MMR in a Nutshell

[Cape]

Its a work in progress

Abusing mods and people rage quitting will always occur in some percent of games.

[beam]

That is even worse. Leaving it up to the host to decide game modes is easily abusable. I've seen beam abuse like 3 times already. Aggi and ogron are constantly rage quitting games ruining ratings for people without consequence. Honestly MMR is turning into a mess it didn't need to be. Keep it pure and simple and collect and compile the true data for analysis. As it is now the low tier games are wildly unbalanced while certain players who are not very good are getting carried to maintain their ratings who are currently higher rated.

What rating setting is the bot using to decide lines for the twdd3 lines it is making? Do hosts get to decide the rating specifics there too?

Make MMR pure again. It serves no purpose segregating games other than to dilute ELO info which is breaking the entire point of this...

Aggi and Ogron were both banned for 24 hours from MMR for ragequitting their games just like you were banned 24 hours for ragequitting your game last night. Without subs being available and the inability to add players after the game starts, there's no option other than to kill a game or let it finish 3v4/4v5 if a player leaves early or specs on go. This is something qan is working on, but as the game mode is still in a testing period, it is not perfect.

As far as the MTT command goes, the general opinion seems to be that it is much better overall. If you feel as though you are too good for the matches in dd4/jd4, then you should be able to win them consistently enough that you climb to a rating in which you get matched to the games in dd3/jd3. MMR is based on an elo system and elo systems are designed for you to face similar skilled competition. The system was not designed so 1000 rated players could get matched with 5000 rated players. The players in both games should be much closer in skill and rating than they were in the original matchmaking system.

[Rab]

You realise that a 24 hour ban isn't really a punishment right? Try 24 days and people might care.

[beam]

You realise that a 24 hour ban isn't really a punishment right? Try 24 days and people might care.

It's basically a warning and the next time will be more severely punished.

[LeeRomeno]

My 2 cents on this. The so called JR arena games are quite often decided by 1 high star player, who somehow also ends up thrown in the mix. For example in jav, there are 7 average joes and then 1 freq has roiwerk or dreamwin or ease or vys or someone else and that freq wins 95% of times, even if other freq has slightly better newbies than other freq. So in reality the MMR rating of the newbies is purely based on if they are lucky enough to be on that freq. In that scenario, mixing 2 arenas with equal freqs makes more sense.
At the same time if that is not the case, i can see why its preferred mode and definitely the main arena game is better in both ways.

[beam]

My 2 cents on this. The so called JR arena games are quite often decided by 1 high star player, who somehow also ends up thrown in the mix. For example in jav, there are 7 average joes and then 1 freq has roiwerk or dreamwin or ease or vys or someone else and that freq wins 95% of times, even if other freq has slightly better newbies than other freq. So in reality the MMR rating of the newbies is purely based on if they are lucky enough to be on that freq. In that scenario, mixing 2 arenas with equal freqs makes more sense.
At the same time if that is not the case, i can see why its preferred mode and definitely the main arena game is better in both ways.

If a player is good enough to win those games 95% of the time, they will climb and not be in those games for very long. This is a disadvantage of everybody starting at 3k and should auto correct itself over time with enough games played.

[LeeRomeno]

If a player is good enough to win those games 95% of the time, they will climb and not be in those games for very long. This is a disadvantage of everybody starting at 3k and should auto correct itself over time with enough games played.

True, but lets say u have 9 top players with !p, they all cannot fit if there are 2 4v4 games. This happens more often than not, but of course over the time and if everyone involved has at least 50-60 games, it should get better.

[Rab]

Lee is right, newbie games are almost entirely determined by draft and not by performance.
The main game is better though because they don't have to deal with newbies, pick your poison.

[Riverside]

A potential use-case that i dislike with the MTT command is that I assume its been used when there's only one game in queue (ie 12 players).

I dont think the purpose of this command is to disclude lower rated players. I believe that is how it has been used at times.

When this was pointed out to the host, they said that this would be their last game hosting mmr for the day as "people were whining too much".

Lol

[beam]

A potential use-case that i dislike with the MTT command is that I assume its been used when there's only one game in queue (ie 12 players).

I dont think the purpose of this command is to disclude lower rated players. I believe that is how it has been used at times.

When this was pointed out to the host, they said that this would be their last game hosting mmr for the day as "people were whining too much".

