So people can choose whether they want a game to be important or not, something you specifically didn't want in your system ("all games should be equal"), and something I did want in my system ("all games shouldn't be equal"). A neat example of how both systems are valid solutions to slightly different problems

The points system is flawed for several reasons.

Firstly, variable pots. This means that you inevitably get skews, for no apparent good reason. It also encourages players not to care about low pot games, which isn't good.

Secondly, inflation. Inflation is a problem because it means that players who haven't been on the site for so long, or take a break from the site, are at a massive, necessary disadvantage. It says nothing about your skill if you haven't been around so long to rise up the inflation curve, or haven't played often enough.

Thirdly, any above average player's rating will tend to infinity. The result of this is that an important factor is simply playing a lot of games, but that isn't something that you want really to encourage or something that reflects on skill.

Fourthly, the risk:reward ratio always remains the same. This is the really big one. Basically, to play Madmarx, TMG or Rait, I must put up the same stake as to play simple Simon. In return for winning I get the same reward. The effect of this is simple- you do better under points if you refuse to play good players. Any good rating system should give the same rating for a player no matter who he plays.

"So people can choose whether they want a game to be important or not, something you specifically didn't want in your system ("all games should be equal"), and something I did want in my system ("all games shouldn't be equal"). A neat example of how both systems are valid solutions to slightly different problems"

I take from people wanting games to be unimportant that they don't want to have to care about the result of them, i.e. they don't want to play to the best of their ability (if they do, why wouldn't they have it rated)

I'd recommend having a simple dichotomy, rated and unrated games. This way, you can have an unimportant game, and you can have an important game.

The only criticism of G-Rating with respect to implementation that I accept is that you can have rating lower than your starting rating, but its not my opinion that matters here, because I'm not running the site. I don't think that that problem is insurmountable.

I think if someone has played on this site for two years and has managed to accumulate over 10000 , that they're more than just a simple point farmer. It's not uncommon for somebody to have a good run and win or draw a bunch of games in a row, which will result in more points, but to consistently perform at that level over an extended period of time certainly gives some indicator to that players skill, does it not?

I think points system and ghost rankings are equally valid. This is my defence of Ghost's critique of the points system.

"Firstly, variable pots. This means that you inevitably get skews, for no apparent good reason. It also encourages players not to care about low pot games, which isn't good." Response, I want some games to be competitive and some to be fun, I like playing variants such as gunboat, but may want to risk less in these games. Some of the 5 pointers I play I take extremely seriously (league matches, for example) but others, like haiku poems I take less seriously. The game is meant to be fun, it isn't all cut-throat.

"Secondly, inflation. Inflation is a problem because it means that players who haven't been on the site for so long, or take a break from the site, are at a massive, necessary disadvantage. It says nothing about your skill if you haven't been around so long to rise up the inflation curve, or haven't played often enough." Response - the points adequately reflect experience as well as skill, high points - high experience, low points low experience. Skill levels tend to increase with more experience and this is reflected in the points system.

"Thirdly, any above average player's rating will tend to infinity. The result of this is that an important factor is simply playing a lot of games, but that isn't something that you want really to encourage or something that reflects on skill." Response - again points reflect experience, players can easily determine who is experienced or not.

"Fourthly, the risk:reward ratio always remains the same. This is the really big one. Basically, to play Madmarx, TMG or Rait, I must put up the same stake as to play simple Simon. In return for winning I get the same reward. The effect of this is simple- you do better under points if you refuse to play good players. Any good rating system should give the same rating for a player no matter who he plays" Response - your argument is only really valid if simple simon is able to steal points. As simple simon would have needed some skill or longevity to acquire the points necessary to compete in the high pot games, beating him should be rewarded the same as beating MadMarx.

Wow...I hadn't thought of having rated vs. unrated games. I get into some public press, gunboat and some smaller pot games with a bunch of new players that I like to play for fun....but always with some angst, because I know they are going to affect my ghostrating. Recently, I have put less importance on my Ghost rating because of that issue. But adding a rated vs. unrated option, would definitely restore some of my confidence in that system.

Once again, brilliant thoughts from TGM. Also, thank you for developing the system and making this the top Diplomacy on the web.

Maniac makes some good points, but I can also refer you to my old post on the dev forum which you never responded to, which responded to all of these issues you raise:
http://forum.webdiplomacy.net/viewtopic.php?f=23&t=47&start=0

Remember "Expected Result= Old formula* 3/4 + (1/7)*(1/4)"? Remember "If people join and don't play 10 games, they might as well not exist" ?

