But where do you draw the line? Think of the difference between 1 day and 10 days. Sure, maybe having more time won't improve your moves but a lot of people don't log in more than once or twice in a 1 day phase, and communication is harder than a 10 day one.

The distinction should be between live and non-live games. There's a difference. A lot more diplomacy goes on in the non-live games.
It would also be nice to have non-rated games.

.... I didn't answer your question dD. We don't want it to be too complicated, but "where do you draw the line?" I think live versus non-live is pretty intuitive.

At some point, you run out of things to say because not enough information is available to you.

Do you run out of things to say after 1 day? 2 days? 10 days?

I don't know, but I think the difference in the amount of things you can say between 15min and 24 hours is greater than 24 and 48 hours.

Also, think of it this way: In a live game, you play the game in one sitting. By the end of the game, you may be tired, or have somewhere to go. Non-live games are not played in one sitting, be it 1 or 10 day phases.

I rather like the present system using both metrics. If your point total/ranking >> your Gr rating you have been picking off easy oponents, while if your point total/ranking << your Gr ranking you have been playing with much higher ranked oponents and been content to settle for a large # of draws. If your points toal/ranking approximately equals that of your GR ranking you have got the balance about right. Couple this with your win/loss ratio, which is what the game is all about IMO, and you will have a pretty good idea on your development as a player.
No doubt I have got all of this wrong and am open to correction.

Just a question, what's the longest phase length that is considered live. (ie; how will dodgy phase lengths like 1 or 2 hours be dealt with?)

But anyway, what I want to say is the effect a phase length has on a game is rather arbitrary and really depends on the circumstances. (people not being able to play through the whole game due to planning incompetence happens in both live and longer games.) The difference between diplomacy and the variants/gunboat are much more evident and solid.

I don't think it's feasible to play 1 hour phase length. But kudos to anyone who can pull it off. Well, not kudos, but my deepest sympathy.

@Ful - Who are you to say that "balance" of GR and points is right? I *lie* playing opponents tougher than me and don't give a fuck about points, so what is right for me is to have a top 100 GR but crappy points. For you to say I'm doing it wrong because I don't have them "in balance" is bullshit, man.

OK, I didn't read everything, so sorry if I am repeating anyone.

First, the point system isn't going anywhere. Kestas likes it and the reasons are that it is a simple system that beginners can grasp.
Yes, it can easily be 'gamed' and doesn't really show the true value of the player AS a player, but it helps make the site run.

As far as GR, that is a script that is run once a month and I don't think it is anything simple. For that to be run all the time when everyone logs in would be a royal pain. To run it once a month would be unfair. Plus, how can you prevent newbies from joining games? Say a max of 100 GR can join? We've all seen games here where we can't get a game with just under 100GR.... so should we be able to 'up' the level before the game starts/cancels?

You are asking to change the fundamentals of the entire system. It isn't a simple change even if it was desired.

Now, if you ask me, I think that GR should be integrated (updated every night at 12 or something), but it should be a secondary ranking, not a replacement. In fact, I think it would be an interesting idea to have multiple types of ranking systems integrated.

@ Draug, that's fine and Iam sorry if you felyt I attacked you by my remark, it was really not intended.

Obviously I fully agree with this from FK - ''I think it would be an interesting idea to have multiple types of ranking systems integrated''. Together of course with that win/lose ratio

@fortknox - GR doesn't need to be updated at every login to be accurate. If properly integrated and run on just the players in the game, with the number maintained in the database, it can be run on the 5-17 players at the completion of the game when points are doled out. It only takes so long monthly because it is run on everyone. Adjusting it each game when the game ends would be very fast and very easy to implement as it's just a serious of 5-17 equations per game.

series, net serious... Where's my brain. Oh yeah! Waiting for vacation to start in 6 hours!

not, not net... Geez!

Oh, and it could recognize game types and not run when a non-GR rated game type (or a non-rated game) ends.

