There is a simple question of what you want the system to acheive. If you want it to be a means by which people can quickly gain significant fame, and where you can have super-high profile games, then it works well enough. If you want to have it as a means of genuinely representing a player's skill, then it catastrophically failing in it's duty. I could open five fifty pot games and get resonable, rational players, but who are not particularly skilled at the game, and hence win-out, or I could open a single two hundred and fifty pot game, and find myself playing against better opponents, who not only know the game well enough to understand my arguments, but also well enough to see through the weaker ones, make arguments for themselves and so on, and I am not going to do as well: there is no greater reward for beating MarekP, Rait or Dangermouse (I apologise for the cynical manipulation of three high profile players to support my cause) than for beating Tom, Dick or Harry.

Now, the argument Kestas makes is that "There's no real way to express how good a player really is in a single number", however I strongly believe that that is wrong. People play to a different standard in each game, however they have a general standard. When I play a certain group, you can't say definitely who will win, but you can say the likelyhood of any player winning, and that can be expressed as a number, a rating. That is the basis of the rating system I have developed:

A rating of 2000 means that a player is twice as likely to win a game as somebody in the same game with a rating of 1000. Thus in any game we can easily calculate the expected result (The likelyhood of their winning) from each player's perspective as Player's rating/ Sum of all players' ratings. This gives a value between one and nought, where one is a victory and nought is a defeat.
Now the following formular can adjust player's ratings.

New Rating=Old Rating + (Result-Expected Result)

A small Modification is this:

New Rating=Old Rating + V * (Result-Expected Result)

The multiplier V is simply a way to change the volatility of the ratings. Too high and the inaccuracy is great because each game dramatically changes a player's rating. Too low and player's skill may change faster than their rating, as not enough games are player in the interim. I have yet to decide on the value for V.

A value I have decided upon is that an average rating (the default for a player yet to play) would be 100, so sombody with a rating of 110 would have 110% of the average skill.

A final thing, I would greatly appreciate it if Kestas were to send me a copy of the games database (something I was going to ask for earlier, but we never saw one another on MSN, as the plan had been) so that I can find the value "V" that works best and hence the ratings of every player here based on this system.

This sounds like handicapping, like in golf or bowling. In bowling, as long as you beat your average, you have a good chance of winning a tournament, regardless of if the guy on the next lane bowled 100 points higher that you.

So if you play 2 games, one with players higher ranked than you and one with all players lower ranked than you, if you lose both, the 2nd game will have much more impact on your new rating, since you were expected to lose the 1st one anyway.

Thanks Zaru; I'm currently trying to pitch the points system over at Facebook Diplomacy, and I'm getting all the same criticisms I got when it was first added here, so it's good to hear something positive about it (even if followed immidiately by something negative ;-)

Ghostmaster: You haven't said what "result" is, so I can't say much about your idea, but I don't really like to encourage debate about the points system since it's good enough and I doubt everyone would agree with your way to measure skill. (Also I don't like magic numbers like V)

Kestas, the only thing I would change is that if I have 1000 points, and I am currently in a game with a pot of 340 and I have 5 supply centers, my rating should be 1050, not 1000. I realize it would be a difficult query to write that would contstantly change, so maybe you could only run it once a day, but a player who has 1000 points and no current games and a player with 1000 points and 10 games going are ranked the same right now.

Zarathustra, you suck up.

Hey, why did Zarathustra's point total just go up for no reason?!

TheGhostmaker: you're not taking into account that some countries win more than others.

I think what Ghostmaker is proposing resembles very much the IRB metod for ranking. I don't really want to explain it though, so go to www.irb.com

The point system is great, but I agree with Treefarn, it should be modified in some way that points currently bet in ongoing games shouldn't negatively reflect a players current point total. ie. they don't come off until after a game is finished/player eliminated. The off-set however would be that all new players would always have 100 points at all times. However, the benefit is that a very experienced player betting all his points wouldn't read '0' points and look like a newbie.

Although then you'd also have to keep your available points to bet separate from your ranking points. Otherwise you could be spending points you don't yet have, and could end up in the negative.