"From the average performances in a first game, a new player’s level is 6/10ths of the average rating. It would therefore not make sense to have new players with a rating of 100, since that would simply inflate the ratings of those who play against them. Thus, a new player should have a rating of 60."
So new players join at 6/10ths the average rating, and for each new game for their first 10 games they get 4 as a given to get them up to the average, going by the formula "Expected Result= Old formula* 3/4 + (1/7)*(1/4)" ?
So I guess if 10 new players join the first has a rating of 60, the average rating is now 60 so the second has a starting rating of 6/10*60, the third now has an average of (60+6/10*60)/2, so he starts at (60+6/10*60)/2*6/10, and so on down to the 10th? What absolute absurdity, how completely arbitrary.

You say the problems are not insurmountable, but you sure haven't said how to surmount them. And that's what bothers me: You have not yet produced a workable alternative. How can you possibly say my system is flawed when you have produced no alternative?! It is infuriating.

If the system has changed since you made that post: What has it changed to? Where can I get an exact specification? If I show a whole bunch of holes in it as I did last time will you ignore it again and come back in a few weeks with something different?

I know it's easier to attack my system with the same tired arguments than it is to defend your own, but if you actually want to replace the existing system your system needs to stand up to scrutiny

And I'm also happy for you to apply your system on top of the existing one, if people value their Ghost-rating more highly (as with Crazyter, who agrees it's obviously better but just doesn't know what it is): But how on earth do you expect me to respond to criticism of my system when your system cannot be implemented as a replacement?!

I thought "both systems are valid" was a good compromise, we could use my practical system and you could come up with your rating seperately, but if you're actually challenging my system as in this thread you need to provide a specification and actually respond to the faults in it, rather than endlessly attack my system with the same arguments.

Kestas- I chose my examples to have extreme betting patterns to help highlight the problems. In most cases the problems wouldn't be so ridiculous, but they would still be there, just to a lesser extent. But really, I don't think betting like this is all that weird. I for one play two types of bet games, 5 point games that are either invitationals or leagues or gfdt or masters. My other games are really high pot games because to get the same skill level without invites I need to keep the bar to get in sufficiently high.

I play all my games roughly equally. I want to win. I try to win all my games. I don't take games less seriously depending on the variance of the pot.

Spell of Wheels point about gunboats is a good one. Gunboats are fun. But I feel like its just handicapping myself because I can't talk. So I don't play them. Just betting 5 to get in doesn't make a difference to me, because the defeat will show up on my record regardless, and no one takes points seriously.

So to conclude, I get it that you think that games should be unequal, and I think cool variants could count under that system, but unrated/rated would solve that problem so much better than points.

I actually think it is a good compromise Kestas. Because I use the points system to help keep the riff raff out of higher games, and it certainly helps the intended purpose of limiting newbie games at the start, and interesting people in acquiring points for when they join the site.

However, I don't think anyone sees it as a rating system anymore. Experience sure. Rating system? No.

Hmm, I take points just as seriously as the ghost rating. The current points system is fine if you ask me. Points measure experience, remember....

And I myself can't respond to technical problems that I'm sure exist in the G-Ratings. I just thought I'd lend my experience with the points system to the discussion. I hope its taken in the way it was intended.

Centurian neither me nor you can evaluate whether it would solve the problem better without a full specification, which TGM has not yet provided. And if he has provided it I sure hope it's different to the original one he proposed on the developer forum, and I sure hope if I show problems with the new specification he won't ignore it and bring out a new one as with last time.

Again I have *no* problem with this if TGM wants to provide a rating system of his own, but if he is saying the Ghost-rating should replace the points system, as his criticism of the points system here suggests, he *needs* to post a full, unchanging specification of it, and see if it really hold up against criticism better than the points system. (It sure didn't last time)
I cannot even begin to critique let alone implement an alternative system without a specification which stands up to some criticism. Until that point TGM is throwing bricks from a glass house

(Bear in mind I wouldn't be this argumentative if this hadn't come up before repeatedly, this is an old, old issue which has never been resolved yet keeps coming up.)

Thats what you get with a Ghost Rating thread
Die issue... Die!!!!!>=[

"I know it's easier to attack my system with the same tired arguments than it is to defend your own, but if you actually want to replace the existing system your system needs to stand up to scrutiny"

The simple answer is that I don't care about replacing the existing system.

