it appears Gobbledydook is a genius

I can argue against the system after Gobbledydook's modification, because of the inherent problems with the original system.

The reason I suggest a rating system is because the point system, although perhaps entertaining, fails in distinguishing between the players. Ratings allow players to actually be compared, so that we know that a rating 2000 player is very strong, a 1600 player average and a 1400 player weak.

The Elo system could be adapted very effectively to Dilpomacy, and the player need not understand its mechanism, because the concepts it upholds are simple, common sense, and what matters.

The formulae may make your head hurt to look at, but the outcome is a simple number, which means that players are comparable properly, and we have achieved the objective of the system.

Of course, if nobody wants this and would rather have a system that allows players to bet, then this isn't up to the job, or we could have the two in parallel, but I assure you that the Elo system is better than the one we have now could ever be. I stress though that I do not criticise Kestas for using it, since I think he got it right in making a simple system and the focusing on the game itself. I just want to help get over this hurdle, by doing something I am capable of.

Well, the idea is that eventually points will relatively indicate the strengths of players. Stronger players will undoubtedly rise in points, while everyone in between will fall into some relative equilibrium (obviously increasing because of inflation). The only real problem is that if ever an incredible Diplomacy prodigy were to emerge, which nobody could defeat, then their points value might start running away from any relative ranking, into an absolute domination.

Gobble's solution is the one that I would advocate, it's simple, and would work.

In the short term, Gobble's fix would sort out inflation to an extent, however I would hope for more. Essentially, I would like to see ratings be a reliable measure of a player's skill. It remains that if I beat six people in a 5 pot game, and then loose in a 500 pot game to the same six, I end up backwards in spite of being the second best player there at worst on the basis of those two results.

I agree with TheGhostmaker -- the Elo-like rating system would be the best to measure players' strength.

Lol I still like my formula better. Maybe if I gave it a fancy acronym as well.... Ok, let's try this:

CRAIG - Common Rating Awarding Intuitive Gain


I have way too much time on my hands....

no, elo has turned every possible game into crap. believe me, it actually promotes not playing the game! the less you play, the better you are ranked

our system makes you play games, because if you don't you will be eaten away by inflation. i believe this is a good thing

also, elo promotes playing only with the equally ranked players, while we are here constantly challenged both by novices and experienced

Look at my formula and try plugging some numbers into it. I think it strikes a very rational balance between win percentage and number of games played. I.e. one example premise of it would be that it's easier to get a 100% win percentage after only 1 game than it is to get a 70% win percentage after 50 games. But then when you get into the higher numbers, like 600 games played vs 650, the difference isn't as great. The exponential variation is the key.

And yah I just took a look at ELO, and I agree it does look like total crap.

Here it is in a cleaner, more simplified form:

Rg = Wg ^ (Wg / Lg)

Where Wg is total number of games won, Lg is total number of games lost, and Rg is the invisible result number used to determine the player's ranking relative to the other players (sorted highest to lowest).

Try plugging in some random values for Wg and Lg, then compare them and see if you find any examples where it spits out a relative ranking you think is unfair or irrational. I dare ya! :P

Err only flaw of course is if Lg is 0 lol. Ok, so let's revise it like so:

Rg = Wg ^ (Wg / (Lg + 1))


Ok, now try it.

I'm quite in favor of CRAIG, as you are penalized for losing games rather than simply only being rewarding for playing them. The only oddity about CRAIG is that everyone who has won 1 game, no matter how many game played, will have a CRAIG score of 1. So, if I play 200 games and only win 1, a player who wins his first game will be my 'equal'. This can be adjusted by doing this:

Rg = (Wg + Wp)^ (Wg / (Lg + 1))

Although we should use this for the rankings (and should a little symbol awarded for different rank), I still believe that points system should be used to regulate what games may be entered.

If you want to just fix inflation by not implementing a new system: take away a point from every person on the top 100 every time a new person joins or points are reset.