I don't think that modifying the system by adding the points You have bet into ongoing games would enhance the system much. The points You have bet are expected to be returned eventually (preferably in a greater amount, in which case Your ranking goes up) or lost (in case You rating goes down). So, their effect on Your rating is simply delayed (and it's not the same as not taken into account). In long term (in case of phpDip, let's say after 20 games) it doesn't really matter (unless You are playing 20 games simultaneously, and there are not many of those people) - Your rating has averaged to some level, either with the upward or downward trend (which is really biased by reset system).

Well, I like the point system. The Ghostmaker, I think your problem lies in the fact that you play with many of the same people. If you constantly play in games with the same people than no new points are added into your closed economy. You are merely trading your own points amongst each other as you each win and lose. The effect is more noticeable in WTA because it's much easier to lose your stake in the game. However, as long as you win more than 1 out of 7 of your games at the same stake of WTA than your rating will increase.

Now, I can see you have a much better win percentage than 1 out of 7, so why do you seem to have a lower score than others with the same win percentage? Perhaps it's because you have more points invested in games than others. Perhaps it's merely a temporary low brought on by a big loss or two, in which case, your statistical trend should soon lift you up to a more agreeable level. Playing PPSC instead of WTA will reduce the effects of temporary lows, though it also reduces the effects of temporary highs. Your rating is not your highest, or your lowest points over time, it's the average. That's why I wish points invested in a game were still shown in your points total.

Perhaps you lose all your high stakes games and win all your low stakes games. This is not unusual, as the high stakes games are probably filled with higher ranked players, while lower ones are filled with lower stakes games. Remember to prioritize and if you have limited time, to devote that time to your higher stakes games. As it stands, that does not artificially lower your rating, if you win low stakes games and lose high stakes games, than your rating should fall between those two extremes. Play some midstakes games and see where you fall.

I didn't actually look at your game history so I don't know if you fall into this category. I personally believe that your rating does not reflect your skill, but I believe it's your fault and not the system's.

The theory is there for a functional points system. My advice is to play with different people, because it only works when there is an open economy of points. As well, don't be down when you're points seem lower than usual, and don't get cocky when they're high. Neither is representative of your true rating, though I think most people like to think it's their high point. I would ask Kestas to display points as a number of points available and a number of points in total, so that the effects of temporary lows as a result of joining a lot of games aren't noticeable.

Chrispminis sums it up well as usual :-)

Regarding the total points sum thing, this has been suggested before, and on Facebook phpDip. The reasons it hasn't already been done are:
- It could confuse newcomers, where are your available points displayed and where are your total points displayed? If you only see your total points you don't know how much you have free to bet, if you only see your available points you think you have fewer points than everyone else, if available is used in some places and total in other places how do you tell which is which? (If you don't think this could then confuse you haven't seen some of the FB phpDip forum questions)
- On FB phpDip I think people are less likely to bet all their new points away on big gambles if they know it'll look like they have no points immidiately once they've bet
- As Rait says, it's only delayed/averaged, and safe earned points probably should be worth more than bet points

I can see both sides of the argument, but I'll always go for the one that doesn't need me to do any work ;-)
(I'd accept a patch, if it dealt with the issues above)

"Ghostmaster: You haven't said what "result" is, so I can't say much about your idea, but I don't really like to encourage debate about the points system since it's good enough and I doubt everyone would agree with your way to measure skill. (Also I don't like magic numbers like V)"


aoe3rules: Nation compensation I think is perhaps a bit of a risk: Diplomacy to an extent should be able to take care of that: If you are Italy, then you can say "look ally with me, I'm small but handy" whereas Russia cannot calm people's fears in the same way.

Chrispminis:

"Perhaps it's merely a temporary low brought on by a big loss or two, in which case, your statistical trend should soon lift you up to a more agreeable level"

See Revelstone... I was Russia (one of my two worst nations) I devoted lots of time to it too... over 3000 words in five hours of press gives you an idea.

Remember, I am not saying that the system doesn't give an indication of skill, it does, and nor am I bitter about my Raiting (Sorry, couldn't resist). The major problem I have is that beating a bad player pays just as well as beating a good one at the same stake, which is in principle totally wrong.

Kestas, can I please have a copy of the database if only to calculate the ratings (I am interested in the system if nothing else)? I would also be happy to make a patch.

