Mabye "discontinue" is too much, how about just keeping our points in our profile and keep the point investment game system, GR can be a number next to our name(take over the space where the points now reside) that way we get to keep the best of both worlds, the point based game rationing system and the GR competitive reward.

People seem to be very confused by what I mean by metagaming. Metagaming refers, by definition, to *anything* outside of the game. Clearly, it would be impossible for anyone to be completely free and clear of metagaming.

What I mean by "points encourage metagaming" or "people are gaming the points system." Is that they are making decisions that, from a diplomacy standpoint, make no sense. The decisions they make are artifacts of points, which are themselves something outside of the game. Perhaps this will become clearer with examples:

-I would like to play these players, but I don't want to risk that many points

-I would like to play these players, but I don't *have* that many points

-I would stab him, but if I come in a strong second and give him the win, I get almost as many points.


None of the above (and more) would happen with GR replacing points. And, again, people keep saying you can game the GR, but I've yet to see anyone explain how that would be done. Implemented properly, GR is a measure of skill, so the only way to game it would be to play better and that's not gaming the system.

In PPSC, you could help another player and do strong second to boost your GR.

GR has the opposites problems though.
In GR, it's the better player who then has to decide "I would like to play these players, but I don't want to risk that much GR"

And yes, GR gives you credit for a strong 2nd.

@Kestas: GR could be calculated by the system without too much fuss so whilst I'm a fan of points (although names should show total not free points) it wouldn't be hard to implement GR

All scoring systems that we have have some merit. Be it points, W/D/S/L ratio, GR and so on.

Many people on here play in a way to maximize their points, but there're also many who play to optimize GR. Usually by profiling the type and amount of games they play in.

For example, and I will use MM here, as a player who's been at the top for quite some time and I personally rate very high:
- I play in all tournaments, MM doesn't
- I play live games, World, Ancient, CD takeovers, new players, cheaters, etc. MM doesn't
- I have sent 24781 game messages over 325 games, MM has sent 24107 messages over 87 games. Arguably I invest 4 times less time per game, which allows me to play more games. If we had a scoring system similar to golf or tennis I'd have a much higher rating and would have accumulated a substantially higher prize income.
- I could also try to optimize my GR or points, but I guess it's not my thing. Some people want to play less games and put their all, some others want to play more games and are ok accepting some of these will be poor ones.

So, in conclusion, if you have even the top and most experienced players having different focus and personal assessment of what it actually means to do well, you simply cannot have a single scoring system encompassing it all and making everyone happy.

Points are valid, especially for new players, more experienced ones in any case look at the broader picture and also take into consideration GR and ratios.

Quoting abgemacht: "Implemented properly, GR is a measure of skill, so the only way to game it would be to play better and that's not gaming the system."

Implemented properly...ah, there's the rub. There's no magical algorithm that's going to be absolutely ironclad, because it's nearly impossible to quantify/capture abstract things like intent, objectives, communication, alliances, backstabs, etc...all those wonderful "three-dimensional" things that project out of a finite set of moves on a two-dimensional map.

No matter how complex a GR algorithm is, there are going to be gaps, and if people figure those out, they can take advantage of them.

@fig: I'd really like to see an algorithm for that. I've seen some fairly abstract formulas but TGM isn't a code guy and I'm not sure how up to date they are
If it's true it would be great to see something that could be more directly translated into well functioning code, at the very least it would let us make more informed decisions and have it as an extra detail in the profile page

I thought the math could be summed up thusly, at least for a rough order of magnitude of accuracy:

Each player puts 1/20th of their GR in the pot. So if every player started with a GR of 100, the pot would be 5 x 7 = 35

In a WTA, the winner gets 35, and ends up at 130 (100-5+35). Losers end up at 95.
In a draw, the pot is split equally among the survivors
In a PPSC, the survivors get 1/35th of the pot for each center the end with.

I'm sure Ghost can correct me if I missed the details.

Of course, the logical conclusion is the only way someone with a GR of 1000 would gain something by playing with some 100 GR players would be to win a WTA - a draw is unlikely to result in a net gain, and a PPSC is still going to be a net loss.

I must say i don't bother playing for points (D) or GR rating, I play for fun, and to win, ok a draw is better than letting someone else solo... but all those other measures are, well useful, but not something worth gaming.

It would take too much effort for me to bother optomising my GR with no extra fun added to my games.

I like to play tough opponents as it tests me more than playing easy opponents.

I admit taking not of the points in my league games, BECAUSE it's a league. Winning a league is a different matter from winning a single game. But having lots of points isn't fun, having a look at my GR is interesting to compare my position to other players, but for comparison after i've done my best to beat the best opponetns i can find...

