Ratings Disputes and Offence Reporting (Discussion)

[HMS-Surprise]

so make me mod then.

@elexis

@Hannibal_Barca

Edited July 1, 2018 by HMS-Surprise

[elexis]

9 hours ago, (-_-) said:

How would that handle host side drops and player kicks?

That's why WFG should host these games, it can reject kicks.

If someone joins a game and leaves it for network issues or bugs, then it's to his disadvantage. There may be a 2-5 minute limit for rejoins and a 2-5min window for abandoning a rated game after the start.

4 hours ago, (-_-) said:

it may still be worth to patch up the obvious flaws and exploits

Such as? I believe it would only change the way it would be exploited, for instance using the taskmanager to kill the process or pulling the ethernet wire.

3 hours ago, Hannibal_Barca said:

after rigourous questioning

If it needs questioning, then we didn't have a solid impression yet.

3 hours ago, Hannibal_Barca said:

short-staffed

To moderate foul language it's ok currently. Anyone claiming to be able to handle rating fakery is lying. I'd like to invite new moderators too, but it's hard to find someone who is trolling rarely and actually helping. Each time I saw a candidate that I'd endorse recently, he does something that he should have punished others for.

3 hours ago, coworotel said:

 Yeah I think so far few people would go through the trouble of faking the information sent to the servers just to improve ratings.

It depends on how hard it would be to fake a rating. Currently the rating is changed if both clients agree that one of them lost. So if one rating report differs or isn't sent, nothing is changed.

If we trust one of the two players, someone can use this to create arbitrary amounts of fake rating and remove points from a selected player, making it effectively worse than now.

There was also the proposal to upload replays, then moderators would have to watch the replay to decide. This is easy to implement, but is extremely timeconsuming to moderate.

We need the code for dedicated servers, players setting up the game anyway regardless of rating fakery, so might just as well stop postponing this for more years to come.

[wowgetoffyourcellphone]

I've always wanted 0 A.D. to have a dedicated server to host games.The benefits seem overwhelming to me. To see a WFG programmer say the same is definitely gratifying, even though I realize the rewrite would be immense. I have no doubt WFG could "afford" such a server. It's all about the time and effort it takes to do it.

[elexis]

49 minutes ago, wowgetoffyourcellphone said:

The benefits seem overwhelming to me.

The dedicated host feature itself should be possible because players want to host for each other anyhow.

But which other overwhelming benefits do you see?

It also has significant disadvantage for WFG to host games. For instance I see it as a benefit that WFG can't look into usergames, it should be players business if they decide to fight each other and I'm lucky that we can't moderate it unless it leaks into lobby chat. But if we host things, we have to host userchat too, need to change the policy.

Second disadvantage is that it costs more effort to setup a custom lobby. In the worst case it costs lots of CPU power (i.e. actual money in some orders of magnitude, possibly even multiple computers) if we are forced to simulate the game too (and I think we might have to if we want to know the winner of the match). Also it prevents rated modded games unless we installed the mod on the host then.

So I want the dedicated server feature anyhow (I wrote a half finished patch, one that added a chat hack to allow players to setup a game in 2015), But it would still be better to not have WFG have to have to host rated games, but I didn't see an alternative yet.

[wowgetoffyourcellphone]

I am not sure why you think WFG would have to moderate user chat for anything except bullying. Can you clarify that point? 

The chief benefit I see for most players is this:

18 minutes ago, elexis said:

if we are forced to simulate the game too (and I think we might have to if we want to know the winner of the match)

A central host for less lag and a more stable connection. No need for host migration or any of that. P2P is bad bad bad for stability and lag. A central server could help a lot in this regard. Unless I am not understanding things correctly, which is possible of course.

 

20 minutes ago, elexis said:

Also it prevents rated modded games unless we installed the mod on the host then.

I'd say as long as the mod has been verified and hosted on the mod.io site, then it should be okay to install on the server too after some more thorough checks of the simulation files?

[Stan`]

Is it sane to use the same rating for every mod combination ?

[elexis]

47 minutes ago, wowgetoffyourcellphone said:

A central host for less lag and a more stable connection. No need for host migration or any of that. P2P is bad bad bad for stability and lag. A central server could help a lot in this regard. Unless I am not understanding things correctly, which is possible of course.

If one client (player or observer) is lagging, all of them are. The game can only progress at the speed of the slowest simulating or laggiest one. So a WFG hosted server for unrated games only had the advantage that the host can't alt+f4 to kill the game anymore.