Ghostmaker :P


Result is 1 for a win, 0 for a loss 1/2 for a two way etc in WTA.

In PPSC 1/34 for one centre, 2/34 for two etc.

The method is one based on statistics alone, and so works. The inaccuracies are a factor of the number of games played and the inconsistency of players.

Thirdly, the "magic number" V is nothing of the sort. It is literally just the sum of the points that change hands for any particular game. High means volatility but fewer games to be played to settle a rating, Low means a more settled rating but longer time to build it up. I need some actual numbers to decide where to put that number, but I imagine a value of about 8-10 would be about right.

Formatting was screwed in that post^^^

aoe3rules: Nation compensation I think is perhaps a bit of a risk: Diplomacy to an extent should be able to take care of that: If you are Italy, then you can say "look ally with me, I'm small but handy" whereas Russia cannot calm people's fears in the same way.

Chrispminis:

"Perhaps it's merely a temporary low brought on by a big loss or two, in which case, your statistical trend should soon lift you up to a more agreeable level"

See Revelstone... I was Russia (one of my two worst nations) I devoted lots of time to it too... over 3000 words in five hours of press gives you an idea.

Remember, I am not saying that the system doesn't give an indication of skill, it does, and nor am I bitter about my Raiting (Sorry, couldn't resist), evidenced by the fact I was arguing for a new system when I was 35 in the ratings. The major problem I have is that beating a bad player pays just as well as beating a good one at the same stake, which is in principle totally wrong.

Kestas, can I please have a copy of the database if only to calculate the ratings (I am interested in the system if nothing else)? I would also be happy to make a patch.


"Ghostmaster: You haven't said what "result" is, so I can't say much about your idea, but I don't really like to encourage debate about the points system since it's good enough and I doubt everyone would agree with your way to measure skill. (Also I don't like magic numbers like V)"

Ghostmaker :P


Result is 1 for a win, 0 for a loss 1/2 for a two way etc in WTA.

In PPSC 1/34 for one centre, 2/34 for two etc.

The method is one based on statistics alone, and so works. The inaccuracies are a factor of the number of games played and the inconsistency of players.

Thirdly, the "magic number" V is nothing of the sort. It is literally just the sum of the points that change hands for any particular game. High means volatility but fewer games to be played to settle a rating, Low means a more settled rating but longer time to build it up. I need some actual numbers to decide where to put that number, but I imagine a value of about 8-10 would be about right.

Alright. After some debate with The Ghostmaker over MSN, I've decided that he does have a point... The debate is on-going but I thought I'd update the forum with some of my thoughts subject to public deliberation.

The point he's got, is that the ratio between bet and pot won is always going to 1:7, no matter who you face. This means that if you put all your points into a super high stakes game and beat out the top players on the site, or if you put all your points into several small games with the lower end players on the site and won them all, then the result would be the same.

Now, I remember when the points system was inducted, one of the main arguments for it's creation was the ratio... because in those days, there were people who liked to play more than 20 games, and those who only played a couple. The reasoning was that, the people who liked to play more, could do so, by playing many lower staked games, while people who played fewer could play a couple high staked games and the mathematical outcome for the two people of same skill level would be the same. However, a side effect of the points system is that the people who play too many games are penalized and hence there aren't too many people like that anymore.

Now there are a few solutions I can think of that would solve the ratio problem and keep the current points system... but there quite arbitrary in nature and I don't like that.

1. Put a cap on total games available to play at a time. The less games one can play, the less the ratio will have an effect. Obviously, this is going to met with resistance, including from me.
2. Restrict higher ranked players from joining lower staked games. This would likely involve some arbitrary percentage of points as the minimum bet. The higher the percentage, the less of an effect the ratio would have. Obviously this represents an even bigger restriction on gameplay here, so I would strongly advise against it.

So here's my last effort to save the points system... I would argue that winning a higher staked game and the relative number of smaller staked games is equal in difficulty. Winning 10 games of 10 bet/ 70 pot WTA against 60 lower ranked players, I think is just as hard as winning one game of 100bet/700 pot WTA against 6 higher ranked players, who would theoretically be 10 times more skilled.

