Latest Posts

Topic: Rating system

trimard
Joined: 2009-03-05, 22:40
Posts: 195
Ranking
Widelands-Forum-Junkie
Location: Paris
Posted at: 2019-07-01, 19:43

That would allow for a permanent rating system, on the website, and bring a bit more competitivity in the game (outside the tournaments)

Why?

  • Motivation, one more game might mean you'll get ahead on the ladder for example
  • Less games where one player overwhelmingly lead the game.
  • Might push for more maturation of the meta of the game "yeah the best players do that and that", we already kind of have that, but it would be represented by more players than simply a few known names

Why not?

  • Motivation, some players could be discouraged by a low rating score. Though nobody would be force to participate of course
  • New feature: time needed, testing, more bugs.
  • Might not be applicable to other modes than 1vs1

How

After a bit of discussion with Worldsavior, he suggested using the Glicko system. The main advantage compared to something like ELO, is that Glicko tries to add variation for people that don't play a lot (which is pretty frequent in the widelands context compared to chess for example)

Glicko system

"The RD measures the accuracy of a player's rating, with one RD being equal to one standard deviation. For example, a player with a rating of 1500 and an RD of 50 has a real strength between 1400 and 1600 (two standard deviations from 1500) with 95% confidence. Twice the RD is added and subtracted from their rating to calculate this range. After a game, the amount the rating changes depends on the RD: the change is smaller when the player's RD is low (since their rating is already considered accurate), and also when their opponent's RD is high (since the opponent's true rating is not well known, so little information is being gained). The RD itself decreases after playing a game, but it will increase slowly over time of inactivity. "

How we could envision a rating system without changing anything in regard to game and server:

We create a new page, specifically for ranking, when you click on it the first thing you see is the current rank of everyone who's ranked. Preferentially in the playtime scheduler "index" page

If you want to be ranked too, there will be a "my rank" page accessible from there:

  • You click on a "get me ranked too" button
  • You play a game against someone who's also already ranked and lose/win
  • You and your opponent both confirm the score of the game in the page "name of player" "win?Loss?tie?"

If we want more automatically update of games

It has also appear the website and the game are starting to be really synchronized (with the password with actual security for example). So I think it's a good time to start and discuss such system. Maybe there could be a way for the website to know directly who you player against and the score of the game? That could pose some problems when the game bug and the game has be restarted?

Thoughts? face-smile.png


Top Quote
king_of_nowhere
Avatar
Joined: 2014-09-15, 18:35
Posts: 1423
Ranking
One Elder of Players
Posted at: 2019-07-01, 20:25

I thinking starting one myself (with elo, because it's what I know). problems are twofold: 1) most people play very little. 2) different win conditions make for very different games.

those two obstacles always stopped me


Top Quote
WorldSavior
Avatar
Joined: 2016-10-15, 04:10
Posts: 1246
Ranking
One Elder of Players
Location: GER
Posted at: 2019-07-01, 22:14

trimard wrote:

About a rating system

After a bit of discussion with Worldsavior, he suggested using the Glicko system.

If anyone is asking himself which version of Glicko was suggested: It was Glicko2. Might be not the perfect rating system for Widelands though; I'd be interested in suggestions for a better one.

If you want to be ranked too, there will be a "my rank" page accessible from there:

  • You click on a "get me ranked too" button
  • You play a game against someone who's also already ranked and lose/win

But I guess that if it will be good if both players have to agree before every match about that? It would be a nice additional feature if one could still play unranked matches.

One important thought about the ranking system: It should be forbidden to artificially decrease or increase the rating, and this behaviour should be prevented by punishments.

king_of_nowhere wrote:

I thinking starting one myself (with elo, because it's what I know).

You also know Glicko2 very well face-wink.png I think that Elo would be not so good, because it takes more time until the real ranking gets achieved... Maybe you can work with trimard together

problems are twofold: 1) most people play very little. 2) different win conditions make for very different games.

One could give every player one rating number for each win condition, similar to one chess website which uses one rating number for each chess variant.


“It's a threat to our planet to believe that someone else will save it.” - Robert Swan

Top Quote
king_of_nowhere
Avatar
Joined: 2014-09-15, 18:35
Posts: 1423
Ranking
One Elder of Players
Posted at: 2019-07-01, 23:44

WorldSavior wrote:

problems are twofold: 1) most people play very little. 2) different win conditions make for very different games.

One could give every player one rating number for each win condition, similar to one chess website which uses one rating number for each chess variant.

I also thought that. but seriously, there are few enough people that can be ranked for autocrat. how many games of wood gnome or territorial time are played?


Top Quote
trimard
Joined: 2009-03-05, 22:40
Posts: 195
Ranking
Widelands-Forum-Junkie
Location: Paris
Posted at: 2019-07-02, 00:26

Yes I agree with kind of nowhere on that point. We don't have enough players and games played of each type to do more than one ranking.