Notice that it's easily possible that WFG hosts games but doesn't simulate them (but then we don't know the winner of the match unless trusting player consensus which might not be given).

WFG hosting everything also means that this service being slow or down affects all matches.

((Still we want the dedicated server ability, but the main question is the feasibility and need for WFG to host rated matches. If there is any effective alternative, we should consider it.)

5 minutes ago, stanislas69 said:

Is it sane to use the same rating for every mod combination ?

Policy question effectively. Saneness of playing on scenario or random maps can also be questioned (survival of the fittest, demo maps)

[Hannibal_Barca]

21 hours ago, HMS-Surprise said:

so make me mod then

Wrong attitude and incorrect form of application 

 

21 hours ago, elexis said:

needs questioning

rigorous questioning = why do you want to become a mod

From that question all the other sanity-related questions are derived.

 

21 hours ago, elexis said:

like to invite new moderators too, but it's hard to find someone who is trolling rarely and actually helping.

At least the current team is a bunch of angels

 

21 hours ago, elexis said:

he does something that he should have punished others for.

This is a good idea - do something bad and then ban others because of it

 

 

On 6/30/2018 at 2:26 PM, coworotel said:

Hi, today I learned a new strategy to get points without winning: user GVIT was losing the game, then he just kicked me. I asked him to create another match and resign but he just kicked me again.

 

By the way, you don't get points by defeating an offline player.

[elexis]

1 hour ago, wowgetoffyourcellphone said:

I am not sure why you think WFG would have to moderate user chat for anything except bullying. Can you clarify that point? 

Currently it's technically impossible for us to moderate ingame chat because we don't have access to the data.

If we process userchat, we have to add it to the privacy policy and then some people will complain that there was bullying or anything else in userchat hosted in our game and then we have to moderate that too or justify ourselves each time it is reported.

I guess if there was WFG hosting we could still only relay the chat and not store it, thus continuing to have it impossible to moderate that. (Seriously we have more important things to do than kindergardening.)

[Hannibal_Barca]

Theoretically as lobby moderators we have no right, nor any intention to moderate in-game chat.

While users are subject to our rules while in the multiplayer lobby, in-game the host is all-powerful and reserves the right to do anything . The exceptions to the rule are rated games and rated games only.

(Game names are moderated as they appear in the lobby game list)

Any in-game law enforcement is to come from persons other than lobby moderators, perhaps some Team Members would like to oversee complete chat moderation.

Edited July 2, 2018 by Hannibal_Barca

[Lion.Kanzen]

On ‎7‎/‎1‎/‎2018 at 4:30 AM, HMS-Surprise said:

so make me mod then.

@elexis

@Hannibal_Barca

[Sundiata]

Spoiler

 

[HMS-Surprise]

Spoiler

 

 

[Imarok]

On 7/2/2018 at 10:02 AM, elexis said:

Notice that it's easily possible that WFG hosts games but doesn't simulate them (but then we don't know the winner of the match unless trusting player consensus which might not be given).

As already said in some other places, there is imho a better solution:

Wfg hosts rated games without simulating it but stores the replay.

If the players send the same result it is counted.

Only if the results differ, which should happen quite rarely, we simulate the stored replay (automatically or manually) to get the real result. Of course we should also admonish the player that sent the false report and ban/kick him/her if it happens multiple times.

[elexis]

3 minutes ago, Imarok said:

As already said in some other places, there is imho a better solution:

Wfg hosts rated games without simulating it but stores the replay. 

You are right; simulation can be omitted in the case of players remaining online and agreeing with the rating and that this is more performant than always simulating.

Simulating only after the game ended would mean that the rating is adapted hours after the rating changed, so it would be necessary that this only a rare occasion.

4 minutes ago, Imarok said:

which should happen quite rarely

Sure we can't construct some exploit? I guess you are right, but we must prove we didn't forget some case.

Is someone is changing the code (to fake rating or by having some semi-legit GUI mod that removes mod version checks) the only case in which one'd have to simulate?

If both players disconnect before the disconnect clause reduces the rating, then the player that disconnected first gets the rating reduction, hence not having to simulate in that case?

Independent of your improvement, what do we do if there is an OOS error? The server can detect it, but it can't determine if the rejoiner computed the wrong state due to a code error or abuse.

[Stan`]

9 minutes ago, elexis said:

You are right; simulation can be omitted in the case of players remaining online and agreeing with the rating and that this is more performant than always simulating.