The difficulty comes in the sheer number of people you have to defeat. As most people know, playing more games than you can handle very quickly leads to a deterioration in the skill displayed in each of your many games. I think micromanaging 10 games with 10 times more players and winning 1 game with players with ten times the skill are probably equal in difficulty, and warrant the same reward.

Playing an equal number of games, you stand to gain points much faster by playing higher staked games, though you stand to lose much faster as well. It still comes down to your average skill represented by an average of your points rating.

this is how I feel about the government
just kidding
I dont really have alot of input about the point system since mine is perpetually low.

generally if someone has morte than 500 points I just assume they're gonna beat me

Ok, Chris puts forward some interesting ideas, and I am pleased to have made my point good and clear.

"1. Put a cap on total games available to play at a time. The less games one can play, the less the ratio will have an effect. Obviously, this is going to met with resistance, including from me.
2. Restrict higher ranked players from joining lower staked games. This would likely involve some arbitrary percentage of points as the minimum bet. The higher the percentage, the less of an effect the ratio would have. Obviously this represents an even bigger restriction on gameplay here, so I would strongly advise against it."

I am in full agreement with you that those are not practicable, and neither really works: with the former, you get a better average gain with people weaker than you (some positive value) than people the same strength as you (in the long run that is nil); with the latter, you still have no compensation: you just limit the range of players to make a rough fix.

" I would argue that winning a higher staked game and the relative number of smaller staked games is equal in difficulty. Winning 10 games of 10 bet/ 70 pot WTA against 60 lower ranked players, I think is just as hard as winning one game of 100bet/700 pot WTA against 6 higher ranked players, who would theoretically be 10 times more skilled.

The difficulty comes in the sheer number of people you have to defeat. As most people know, playing more games than you can handle very quickly leads to a deterioration in the skill displayed in each of your many games. I think micromanaging 10 games with 10 times more players and winning 1 game with players with ten times the skill are probably equal in difficulty, and warrant the same reward.

Playing an equal number of games, you stand to gain points much faster by playing higher staked games, though you stand to lose much faster as well. It still comes down to your average skill represented by an average of your points rating"

Well, that can be countered by the relative amount of effort for one large game and several smaller ones. People are less likely to engage in lengthy press for a 70 pot game than a 700 pot game. And the ratio problem remains: Against people who you are the same standard as, you come out with no net gain (as should be the case), but against people who are weaker than you, you benefit from winning, even should your skill not increase. Simply because a huge exploit may be difficult to achieve for most players, the proplem remains the same, you do better in the system against weaker players as there is no incentive to beat stronger players, and your rating can change without a change in your skill (whereas should your skill remain the same your rating should broadly remain the same)

> Kestas, can I please have a copy of the database if only to
> calculate the ratings (I am interested in the system if nothing else)?
> I would also be happy to make a patch.
Don't really want to encourage this pursuit tbh, I just can't see it happening. Even if I thought your system was an improvement, the current system is still good enough and everyone is used to it

> Ghostmaker :P
Oops, sorry about that

> Result is 1 for a win, 0 for a loss 1/2 for a two way etc in WTA.
>
> In PPSC 1/34 for one centre, 2/34 for two etc.
Okay, so to bring it back into more familiar terms it's like the points system, except the bet is relative to the number of points other players have

I don't really get the divide by thing. It's like by having an expected value you're setting a target for that player, which is like setting an objective. A weak player in the game could get points just by clinging on, a strong player may need to be determined to dominate the game just to break even. It seems to put people into roles like a Risk mission card, skill is no longer the only difference between players when a game starts

There's also no built-in way to deal with taking over civil disorders, which is another thing that'd have to be added on. It couldn't charge the same rate to join a game at the start as to take over a powerful CD player. What happens if a small player takes over a CD in a large game (or vica-versa); do all the large players suddenly have a higher expected value and a sudden bet size increase?

It also doesn't deal with people joining too many games (Unless your reputation points immidiately go down by the expected result amount when you join. And if that happens players could start lots of new passworded games, lose all their expected result points (which would have to be refunded after the passworded games expire), then join many games with no risk because there's no expected value)

Also if the V value is fixed that means people will always go up in points by the same steady amount per game; players that can play lots of games at a time will pull ahead, without offering a choice to other players to play fewer larger-stakes games.