I was going to work on my live games tool over the holiday, but perhaps I should look into incorporating GR in the code, if Kestas will let me. Problem is, my PHP is pretty much non-existent.

Draug, read my comment entirely. Updating it nightly is probably the best option. Remember, the win/loss of someone else slightly affects my GR if I've played them recently.

I suppose that this all bois down to two things - 1) how highly do you value a win over a draw and 2) how highly do you value the callibre of the opposition. I suppose the answer to both questions is a purely personal one.

No, FK. It doesn't The way it is computed is one game at a time as it is. Someone else's win does not affect you unless you were in the game. Their GR is what it is when the game ends and that is what affects you. To run the entire sight nightly would be excessive and much less efficient than to run the changeset immediately upon the end of the game. I've talked to Ghost about this exact issue in the past and he confirme d(and can confirm now or clarfiy if things have changed) that each game is run with the GR at that moment and that the routine looks at the prvious months run when doing the differentials but those differentials have nothing to do with the actual GR computation, just the poeple in that game and their GR at the moment the game completed.

Now, if you run it nightly and only run it on games completed in tyhe last 24 hours, that works, but why not just do it the moment the game completes on the players in the game. It is a very fast and efficient way to do it without adding a nightly process to the schedule or having to deal with the crossover games where the last turn ran before midnight UTC and the game drew after midnight UTC so the system accidently misses it.

hrm... I thought he put in a 'strength of schedule' type thing where he constantly updated your weight based on the people you played, but I guess that'd get overly complex.

I thought there was some chicken blood, full moon and rolling of the bones that took place. It did explain the once a month thing.

it's goat's blood czech... and you forgot the tears of young children as well...

SpeakertoAliens "However, variations (Gunboat, Ancient Med, etc) count as less valuable in GR than pure games do, so if you play a lot of Gunboat your GR will be less than it would have been if you'd stuck to pure WTA/PPSC. The standard points system doesn't care"

Not actually true if you play enough games. Your rating will move slower, but you do not have a disadvantage in the sense that getting higher requires better skill- it just requires more evidence of that skill level. Nevertheless, this doesn't have to be the way things are, and its just a matter of my own preference for GR

Abgemacht “Both would only could WTA, standard map, full press. These are the 2 closest variants to the boardgame. It think it's silly to lump other variants into the GR. Gunboat is not Diplomay; it is a different game. Why would you have your checker wins count towards your chess score? It makes no sense.”

The reason other things are included is to try to make the system more inclusive, for me GR has always been there because points aren’t good enough and I wanted something better. It isn’t completely unjustified because (I should imagine) there is a reasonable correlation between skill in each version of the game.

Spyman “Certainly there needs be different types of GR, as you you have just mentioned Abe”

I’m not sure that this is any more necessary that it is in points, after all, points is GR where you can choose how important a game is and the expected values are all the same.

What I would like to see would be (as you mentioned) rated vs unrated.

I have given some thought into developing a Bayesian system (similar to Trueskill, for instance) which would sort out these issues for good by properly dealing with the idea of correlation between the games.

Fulhamish “I rather like the present system using both metrics. If your point total/ranking >> your Gr rating you have been picking off easy oponents, while if your point total/ranking << your Gr ranking you have been playing with much higher ranked oponents and been content to settle for a large # of draws. If your points toal/ranking approximately equals that of your GR ranking you have got the balance about right. Couple this with your win/loss ratio, which is what the game is all about IMO, and you will have a pretty good idea on your development as a player.”

How I view it is that there is no right balance, but the rating system/points system shouldn’t penalise you for whatever choice you make. Ghost-Rating, whilst it doesn’t perfectly meet this, comes far closer, which is why I think it better.

Fortknox “First, the point system isn't going anywhere. Kestas likes it and the reasons are that it is a simple system that beginners can grasp... Now, if you ask me, I think that GR should be integrated (updated every night at 12 or something), but it should be a secondary ranking, not a replacement. In fact, I think it would be an interesting idea to have multiple types of ranking systems integrated.”