@Kestas - Strat is pretty well there to the best of my knowledge

I’ve only just noticed this thread, and although I can’t give a full response, here are a few notes:

PPSC remains under Ghost-Rating. One aim of Ghost-Rating is to make it impossible to improve your position by carefully choosing which games to play and which not to play. If I wanted, I could avoid difficult games and increase my rating ad infinitum.

Another aim is to make it possible for good players who joined later than similar standard players compete in the rating system. At present, points tend to grow exponentially for good players who play consistently. This is not good for encouraging new players on the site.

A game deposit would be a very good idea if we keep the points system.

I understand your argument that it's a more meaningful system, but it carries the same problems and requires (as I understand it) a greater amount of effort from those running the system.“
Both the points and Ghost-Rating systems do not need maintenance after the code is written.

“I may be in the minority, but I don't mind the points system. As a player for a number of years, but relatively new to this site, I think it helps delineate what kind of game you are in for depending on the pot”

Though other systems could easily emulate this- for example by having different modes (rated vs unrated)

“I also think that the way GR is calculated could be improved. It may look similar to, but it behaves actually quite differently than Elo.”

It is true that GR is quite different to Elo, but for a very good reason, namely that diplomacy is not like chess. It is not the case that by playing better than any of your opponents you can guarantee victory, rather, it is the case that by playing better you can improve your chance of victory. This makes the analysis of normal distribution curves used in Elo-rating impossible to apply to diplomacy.
“Right now, points and GR are both just arbitrary measures of skill. One can point out the flaws with both of them. Some would say that GR is better, since it treats games "equally" and it puts more pressure on the skilled players playing against weaker players. Some would say that points are better, since it allows players to play practice low-point games while also being able to play important, high stakes games.

Points, GR, neither is perfect or better. Removing one in the favor of the other only reduces the amount of information available if you are concerned with getting an accurate picture of a player's ability.”

I’m afraid that this is bunk. For starters, as I have said, I would support the option to have unrated games. Secondly, the idea that it improves a system to arbitrarily consider some data more than other data is pretty absurd. If someone bets all their points in a game, the difference in their rating between victory and defeat is in no way a measure of their skill in either circumstance. That GR is not exact does not make it equally good at rating players as any other system, otherwise I might as well run a raffle every month.

“The reason why the system STILL hasn't been changed (I've seen many threads like this in the 3 years here) is because it works”

Actually, this is flawed logic. All we know is that points are better than having the previous system, which was based of raw wins.

“I've moved back and forth on this quite a bit, as has TGM and the general community,”

Actually, I’ve just decided not to bother you with it anymore since there is nothing more I could tell you to try to convince you of my view that GR would be a big improvement to 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 is currently scripted in perl (by Tantris)

“Quoting abgemacht: "Implemented properly, GR is a measure of skill, so the only way to game it would be to play better and that's not gaming the system."

Implemented properly...ah, there's the rub. There's no magical algorithm that's going to be absolutely ironclad, because it's nearly impossible to quantify/capture abstract things like intent, objectives, communication, alliances, backstabs, etc...all those wonderful "three-dimensional" things that project out of a finite set of moves on a two-dimensional map.

No matter how complex a GR algorithm is, there are going to be gaps, and if people figure those out, they can take advantage of them.”

GR is a measure of how likely you are to win against an average set of opponents, which is what is taken to be the definition of your skill.

“I thought the math could be summed up thusly, at least for a rough order of magnitude of accuracy:

Each player puts 1/20th of their GR in the pot. So if every player started with a GR of 100, the pot would be 5 x 7 = 35

In a WTA, the winner gets 35, and ends up at 130 (100-5+35). Losers end up at 95.
In a draw, the pot is split equally among the survivors
In a PPSC, the survivors get 1/35th of the pot for each center the end with.

I'm sure Ghost can correct me if I missed the details.

Of course, the logical conclusion is the only way someone with a GR of 1000 would gain something by playing with some 100 GR players would be to win a WTA - a draw is unlikely to result in a net gain, and a PPSC is still going to be a net loss.“

This is not correct:

In WTA, effectively everyone puts 2/35ths of their rating into a pot, and then it is divided in the normal WTA way.

In PPSC, it is more complicated, everyone puts in an amount which takes into account how likely they are to win and how likely they are to get part of the pot by surviving, and then the pot is divided in the normal PPSC way

@Ghost - re the PPSC, it stills sounds like something that should be fairly easy to code using simple calculations.The big difference seems to be the 'amount you're putting in' calculation, which I assume is based on the GR of those you are competing against, but that would (I assume) be simple from a mathematical perspective....