Anyway, I could go on. By the time this system addressed all the potential problems I think it'd either have to have a regular points system running alongside it, or have loads of configurable options and formula modifiers to try and balance out the various inequalities.


Regarding the flaw that's supposed to make all this worthwhile; good players can play many games against small players, instead of playing with people of the same skill.

To an extent this is a good thing; it means new players get an idea of how good players play, and aren't faced against only CD players while the fun players are unreachable in high-stakes games. Remember it's about having fun games, not a serious exclusionist Diplomacy league.

Has it gone beyond that and being abused just for points? Well the highest ranking players didn't get all their points playing with <100 points newcomers; the distribution of pot sizes in games follows a fairly smooth exponential curve upwards as you would expect.

I just don't see what's worth all this trouble, and it seems like most people are okay with the points system as-is


Please don't write back with modifications/tweaks; we're not brainstorming a new points system, I'm just trying to give you a proper, better thought out response.
If you do have the skills to make a modification like that consider something that everyone agrees on and would all appreciate :-)

kestas: "It could confuse newcomers, where are your available points displayed and where are your total points displayed? If you only see your total points you don't know how much you have free to bet, if you only see your available points you think you have fewer points than everyone else, if available is used in some places and total in other places how do you tell which is which? (If you don't think this could then confuse you haven't seen some of the FB phpDip forum questions)"

Fair enough, but I think it would be better if the hall of fame, at least, was constructed on the basis of total points, i.e. points available + points in play.

I figure you probably wouldn't thank me that much for an in-depth response, but there are a couple of things I would like to make clearer.

Firstly, the "expected result" gives a "target" in the same way as winning 5 or more SCs is a target in PPSC- it shouldn't make a difference as you still play for the greatest benefit.

Civil disorder is a hard one Leaving a game obviously counts as a loss, but you have to adjust the expectation for the joining player. It does get a bit messy, but not unmanageable.

You do not loose any "points" on joining a game, so this doesn't stop you from joining many games, but it does mean that when you do so, unless you perform well (in which case there is no problem) your rating will plumet to represent your poor play.

There is one thing that I really want to make clear, which is that this system looses the idea that people's ratings go up and up and up the more they play. That means that playing more games will not inflate your rating above that of a simillarly skilled player who plays one game at a time.

This system doesn't stop playing with weaker players from being profitable, just stops it from being disproportionately so. "Abuse" is perhaps the wrong word, simply the way you play on the site, the games you like to choose, helping or hindering your progress.

And anyone can join any game under this system, so it is more inclusive, not less :)

Just finally, I would like to make it quite clear that, although I think this could be a better system, I do not want to hinder the introduction of other, bigger and less devisive things (moderators, for instance), and nor do I want to compell you to do something that you yourself do not support.

"I think it'd...have to have a regular points system running alongside it"

That's exactly what I've suggested. The point system works fine the way it is, I'm not knocking it one bit. But a ranking system based on some calculation other than bet-able points is the answer to this dilema, I'm sure of it.

There's no need to alter the current point system, just add a ranking system to run along side it.

"I think this could be a better system, I do not want to hinder the introduction of other, bigger and less devisive things (moderators, for instance), and nor do I want to compell you to do something that you yourself do not support."

Agreed.

I too would like to see a rating rather than betting or reward system existing in some form, even if as a secondary one.

I'm gonna hafta reference the Xbox 360 game: Halo 3. :)

It does this rather well. In Halo 3, there are levels and there are ranks. Your level reflects how well you do in ranked matches (as opposed to playing in a social, "for fun" match)

Your rank (army style rankings...private, corporal, sargeant, etc.) then, is the amalgam of ALL your games (ranked and social).

When you win in a ranked match you level up. When you level up, your rank goes up at certain predetermined intervals.

When you win a social match, you get points. More points = higher rank.

Those are the two ways to get a higher rank. Leveling up (winning ranked matches) is a faster way to get a higher rank than with points (winning social matches).

So to make this relevant to diplomacy...the point system ranks phpdip players via the Hall of Fame. More points = higher rank in the Hall of Fame.

All you need is a secondary ranking system so that it reflects your games when accounting for playing against better players, win/loss ratio, surviveability, draws, etc. etc.