I think we should allow people to play any type of game, and that would count at the same weight. The principle isn't only to have a good rank, but to promote play. I feel like limiting player as to what they can't play, isn't really helping at that.

At the same time, the less rules we put, on which game can be played, the less representative it becomes...

What about actually two ranks? Competitive and general:

  • Competitive: A few selected 1vs1 maps. Changed pretty regularly. Only played with one predefined type of game (each map is particularly interesting for one specific mode generally)

  • General: Can play every official map as any type. Do we restrict to 1vs1 here too though?

No idea if we could have enough players. Though, we should keep in mind, such system has for main interest also to attract new players face-smile.png

@king_of_nowhere you think ELO is better?

But I guess that if it will be good if both players have to agree before every match about that? It would be a nice additional feature if one could still play unranked matches.

YES

One important thought about the ranking system: It should be forbidden to artificially decrease or increase the rating, and this behaviour should be prevented by punishments.

Huh, didn't think about that. How do you fear people will cheat? Fake games between themselves?

Huh, well that really brings the interest to an automatic system... But before that... How do you suggest punishments?

Edited: 2019-07-02, 00:29

Top Quote
einstein13
Avatar
Joined: 2013-07-29, 00:01
Posts: 1083
Ranking
One Elder of Players
Location: Poland
Posted at: 2019-07-02, 12:44

The rating system topic is going back again. Good. That means that we need that in the game.

Unfortunately we don't have any well known rating system for teams or multiple players at once. Maybe other games has this solved already? Recently I have seen that Age of Empires II game was restored (HD version) and has pretty good amount of plays every day. Maybe they did some ratings already? (They also are playing 1 vs 1 and teams)

Another thing is idea how to validate the results. There are few possibilities (that came to my mind):

  • External validation (administrator)
  • Self validation
  • Strict rules

First, we can send all replays to the server and an administrator will go through the replay and see who played and who has won. It takes lots of time, but for sure make cheating almost impossible.
Second, we can try to force people to send replays to the server and say who has won. If two people (winner and looser) answers the same, the system will accept the game. Unfortunately it is possible that a person who lost will not do anything. Maybe trying to validate the replay will help a bit? (first solution)
Third, we can make "multiplayer challenges" as another type of the game (or part of multiplayer game) that will be sent to the server automatic with replays, names and results.

The more we automate, the easier will be to cheat. But should we consider cheating in the first place? I am not saying "no" to this, but rather "let's try and see what is needed".

Edited: 2019-07-02, 12:44

einstein13
calculations & maps packages: http://wuatek.no-ip.org/~rak/widelands/

Top Quote
king_of_nowhere
Avatar
Joined: 2014-09-15, 18:35
Posts: 1423
Ranking
One Elder of Players
Posted at: 2019-07-02, 13:45

league of legends used a ranking (it still uses one, though it's hidden), and it is based on the team average, so it is possible to do it. it's going to be a problem if we have to adjudicate the ranking of new people entering in a team game, though.

I would avoid human intervention in checking eveyrthing. too complex, and too dependent upon an individual willing to spend time to do it. I also would not worry too much about cheating in this early stage, because frannkly, I don't think there's enough of a reason to cheat here. And if someone do try to cheat at their ranking, send them against worldsavior or me, it should rectify the matter face-smile.png

trimard wrote:

@king_of_nowhere you think ELO is better?

I don't know the alternative system, so I cannot judge.


Top Quote
trimard
Joined: 2009-03-05, 22:40
Posts: 195
Ranking
Widelands-Forum-Junkie
Location: Paris
Posted at: 2019-07-02, 14:15

Thanks to Einstein13 I realized there was already a topic on implementing such system from... 2010.

A few things to get from there:

  • The was an attempt at making point in function of how the game was played, ware produced, strongest army etc. I personally am against such system, because it would take a really long time to equilibrate and might need to be changed after every new build... Moreover my most recent games with the-x clearly showed that army level and winning aren't always the same face-tongue.png
  • Creating a hash, per game, so that one game can't be submitted more than once, and thus allow for taking into account game that needed to be loaded again (for bug or other)
  • There was already this idea of having more than one ladder. Notably for 1vs1, 2vs2, 3vs3, ...
  • wtf is GGZ? What is the currently status about that?
  • Apparently there was already a lot of data exchanged with the server after each game. Is that still the case?

@Einstein13

External validation (administrator)

Might take a lot of time from that administrator. But yeah that avoid problems

Self validation

I think that's much better, with an arbiter looking at the replay if a problem arise.

What about this: each game have a hash. We give one way or another for a player to get the number of that hash.

Now after you played the game, you send to the server: Hash + result of your game.

If it appears there is a problem with that game. Players are invited to send the replays to the arbiter. This one first check that hash correspond to game and then go on to judge about the problem.

Strict rules

TBH, I prefer this idea even more. But how do you think it can be abused?