Also 0^0 will produce an error in a computer evaluation, simply don't run the formula if someone has won no games and award them a position based on the number of games they have played (so that players who play more games are lower)

Revised:

Rg = (Wg + Wp) ^ (Wg / Lg)

45 replies, 45 ideas for new ranking systems (probably), 0 problems with the current ranking system that are based on more than opinion and require a new system (the 100 points minimum bug is known and doesn't require a new system)

No ranking system can sum up a players worth in a number. The current system for all its simplicity, and other benefits such as moderating the number of games joined, does as good a job as is needed :-|
If you think your formula really can express how good a player is as a number you're kidding yourself. Other phpDip forks that use different rating systems still get these suggestion threads; it's a problem with a solution that varies depending on opinion. No matter if you add taxes, points equity and hedge funds, require new players be given bonds with a friend's points used as collateral, have points backed against visits/day or gold in the US treasury, whatever; it'll still be unfair and all these modifications just make things more complicated.

Inflation: If someone makes it to the top of the ladder and leaves, should they stay there indefinitely? No, of course not. Inflation solves this problem; it's a clear outcome of a constant flow of points into the system and not out, and was seen as a positive advantage of the system :-|

I think I need to add a points system complaint question to the FAQ & points page, we've had this discussion before :-(

"Inflation: If someone makes it to the top of the ladder and leaves, should they stay there indefinitely? No, of course not." Why not? If this person was/is so good that they got so high up in the rankings that no one else can reach them, how does leaving make him a worse player? If the player is no worse (and presumably the other players around him are no better) then why should he drop down in the rankings below these other, lesser, players. I understand why you would want to have a constantly changing ladder, but as far as rankings go, it sort of defeats the purpose. In my own, very humble opinion.

Surely it is something of value to be able to see that Rait was the player of his day, should he not return?

Comparison between people who have relative inactivity and those who are playing 24/7 is the prime purpose of avoiding point inflation.

To questions/comments about Elo:

1. It is actually not an acronym but the name of the theoretical physicist who invented it.
2. Elo does not favour not playing unless it is used in a bad form that causes deflation. Your rating will remain the same in value forever if you don't play.
3. In Elo you are rewarded in proportion to the strength of players who you play. It is a symptom of the fact that I get as many points beating Rait as I do beating a newbie that causes uneven matchups, for the same bet, so playing Rait is plain stupid.
4. In Elo people generally play the same rate because it means a fair game. This is particularly important in a negotiation game where people might catch fright from the rate causing unintentional meta-gaming.

In Chess, Elo is used very effectively. In diplomacy, any rating is more circumspect, but the theme remains the same. I suggest it because it has been proven to work very well. It was not dreamt up by a layman but by a man who knew alot about statistics, and that is why I have recommended it.

I might program a simulation or two to show how powerful it can be.

I still stand with Kristopher on this one.

Although, CRAIG doesn't sum up a player's ability in a number, it is an arbitrary way of ranking relative performance.

What you could do it take away the number after people's names (that would reduce the ego-trip of having lots of points). Then the system would be truly only a way of differentiating ability in games.

> Your rating will remain the same in value forever if you don't play.

yes, that's why it promotes reaching high rank, and then making another account :-) i have seen it too many times

> In Elo you are rewarded in proportion to the strength of players who you play.

this just does not work with diplomacy. even if you are Rait, there's a probability of losing even when playing with noobs. diplomacy is not deterministic, unlike chess, where you can never beat a higher ranked player.

it would rather promote higher ranked players avoiding to play with noobs. the reward is small, and the loss could be great

> I might program a simulation or two to show how powerful it can be.

ask Kestas to give you access to the database, then calculate elo ratings for each of us. it might be fun to see the numbers!

Point 1:

Rating decay could be used for players who are inactive for a long time. Multi-accounting is not allowed, so this shouldn't happen. In other games it is allowed.

Point 2:

I understand that it isn't deterministic, but all that means is that the normal curve is flatter. At present we cannot tell the difference between the noob and rait, because of the relatively poor system.

Point 3:

That would be good to do, if I can work out how to program it. My knowledge is very basic, but we can see how it goes. eg. in Excel I should be able to do it.

and other than that I only know a bit of vb.

Ghost, in reply to your comment about not wanting to play Rait because the reward would be the same as beating a total noob, this problem is solved by the current system, because Rait has many many more points to bet than any noob.

could you define more precisely what faults do you find in the current system? about the difference between rait and a noob, it's easy because rait has 8000+ points and noob only 100

so? Rait played many 100 point games...

And with inflation, more and more people will have more and more points to bet.

yes, but you said we cannot tell the difference. what do you mean by that?