symmetrical maps like Duo, lepanto, etc. And those maps tend to be less interesting than other maps.

Imbalance is part of the game. There is no way around it.

@Devonian: Sounds like the GhostRating is an Elo-system with a K-value of 22.5

Oli,
I read the Wiki description of the elo system. I think it is the same system as the Ghost rating system. The descriptions of both systems are very similar.

Yes, I think the difference is the k-value.

Here is a quote from the elo wiki page that corroborates my opinion that this system can be used on an ongoing basis, not just monthly (like the ghost rating on web dip)

Found under: "Theory - Mathematical details":
"This update can be performed after each game or each tournament, or after any suitable rating period"

imagonnalose and Devonian -

I am well aware that no game is truly "balanced' except those symmetrical maps Devonian mentioned. However, I am referring to maps that are SKEWED, not imbalanced. Does anyone REALLY argue that soloing as Venice is way way way easier than as France? Or pisa? or siena? Any ranking system that considers a solo as Venice is equal to a solo as Pisa is a ranking system that's flawed. Some maps stats show that things just are NOT equal, or even close. Perhaps these maps should just be not counted in the rankings? I know webdip does that with the Ancient Med variant.

Or we could use the statistics we already have about the variants to adjust.
So this would only change the outcome of some new variants, but with no strategies and all new to everyone it might not that important in the first few games...

Drano,

I havent played those maps. But if they are as skewed as you say they are, maybe they should be excluded.

@drano - the stats are misleading, though. The stats are based on mostly very small sample sizes. Furthermore, quality of play is way varied between different players. So, it is hard to know what is really unbalanced and what just appears unbalanced.

Jimbo, I agree.

I think that the skills of the players that played the country has a big impact on small sample sizes.

I suggest that the new ranking system be implemented as is, then after it is up and running we can then start to look at tweaking any anomalies that may pop up due to variants like Fubar and Rinascimento.

*also implementing a running Ghost Ranking would make the guys over @ webDip green with envy :)