Am I incorrect? I freely admit I'm no longer a code monkey, and I also haven't looked at the calculations you've posted recently in any detail...

@strat - it is quite simply encoded yes, what he means is basically that its not worth explaining the details of.
Merely the fact it is currently scripted in perl is enough to show making a php version realistic

Soo... one set of concerns are answered: whether it is feasible to integrate GR into the system.

I would assume a good first step would be to integrate GR into webdip on a trial basis. For example, I would assume that GRs would change after every concluded game, instead of on a monthly basis.

That, of course, requires development resources, but fewer than tackling the entire project all at once. Call that one unit of work, that should have a minimal impact on the rest of the site - it's basically just going to calculate GR on the fly instead of forcing Ghost to do it all the time, and display them on the player's stats page

The actual integration of GR into the game system would, I would think, be considerably more challenging - probably 4 or 5 units of work *at least*. We're not talking about a bolt on in that case; we're talking about changes to just about every module on the site.

Note that this is me looking at this from a project manager-type standpoint, not a coding standpoint. Again, I'm not going to say it's "easy" to do something that I can't do.

If it is already in Perl, the move to writing a method in PHP that passes a game number in should be fairly straight-forward. The trickiest issues I see are when it should be called. And do you use the GR of a take over position? Or the GR of the original owner of a position, or both?

Actually strat, implementing it when a game finishes is both less work and better than doing it monthly.

I don't understand why you think it'd be so hard to add?
All you need to do is modify the user table in the database (add 1 new field for GR), possibly modify the membership table (used for membership of games not the site) if you want to use the GR from the start not end of a game, modify the hall of fame to add a second sort option (1 if statement & a toggle button), add it to a users profile (again, 1 line code) and possibly change the number in your brackets if GR is prefered (change one word).

That's not a lot of work at all.

The formulae are all based on +, -, * and /

How to deal with CDs when calculating other players' ratings? You take the average rating for each of the diplomacy turns (CD power=0 GR) for each country.

But a higher GR player takes over a 2SC CD in Spring 04 and the game runs for 12 years... He is in for 2/3rds of his normal GR buy in but in a position that'll be lucky to survive, much less draw or win. Does that seem fair?

It needs to be adjusted based on the strength of the position at take over and, as a punishment, take the GR from the player who abandoned. Kind of like points in that regard. It would be easy to do. Compute the take over's normal GR buy in, divide by the average owned SCs per nation (living nations only) and multiply by the number of SCs he is "buying". Take over a strong nation and you risk more. Take over a 1 SC late in the game and you risk very little.

@Draugnar; you're definitely going down the same thought process I've gone down whenever I've thought about replacing points with GR here, the devil is always in the details, and it's important to ask these questions, though I think if it's implemented here as an alternative ranking system we can weigh these things up and see how it works live, and answer these questions in practice rather than in theory

@stratagos; "Again, I'm not going to say it's "easy" to do something that I can't do."
I appreciate that. Frankly I don't think anyone knows how to do it right now, because it wasn't really intended to work on a live system. I suspect the Perl script TGM mentions works on a list of games/players/wins/losses, instead of points which work on events like "player joined", "game won", "civil disorder", etc, and turning GR from a bulk-processing algorithm into something live could be tricky (then again maybe it would be straightforward. It would be great to have some help with it since that would need intimate knowledge of GR more than webDip)

@TGM; this is the last time I brought it up as far as I can remember:
kestas.j.k: the G rating could be calculated sure
kestas.j.k: as part of the tournament system and put in the profile page perhaps
TGM: And that wouldn't be a problem to you?
TGM: If so, it might actually be better to have me upload the ratings on a monthly basis, and then we don't have to program a new calculator
kestas.j.k: well it depends how it's calculated
kestas.j.k: that approach may not be easier
TGM: It is fairly involved

And a year ago:
kestas.j.k: but Ive been thinking it over and have realized that you have answered most of the points I brought up about the GR
kestas.j.k: some time over the hols Ill simulate it and see if anything sticks out, if not Ill cast a vote and see what people think

So the idea that the debate is dead because it's not worth discussing isn't really right. I still think both systems have pros and cons (and you've also admitted this), and we probably couldn't agree on what those pros and cons are, but we do agree that both systems have their advantages and we both want to see some level of official support.