Apart from the aspect of recreating accounts after a bad start, I think mine would be better, indeed I am convinced it would be. But I don't mind the status quo, for myself. I have Ghost-Rating to measure players' skills to a good enough degree, although it isn't perfectly implemented; I have tournament systems to guarantee the standards of commitment. I have, by a deliberate and coherent plan, made a position for myself, through my own work, where the points are irrelevant to me.

If you want my explanation of why Ghost Rating would be better for promoting responsible play it is this:

The aim of a rating or points system should to promote best quality play, and discourage quantity-style play; the points system was introduced because players didn't care to play to a standard if they were loosing a game. Therefore the best reward system is to reward playing well and penalise playing badly. This means that the best reward system to use is a system that accurately measures a players skill, because then they are encouraged only to play well, which is what we want. Ghost-Rating is the best system we have to measure player skill. Therefore, Ghost-Rating is the best to encourage the right sort of play.

That means that Ghost-rating ensure that with respect to the actual game, Ghost-Rating is the best system. You can criticise it because it will encourage people to stop playing, or open a new account, but there are no other grounds for criticism that can even be considered until you demonstrate the fallacy of my above argument.

I can respond to the idea that people will stop playing by the fact that, with the system already respected by a number of top players, nobody has quit to just stay at the top. The latter, is, I accept, impossible to answer without giving it a try in full implementation. You don't want to do that, so that is it. We could have "not yet rated" for 8 games or something, to tackle the problem before it comes about too, if you wish.

Wow... I really need to start reading the entire threads I post on :)

As a user if the website, I would advocate for dual rating system as I see the advantages and disadvantages of both as a player. I liked the idea of putting some graphic to easily identify the "skill" level of the individual player. I think that symbol could be based on the GR. I don't understand the math involved or the programming required to base or incorporate these rating...I can only give an enduser opinion. Perhaps a minimum number of games is need to acquire to GR baseline number.

Great information, Kestas. Thanks for making this the best Diplomacy site on the web.

>you demonstrate the fallacy of my above argument.
What argument? You haven't defined your system. Define it, I'll rip it to shreds, you ignore it. It'll be fun

Thanks for the kind words SoW, I will honestly try to find a compromise that exists, as long as it doesn't involve putting up with too much nonsense ;-P

Also...one more comment. I really think we need to consider the GR effect on the entire community. When I first came here, I could play with some top rated people and really learn alot quickly from them. It's hard to find any top rated people in low stakes, low GR games anymore. That's my primary reason for advocating for a rated vs unrated game.

Is the reason top rated people don't appear in low stakes, low GR games because the game is rated?

My understanding is that every game I play affects my GR.

Well I try to stay away from low stakes games because the play is worse and its easier to say no to those games than a high ranked invite. But yeah, I think GR is a factor in not playing in them, just not the main one. I think your likelihood of doing well mostly negates that, but sometimes bad players won't listen to reason like good players will, so you can just get unlucky.

Also my understanding is that I put up more risk to my rating by playing lower rated people than if I play similarly ranked players. Ideally, I would only want to play only players who are higher rated than me. I have less to lose.

The idea is that you have less to lose if you play higher ranked players and more to win. In theory this is balanced out by the fact that you are more likely to lose and less likely to win against such players.

Centurian...remember the public press game with you, me, TGM, statagos and MJT123. One of my first games and very memorable. I just finished up another memorable PP game. I was by FAR the highest ranked player there. I drew that game while I was ahead....just like I did for you and TGM. Some games should be played for fun, help a new guy...whatever.... Without having to worry about my GR.

Having said that, I value my GR. It seems like an appropriate ranking. I just hate that it follows me everywhere I go.

I even recently used TGM's failed strategy of raising the "Juggernaut" alarm with the same failed outcome ;)

Here is a question that needs answering by people who engage in ranking systems....
"If you were the world's 11th best Poker Player would you
(a) only play with people from the top ten,
(b) never play with people from the top ten
(c) only play very high stakes
(d) sometimes play for matches with the kids around a kitchen table.

(Choose all that apply)

In my opinion ranking systems invariably highly rate the kind of play that the ranker prefers and tries to lowly rank other kinds of play. Surely having two systems is fine if they reflect different people's interpretations of how the game 'should' be played. But because no one can (should) be dictatorial about how the game is played a perfect ranking system is impossible.

Well said Maniac....can I have my matches back? ;p