This is very possible. Kestas and I are both much less fight-y over the issue of points and GR at the start (I’m less obnoxious). I will include a post of the last thing Kestas has said on the matter (a little over a year ago).

Fortknox “As far as GR, that is a script that is run once a month and I don't think it is anything simple.”

With the current script we are looking at about 5-10 minutes to calculate the entire history of webdiplomacy, so it doesn’t take much processing power. Implementation would change the script slightly to improve it (deal with CDs & peoples’ ratings changing during a game etc.), but that would be only a handful of operations per phase.

What Kestas said November 2010 (all in his replies, I've redacted for brevity):

"I've moved back and forth on this quite a bit, as has TGM and the general community, but here's where I think things stand at the moment:
- The points system is pretty effective for what it does, it's efficient, simple, and serves multiple purposes
- GR is pretty effective for what it does, apparently some people place more weight on that which is fine, and it still correlates with points to a reasonable degree so it would be a refinement and not something entirely different...

Re implementing it within the site:
- I would like to do this at some point so we can host tournaments on a more official, well supported basis, but apparently the way GRs are calculated is based on a spreadsheet that would be difficult to code(?) There has basically been a back and forth over whether and how GR could be officially implemented, and how, and whether it should be
- GR isn't perfect either, and I think it gets a lot of consideration because it's the alternative to the points system (which also isn't perfect).
The points system is simple and open enough (as well as being the default) to be, I think, more open/subject to criticism than GR. I'm convinced if GR was the default system it wouldn't put an end to these debates, because it's a subjective thing. For example the recent thread where someone said they wanted to be banned because they had a negative GR; if it really was a no-brainer we would have switched, but it's a lot more complicated than that"

"I've definitely shifted my position as GR has got more accepted and stable, and carved out an independent niche for itself, yet there are still many disagreements on whether it could replace points which can confuse things" (This is important to note I think, because in that thread, there two debates as always, should GR be implemented and should GR replace points, the first I think has a strong majority in favour of GR implementation, but people are far more tentative about GR replacing points)

"The reason I'm interested in also implementing GR is that it has carved out its own valid niche, and that it'd be good to open the code up to allow people to use alternate ranking systems and run tournaments etc"

TGM by different types of GR I mean separating GR into divisions, thus there could separate rankings for live games, gunboat, ancient map, standard non-live games. You could still bring it together into one overall GR if you wanted to, but the ability to break it down would be interesting (not essential though).

Spyman, I understood what you meant. The issue is that not enough games of ancient map, say, are played to give meaningful results.

(RE point loaning, it started with points being loaned to players in tournaments, by me as TD.)

Fair enough. If there could be just one distinction I would say Gunboat versus Press. But regardless I would just like to see GR made part of the site.

@ ghost, I take your point, but in my view the game should be about solos. Therefore, a system which allows a player to maximise his ranking by unambitiously going for draws, albeit with tough players, has a little flaw in it too. Hence my point about the comparison/consideration of two slightly inadequate systems may give a better indication.
Of course if one thinks that the game is about forming and keeping good alliances rather than solos, and this is an equally valid approach, then the GR system as it stands is a pretty good metric.

fulhamish: "I take your point, but in my view the game should be about solos. Therefore, a system which allows a player to maximise his ranking by unambitiously going for draws, albeit with tough players, has a little flaw in it too. Hence my point about the comparison/consideration of two slightly inadequate systems may give a better indication.
Of course if one thinks that the game is about forming and keeping good alliances rather than solos, and this is an equally valid approach, then the GR system as it stands is a pretty good metric."

Actually (which surprises a good many people), this is irrelevant to the points vs GR debate. Note that both systems assume diplomacy to be zero-sum (i.e. all games are worth equal in terms of total pay-out). If you agree with this in your idea of how payouts should work, then both systems work fine just by using that version of payouts (in Ghost-Rating you need to make a model of how the game works, but that is rather easy to do). If, however, you disagree, then neither system will satisfy you.