@king_of_nowhere

league of legends used a ranking (it still uses one, though it's hidden), and it is based on the team average, so it is possible to do it. it's going to be a problem if we have to adjudicate the ranking of new people entering in a team game, though.

Yes, but there is no 1vs1 AFAIK? Or do they have different ranking in function of the mode in which you're playing 1vs1 or 5vs5?

I would avoid human intervention in checking eveyrthing. too complex, and too dependent upon an individual willing to spend time to do it.

Yes, but I think as a first stage and prove of concept it would be pretty useful!

also would not worry too much about cheating in this early stage, because frannkly, I don't think there's enough of a reason to cheat here.

Yes! Even in the tournament I think you said you never actually needed an external arbiter. And we can still implement anti-cheating system afterward

Edit: If nobody has a problem with Glicko2, I think we should start with that system.

Edited: 2019-07-02, 14:16

Top Quote
kaputtnik
Avatar
Joined: 2013-02-18, 20:48
Posts: 1762
Ranking
One Elder of Players
Location: Germany
Posted at: 2019-07-02, 17:19

trimard wrote:

  • wtf is GGZ? What is the currently status about that?

face-grin.png As far i know GGZ was an external service for online gaming with the ability to rank players. This was used by widelands for some time (don't know how long), but the ggz service was discontinued. No idea why. The only resource i found is an old manpage. All other links seems to be dead...


Top Quote
WorldSavior
Avatar
Joined: 2016-10-15, 04:10
Posts: 1246
Ranking
One Elder of Players
Location: GER
Posted at: 2019-07-02, 21:17

king_of_nowhere wrote:

WorldSavior wrote:

problems are twofold: 1) most people play very little. 2) different win conditions make for very different games.

One could give every player one rating number for each win condition, similar to one chess website which uses one rating number for each chess variant.

I also thought that. but seriously, there are few enough people that can be ranked for autocrat. how many games of wood gnome or territorial time are played?

Not many games, as far as I know. But different rating numbers could encourage to play even more, like "I'm not so good at autocrat, so I'm trying to become highly ranked in collectors".

trimard wrote:

I think we should allow people to play any type of game, and that would count at the same weight.

Even on imbalanced or inofficial maps? I think that they shouldn't become rated. If you really want you could start with only one rating, and as soon as it seems pointful to create a second rating number for another win condition one could still split that.

At the same time, the less rules we put, on which game can be played, the less representative it becomes...

Yes, that could be a problem.

What about actually two ranks? Competitive and general:

  • Competitive: A few selected 1vs1 maps. Changed pretty regularly.

Why changed?

  • General: Can play every official map as any type. Do we restrict to 1vs1 here too though?

I don't know how to apply a 1vs1-rating in a match with more players. A possibility would be to use an own rating number for 2vs2 and maybe another one for 3vs3 and so forth face-wink.png

No idea if we could have enough players.

The rating system could encourage more playing. And if tournaments get ranked, it's already a good number of games ...

One important thought about the ranking system: It should be forbidden to artificially decrease or increase the rating, and this behaviour should be prevented by punishments.

Huh, didn't think about that. How do you fear people will cheat? Fake games between themselves?

Yes, if somebody lets the other always win and the winner continues with getting this free wins it's cheating from both sides.

Huh, well that really brings the interest to an automatic system... But before that... How do you suggest punishments?

First of all the changes of rating points should get reverted. Other punishments would be temporarily bans or in really bad cases permanent bans...

Before I forget it: It can also happen that somebody who is about to lose quits the match and never continues it. After a while this should be considered as loss...

einstein13 wrote:

Maybe other games has this solved already? Recently I have seen that Age of Empires II game was restored (HD version) and has pretty good amount of plays every day. Maybe they did some ratings already? (They also are playing 1 vs 1 and teams)

Age of Mythology used a rating system which also worked for 2vs2, 3vs3 and so forth, this is really not rare. But I don't know how it works...

Unfortunately it is possible that a person who lost will not do anything. Maybe trying to validate the replay will help a bit? (first solution)

Yes, replays are very useful if someone refuses to answer...

But should we consider cheating in the first place? I am not saying "no" to this, but rather "let's try and see what is needed".

It may be possible to implement the rating system first and to implement anti-cheat-measures afterwards.

king_of_nowhere wrote:

And if someone do try to cheat at their ranking, send them against worldsavior or me, it should rectify the matter face-smile.png

Lol

trimard wrote:

@king_of_nowhere you think ELO is better?

I don't know the alternative system, so I cannot judge.

So you meant that you don't know it well enough, I see...

trimard wrote:

@king_of_nowhere

I would avoid human intervention in checking eveyrthing. too complex, and too dependent upon an individual willing to spend time to do it.

Yes, but I think as a first stage and prove of concept it would be pretty useful!

yes


“It's a threat to our planet to believe that someone else will save it.” - Robert Swan

Top Quote