So, new players will start out as 100 pt Noobs (or whatever label you want to use, with cool shiny graphic!).

Points stay the same!...to be used to barter. You merely add certain benchmarks to "level up".

It's practically already being done with our level designations in our profiles!!! I'd say just enhance it a little by not making it linked to total points AT ALL, but rather, with some calculation of wins, eliminations, survivals, total games, etc.

Did any of that makes sense? :)

Why would that be any more representative of skill than the current system?

Because a system that includes data from actual results (wins, losses, survivals, draws, etc.) is measuring not your ability to gain or lose points, but your ability to win, or lose or draw by taking into account such.

Imagine if joao (who I've seen using 1000 of his 1053 pts to start a game) loses his 1000 pt game. His points are no longer reflective of his skill.

And that leads me to my only problem with the point system...there are many good players with low points. Therefore, points are not necessarily a strong indicator of skill.

Don't get me wrong, I don't mind the point system as it is. But if ranking skill is partly its goal, it fails in some circumstances.

And the solution, I think lies in the addition of a secondary ranking that takes into account a number of things (to be decided upon) that ARE reflective of actual skill (ie. defeating better opponents, high win percentages, low eliminations, etc.)

Well Darwyn, I've discussed what you're talking about for quite some time. It's not that the points system isn't representative of skill, because it very much is, just as an average over a length of time. joao losing his 1000 pt game is quite likely as he'd be playing against players of a 1000 bet calibre. However, he'd likely rise up again to reflect his skill. If you look at the Hall of Fame, the ranking is very similar to win ratio. The points system encourages a high win ratio as well as prioritizing higher staked games. If you look, the top 10 all have around 40% win percentage, the top 30 have around 30% win percentage etc. There are a few anomalies, but none that are gamebreaking.

However, there is a problem with the points system that The Ghostmaker has shown me, but that problem is inherent in any win ratio system as well, so your secondary one would not resolve it.

The problem with the points system is as follows. Theoretically, if you played players who were exactly your skill level at the same pot level, than over time your points should remain fairly static as you win and lose in equal ratios to people of your own skill level. However, if you consistently play stronger players, you might win some, but you will mostly lose them as you are the weaker player. So on average your points will continue to decline, despite that your skill may not have declined, in fact it is probably better for playing the stronger players. In a similar fashion, if you consistently play weaker players, you might lose some, but you will win more ratio-wise as you are the stronger player. As a result your points will slowly climb despite that your skill may not be increasing at the same rate. As such, the player who plays stronger players has a lower rating than the person who plays players of a similar skill level who in turn has a lower rating than the person who consistently plays weaker players.

The same problem is inherent within any ranking system relying solely or mostly upon a simple win ratio. As playing weaker players guarantees a higher win ratio, and playing stronger players guarantees a lower win ratio. Of course this problem is purely theoretical, but we can see some of the effects as players who consistently play against stronger players are pushed to the ground.

I wouldn't overhaul the points system without a flawless system coming in, as this is it's only flaw that I can see, and despite that flaw we have a fairly decent Hall of Fame.

I'm interested in what results The Ghostmaker's proposed system will have now that he has the database. I'm not quite sure it's flawless system yet, but it certainly addresses the problem that the point system has, though it'll have to replace the point system in all roles before it replaces it.

I don't pretend that this discussion will actually lead to any change, but I hope the debate leads to some ideas. I don't want to incite any dissatisfaction with the points system either, as it is a very good system. Extremely impressive, and I think quite the step up from the simple win ratio system that was advocated by me prior to the point system.

I am afraid that, unless Kestas allows me to, I am not at liberty to publish the results of the test I am doing- it was a condition on my getting the database was expressly that "I will not refer to the results from the DB test at any point (on the forums)"

Ah well. =D

I never said that, there were no terms :-P

I said I didn't want to send the DB because I knew you'd be back in this thread today using the DB to claim to have verified that your system is flawless, or created a new system that is

Right now I can't think of anything more boring than critiquing system after system or analyzing numerical test results; if you want to post some data go ahead, but my response is in the FAQ :-(

Really. The impression I got was that you were saying that I shouldn't do so.

Which, note, I was perfectly content with.