oh, and i forgot to mention: don't expect Kestas to help.

I, belsherj, vote for option#3 (Elo)

I play a variant of Risk online called conquer club, and they have a points system that works pretty good, you might want to look into it.

I vote for option #3 Elo.

#3 for me, definitely the Elo system.

sadly, even if every single person votes for a particular method, the code will not magically appear.

McCain, are you voting for that? if so, post your suggestion as #4.

Is there anyone in particular capable of writing it?

capable AND willing? we'll have to see. if no one else volunteers, i'm not doing this by myself. so we'll just forget the idea and give up.

Clearly I've touched a nerve with my dismissive response to the Elo rating system, but this post is pretty inflammatory aoe3rules..


Without turning this into a rant there are a few implicit assumptions which I'll try to address as impersonally as I can:

- This vote is binding ("if an alternate ranking system wins, everyone who is willing to help develop it should start working")
This vote isn't binding: Only people interested in the point system will be voting, this vote isn't a fair representation. Not everyone understands the role of the points system and what the Elo system is, the change would affect people not using this site, and democracy is well known to not always be a good approach to software development.

- The Elo points system "solves all problems of the current point system"
This is, at best, debatable. Listing it as a "pro" is a childish tactic

- I owe my help ("don't expect Kestas to help", "Cons: too much effort.")
I don't owe anyone my help: I'm busy, I do this in my free time. The fact that you think you need to make it explicit that you can't expect me to help is revealing, and disappointing, to say the least

oops, sorry. i meant "don't expect kestas to do the whole thing for us".

I think that with some form of ELO system, you don't have to abolish the current points system. They can work together.

I vote #3.

what portions of the Elo system? and do you mean we can just use some of the old code, or actually combine the two systems?

All portions of the ELO system can be implemented without adversely affecting the points system...I think.

As far as I can tell, the points system is a great bartering system for this game that does one thing very well...it keeps noobs from joining skilled games...it makes them earn it. If kestas's argument that the point system is fine as it is, I think he's partly correct.

What the point system doesn't do very well is accurately reflect a player's skill. Is this not why all these threads and debates have started?

We need a system that accurately reflects skill at all times. I think the ELO system does this well.

But do we need to abolish points? I don't think we do. In fact, it might give the ELO rating a bit more depth.

Take Rait...he'd be a 8000 point Grandmaster (or some label)...theoretically, if he bet all his points in one or two games the current point system would have him at 0 or 100...a noob. But we all know that's not the case. Instead, he'd be a 0 point Grandmaster.

All you need to do is add the code (obviously) and swap out the little "D" graphics denoting points, with your corresponding ranking graphic. We all know what the numbers mean, but now when we go into games, we'd know exactly the skill of a player regardless of points.

Does that make sense? Sorry, it looks like I just rambled...

no, you didn't just ramble; that makes perfect sense.

instead of the "V" variable, we could use the bet for the game.

this would work perfectly! (okay, not literally, but...)

Vote for ELO, option 3

@Kestas: i know you do this in your spare time. i'm busy too. i wasn't being sarcastic. sorry about any misinterpretations.

@Darwyn. I agree completely, the ideal solution would be to have some sort of ranking system, maybe Elo, maybe some thign else that's based on ones win percentage, and leave the point system in place as a separate thing entirely.

I prefer the Elo system, but I'm not voting for it. I think it's pointless to have a democratic vote on such an issue for the reasons that Kestas has outlined.

I find this thread quite affronting and an insult to the hard work that Kestas has put into this site.

That said, a lot of effort is not a problem with the Elo system, it's actually quite simple to implement. The real con is the con of changing the whole system. Is it really worth it? Will the community make it through? I don't mean to be a downer but I would advise that advocates of the Elo system, including myself, educate more people on the subject before any changes are made.

Democracy can be a wonderful tool, but it's not the one for this job.

Really this should be less of a "voting" situation, and more of an organizing post, for people who might be interested in working together to implement some new ranking feature.

I mean, the code is there for you to look at and work on as much as you'd like, so if you want to try to work on a new feature, download the code and get working! Kestas wil have to decide once it's done and submitted if he wants to implement it on this particular site, but there is absolutely nothing to prevent you from hosting your own version on your own site that uses the ranking system you've worked out. That's the great thing about an open-source project!

Yup that's definitely an option, and thanks to the AGPL if it were to work out I could use the modified code back here, share-alike licenses work very well for this sort of thing :-)


Since talk about modifications to the Elo system are now in the voting thread this seems to have become another generic points-debate thread, so I'll throw in some thoughts and clarifications:

It's the points systems' first birthday this week (not sure what day exactly), and a year ago there was loads of civil disorders, individuals were playing in 40 games or more, the number of players was falling slowly, and the quality of gameplay was uniformly terrible because good players had no easy way to exclude novices. (It was the same on Facebook phpDip a few weeks ago)

The points system solved all those problems in one go, and more than any other feature it created the phpDip community as it is now. I don't want to risk undoing that, especially for the sake of something as subjective as a better way to rank skill.

A replacement system has to do these things:
- Seperate civil disorder players from players who rarely go into civil disorder
- Allow players of more skill to play with each other and exclude players of lesser skill
- Allow players to choose between playing many insignificant games, or few big games
- Don't allow a group of players to gain points just by playing amongst themselves
- Provide a way to rank players


As Darwyn and Chrispminis have said, the points system works, and it does all of the above. The only complaint is with the last criteria.
Last time I checked the Elo system I seem to remember it wasn't as good in some of the other respects, while being better at ranking, so it essentially is a different trade-off at best. (Also it wasn't fully developed when I last checked on it, it didn't account for draws etc.)