But it seems like whenever I come to pulling the trigger there's always a lukewarm response, and it feels like if I want it done I really need to chase it up. I can't trace it down but I remember not being able to easily get access to the spreadsheet used to calculate it etc

As far as I'm concerned implementing GR here would:
- Provide another stat for the profile page
- Give another way for people to rank themselves as with the hall of fame
- Would be useful for future tournament or league support, which would let players organize themselves into "pools" of players which could be more exclusive and have better multi/meta-gamer protections (i.e. it could be good for future development)

So what's standing in the way?
- Can I get access to the Perl script you use?
- Are you willing to host the ranking here?
- Can it be converted to be event based rather than bulk-processing?

I've been leaning more to the idea that you prefer it as an off-site hosted thing, posted monthly, and separate to the site itself. I would quite like to see it get more official support, but it's your system.
If that is the case though I think you should be more open about that in these threads which call for it to be adopted officially, and if that's not the case then let's get the ball rolling

Well, I'm really glad I made this thread, because my understanding of the situation was completely wrong.

I though there was tons of support for GR integration from the community and Ghost, support for ditching points, and little support for GR integration from the Devs.

But, after reading, it seems that people still like points, are wishy-washy about GR, the Dev team supports GR integration, and Ghost is unaccounted for.

Interesting.

Well I don't want to put words in anyones mouth, it could well be a misunderstanding. 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

I think GR is great and would enjoy seeing it integrated into the site, but I don't think it should replace points, though I suppose I do fear change and get attached to traditions and the points have been around as long as I've been here.

nor, I suppose, would you want that fear-inspiring point total to disappear =P

seriously, you should consider waiting for a "newbie game" to get six people, join it so no one can leave and intimidate the crap out of them

the epic meltdown would be amazing

Sorry, kestas, you seem to misunderstand what I was saying in a few places. I’m not making the argument for adding Ghost-Ratings because you already know what the arguments are.

“So the idea that the debate is dead because it's not worth discussing isn't really right. I still think both systems have pros and cons (and you've also admitted this), and we probably couldn't agree on what those pros and cons are, but we do agree that both systems have their advantages and we both want to see some level of official support.”

Basically what I was saying is that I have nothing further to contribute to the debate, not that it is dead.


“It is fairly involved”

Basically was referring to the fact that the formula seems difficult to explain (and very long to write down in one go), and also that a full implementation does involve more than just adding a bit of code to turn a csv input into a number by each username in the database (I’d have thought). Also, as regards tournaments, I thought it might be better to have it so that forks could have their own rating system (that wasn’t GR) if they so wished. (I also am all too aware that development time is limited)

“we both want to see some level of official support”

I really wasn’t aware of this.

“- Can I get access to the Perl script you use?”
Yes
“- Are you willing to host the ranking here?”
Not totally sure what this question means, because an implementation would just run itself, but assuming you mean the same thing as “are you willing to have it hosted here?” then the answer is yes.
“- Can it be converted to be event based rather than bulk-processing?”
Yes

I've sent you the perl script, the GUI for it (not that you need that, I suspect) and the two csv files that allow banned players to be removed from the calculation and determine the weightings of the variants in the calculation.

When I joined the site, I noticed the Hall of Fame and understood part of the motivation of playing was to acquire points. I'm not saying that's the only reason to play, or even the main reason to play, but it certainly was an aspect to the site that measured success, as I understood it, so to take that away after years of playing with that as a significant consideration on this site would be a shame and certainly disappointing, but it is what it is and we'll all move on if need be.

PE, I could not be less intimidating. I tend to underwhelm rather than overwhelm, and significantly at that. I hate to disappoint you, but I am very much used to disappointing others on this site, I'm somewhat set up for failure in that respect, sorry in advance.

"I though there was tons of support for GR integration from the community and Ghost, support for ditching points, and little support for GR integration from the Devs.

But, after reading, it seems that people still like points, are wishy-washy about GR, the Dev team supports GR integration, and Ghost is unaccounted for."

abgemacht, I still think there is tons of support for GR integration from the community, why do you say otherwise?

If I have discerned who you are in the Invitational round I'm in -- not going to discuss that, just saying that if you're who I think you are -- your definition of "underwhelming" is still scary good >_>

And besides, I was talking six totally new players to the site, still 100 and all, and you just walk in and be all "hey guys, I'm MadMarx, how you doin'?" It'd be like walking up to a group of single ladies at the local bar and being all "hey, girls, I'm Brad Pitt, nice to meet ya." Might even be as much drool and swooning if the newbies are sufficiently enthusiastic about webDiplomacy.net...

I dunno. I'm speaking as an average Joe here. Maybe it's not as big a deal once you're there as it is looking up there. It just seemed like a fun little troll move to try =P