Simulating only after the game ended would mean that the rating is adapted hours after the rating changed, so it would be necessary that this only a rare occasion.

Sure we can't construct some exploit? I guess you are right, but we must prove we didn't forget some case.

Is someone is changing the code (to fake rating or by having some semi-legit GUI mod that removes mod version checks) the only case in which one'd have to simulate?

If both players disconnect before the disconnect clause reduces the rating, then the player that disconnected first gets the rating reduction, hence not having to simulate in that case?

Independent of your improvement, what do we do if there is an OOS error? The server can detect it, but it can't determine if the rejoiner computed the wrong state due to a code error or abuse.

We could hash mods. But of course if the guy recompiles with the right hash and always send that there will be abuse.

[Imarok]

5 minutes ago, elexis said:

Sure we can't construct some exploit? I guess you are right, but we must prove we didn't forget some case.

Sure you can ofc fake the report, but then we'll get a report and if someone fakes multiple times, we punish that to disencourage such behavior.

6 minutes ago, elexis said:

Independent of your improvement, what do we do if there is an OOS error? The server can detect it, but it can't determine if the rejoiner computed the wrong state due to a code error or abuse.

As nowadays OOS are quite rare, I think we just handle that also as faking, unless the one who "faked" provides his replay and we can debug it. If so we must decide on each case.

8 minutes ago, elexis said:

so it would be necessary that this only a rare occasion.

Don't know why this should happen often...

9 minutes ago, elexis said:

If both players disconnect before the disconnect clause reduces the rating, then the player that disconnected first gets the rating reduction, hence not having to simulate in that case?

Definitely no need to simulate then anyway, but I'd say if both players disconnect nearly simultaneously, we just do nothing, so no rating for nobody.

[elexis]

1 minute ago, Imarok said:

As nowadays OOS are quite rare

The point I was trying to make is that we not only have to consider OOS that are bugs on our end but also the possibility of players utilizing it as an exploit in order to not get a rating change.

3 minutes ago, Imarok said:

Don't know why this should happen often...

We must know that it cannot happen often before we rely on that fact. I also have the impression that we'd only have to simulate very few games with your proposal (but we must be certain).

4 minutes ago, Imarok said:

if both players disconnect nearly simultaneously, we just do nothing, so no rating for nobody. 

Maybe, but by definition it was a rated game, so it were true to the name if it would apply a rating in any case.

If we don't apply a rating in some cases, we must check for and rule out exploitability. In this case the player would have to know how long he would have to remain connected to the game just to get the rating of the leaver. For instance if we leave a 5 minute margin before marking the disconnected player as a loser, then the one sitting in the game must be aware when the 5 minutes started and how long he has to sit there.

(Another edge case: What if both players have computed the same gameresult but the report wasn't sent quickly enough before the disconnect? I guess that's also rare enough, especially if the "you lost/won, leave?" mesagebox would only come after the report was sent and received.)

[Imarok]

22 minutes ago, elexis said:

Another edge case: What if both players have computed the same gameresult but the report wasn't sent quickly enough before the disconnect? I guess that's also rare enough, especially if the "you lost/won, leave?" mesagebox would only come after the report was sent and received.)

Sure that needs investigation. I thought the server knows, when a game ended, but after thinking about, I'm not sure...

One possibility could be that the clients get some rating accepted (rating not accepted) message by the lobby bot. Then we could pop up a message box in case a player wants to quit a game before the rating has been accepted.

[thankforpie]

afaik if u finish him then u get points even if he left

[Boudica]

23 hours ago, thankforpie said:

afaik if u finish him then u get points even if he left

That's false.

[Thorrky]

I played "huabidbs" earlier today and he disconnected the game after clearly losing.

(is this the correct format for this - what information can I give?)

Thanks,

Thorrky

@user1

[Boudica]

I think that it's usually required to attach the replay files as a proof. Each game would usually generate a commands.txt and metadata.json files inside a new subdirectory, which you'd optionally zip and mandatorily attach with your post. See https://trac.wildfiregames.com/wiki/GameDataPaths for replay file location.

[thankforpie]

11 hours ago, Thorrky said:

Additionally, the same occurred with an 'Arthus' today, replay file attached.

metadata.json

commands.txt

ye people are so scared of losing 11 points that they will act like that lol it kills the mood and annoys really, especially when they try to run with their last woman or cav for 10min trying to make you chase it lol or just leaving/afking

 

it would be good if leaver lost automatically (same about afkers)

Edited September 25, 2018 by thankforpie

[Jofursloft]

@user1