The other idea is having them both alongside, but then which system's points get displayed? Both? How does a user know which type of points are being bet? It seems rather messy to me.


I'd be interested to hear exactly what the current iteration of the Elo system for phpDip is, if it has been "fleshed out" since I last saw it, and how it would fit into/replace each of the roles the current system fills.

So much for this being a vote...

I vote to implement ELO alongside the points.

By the way the reason I bring up how bad things were before the points system is because I think people suggesting changes now are making suggestions from the perspective that the points system is just for ranking, and weren't around to see how bad things were before it was implemented.

If the points system was just for ranking I'd definitely merge in a patch for Elo, because I don't particularly care about ranking, it was more of a secondary feature of the points system. But it has to take care of much more than that, which makes altering it unsettling if you were around to see things before the points system was added.

@Kestasjk

Those are some very interesting observations, not having been around before the point system I can't speak to how terrible it was, but it sounds pretty bad, and it makes sense that it would have been, not insentive to not just give up on a game, and/or no reason not to join 50 games and just see if you could keep up.

I do think for the first bunch of criteria, the point system does a very good job, and for rank, a terrible job. Personally, I don't really care about rank all that much so for me it would be fine to leave the current system without anything else, but other people really seem to want to be able to have ranks, so my suggestion to them is to try to come up with and implement their own system, either to replace or to exist alongside the current point system (my personal preference would be to have it alongside the current point system, but maybe there is a better way). That being the case, someone who wants to do this should either just get started working, or should start a separate thread to discuss how to go about implementing these features, "voting" doesn't really make a lot of sense in this context.

So perhaps a good way to start a new thread, would be with a description of the basic Elo system, and then a question asking people how they would go about implementing it in the actual game, from both a design and engineering perspective.

sorry somewhere after the first paragraph or two that wasn't really directed at you Kestas anymore, and was more just a general statement to everyone.

Boy for someone who doesn't care about ranks all that much, I sure have talked a lot about them today... Think I'll be taking a break now...

#1, stick with current points system.

Zarathustra's proposal has some things to be said for it, but also a couple big disadvantages. The biggest disadvantage is that it serves to amplify point inflation by multiplying the two scales. (If the new rating based on points per win/draw/survival were divided by the number of games played, this problem would be much reduced because one's point multiplier would stay within the range 0 to 5 instead of increasing indefinitely.)

Anyway, Kestas rightly points out that the current ranking system is not just a way of determining player quality, but actually serves a variety of specific purposes. Before we vote on a new measurement system, it would probably be a good idea to agree on what it is that we really want to measure.

Well, to address the fact that the point system does a "terrible" job of representing rank, TheGhostmaker graphed the correlation between his Elo ratings and the point system ranks, and he found a very strong correlation, which shows that the problem demonstrated with the points system is mostly theoretical and doesn't happen too much in practice.

I've made 2 posts in the thread inflation that have a bearing on the current point system.

I agree with Kestas about the aims of the system, but would like to refine them:
Ranking is secondary to Kesta's other declared objectives of the points system, I believe. After all, you can look at a players profile an see what % of games they have won/drawn/survived/lost. Hell, you can even go and look at every move they did in every current game to see how they play. So if you want to know how good someone is, you can tell.

Having said that, kestas, can we please have the % be of finshed games only, so your percentages don't change when you start a game?

With regards to the other objectives, they mostly revolve around the problem of new players and established players.

New players are obviously the life blood of a site, we want to engage and get them interested, so they stay and make the place the busy place it is.

The problem is new players are often unreliable, they are often not challenging to play, often go CD, don't bother entering into the spirit of the game (not replying to messages or swearing at you for example).

We all want to avoid playing this sort of player and the system's main objective has to be:
1) to allow experienced, reliable players (whether they are good at the game or not) to play together and have enjoyable games without suffering the player described above too much.
2) To allow new and inexperienced players to play the game against similar people and give them a chance to prove to the system that they are reliable (and hopefully also skillful) players.

As someone that was a new player not long ago I can see both these objectives are as important as each other. I can see that the existing system does fulfill many of these criteria, it's just no very good at ranking.

If you are lucky (or skillful) enough, the system can work quite well for you - once you get a couple of hundred points you could choose to play in games without the newbies and have some challenging games against reliable, skillful players. Winning here gives you more points (higher bets) so good players climb away from the newbie pack.

Or, you could hang around the newbies, playing small bet games with new players, who often go CD, building up your points (and thus ranking) without actually playing anyone that good.

It is differentiating between these 2 types of players that the ELO system is good at but the current system is bad at.

However, what happens if you are a reliable player, that never goes CD, but you play your first 7 games and you haven't won one (or won one, but the bet was lower than the other games so you are still languishing low in the sub 100 zone.

You are stuck playing with the newbies and CD merchants. Are you going to feel enthused and carry on playing when every turn takes the full 24 (sometimes 72 with retreats, building etc)hours?

I don't think so. Now neither the current system or the ELO help you here. Perhaps we need to log when people go CD and allow players starting a game to set a CD threshold, the max % of games that someone can have gone CD and still play in your game. (Only players that have finished (not necessarily survived) a game or 2 can set a threshold). You can't set a threshold high enough to exclude yourself, and you can't join a game with threshold >0 until you have finished a game without going CD.

Only an idea. Could be completely separate from the rating system.
What do you think?

Oh. One more idea for an easy modification to the current system:
When players with 0 are topped up they only get 100 if they never went CD, otherwise the hand-out is capped.