Lol

qan is also working on this so it will see who was benched in the previous game and try to prevent them from being benched in consecutive matches. Currently the possibility exists for the same player to get benched every time using both commands as the same formula is being used every time and there's nothing preventing it from happening.

[Riverside]

qan is also working on this so it will see who was benched in the previous game and try to prevent them from being benched in consecutive matches. Currently the possibility exists for the same player to get benched every time using both commands as the same formula is being used every time and there's nothing preventing it from happening.

That's great.

I dont think its helpful to use the mtt command when there's one game (ie 12 players).

[Jessup]

Aggi and Ogron were both banned for 24 hours from MMR for ragequitting their games just like you were banned 24 hours for ragequitting your game last night. Without subs being available and the inability to add players after the game starts, there's no option other than to kill a game or let it finish 3v4/4v5 if a player leaves early or specs on go. This is something qan is working on, but as the game mode is still in a testing period, it is not perfect.

As far as the MTT command goes, the general opinion seems to be that it is much better overall. If you feel as though you are too good for the matches in dd4/jd4, then you should be able to win them consistently enough that you climb to a rating in which you get matched to the games in dd3/jd3. MMR is based on an elo system and elo systems are designed for you to face similar skilled competition. The system was not designed so 1000 rated players could get matched with 5000 rated players. The players in both games should be much closer in skill and rating than they were in the original matchmaking system.

I didn't rage quit any game you corrupt liar, wtf are you talking about? I wasn't even added to any game. You had just killed a game I was in that had started and I logged off out of disgust because I've seen you do this type abuse multiple times when either you don't like the team you have been put on or one of your buddies on pandora chat complains. More staff abuse by beam. Show me my "ragequit" during a game. It didn't happen and you are abusing again, you lying, corrupt, troll, awful staffer.

You are also being idiotic in your defense of destroying the ZONE ELO ratings that we all have worked to compile for weeks now which you and your buddies destroyed by abuse this last week by secretly changing the system up to where you and your buddies who control hosting decided to implement this segregated system. ELO can't work without the games compiling EVERYONES ratings evenly. MMR was an ELO system meant for EVERYONE to play together to compile data not for just you and your troll Pandora twits who got mad because their ratings were going down and they couldn't handle playing with lower tier players in a game. You guys by doing this literally sabotaged and ruined MMR and it now needs to be reset due to completely corrupted data.

[beam]

I didn't rage quit any game you corrupt liar, wtf are you talking about? I wasn't even added to any game. You had just killed a game I was in that had started and I logged off out of disgust because I've seen you do this type abuse multiple times when either you don't like the team you have been put on or one of your buddies on pandora chat complains. More staff abuse by beam. Show me my "ragequit" during a game. It didn't happen and you are abusing again, you lying, corrupt, troll, awful staffer.

You are also being idiotic in your defense of destroying the ZONE ELO ratings that we all have worked to compile for weeks now which you and your buddies destroyed by abuse this last week by secretly changing the system up to where you and your buddies who control hosting decided to implement this segregated system. ELO can't work without the games compiling EVERYONES ratings evenly. MMR was an ELO system meant for EVERYONE to play together to compile data not for just you and your troll Pandora twits who got mad because their ratings were going down and they couldn't handle playing with lower tier players in a game. You guys by doing this literally sabotaged and ruined MMR and it now needs to be reset due to completely corrupted data.

I'm not sure you realize how you sound so let me give you an example of how you are coming across.

A brand new player named "Jesse" starts playing chess after watching Queen's Gambit. Jesse plays for 100 hours on chess.com practicing hard day and night. Months pass by and Jesse has managed to achieve an elo rating of 635 after 1300 matches, which is a fairly average rating for a new player. Jesse finds out about a tournament where he can compete on the website to test his ability. Chess.com creates various brackets for the tournament for players to compete in to make it more fair and fun for each player. As Jesse is rated 635, he is assigned to the under 800 elo bracket wherein he will only face off against other new/casual players below 800 rating. Once Jesse finds out about this, he petitions to Chess.com that he is being discriminated against because he has just as much of a right to face the grandmasters as everyone else. After reviewing the inquiry from Jesse, a representative of Chess.com responds that it would not be fair or fun for either player to compete against each other in the tournament as the grandmaster would obviously win easily and that the brackets exist to make the overall experience more fair and enjoyable for all parties involved. Jesse then responds by saying that he is being "segregated" and refuses to compete in the tournament because he believes that if only he got a chance to face those grandmaster players that he would not be 635 rating and could take all of their points and be grandmaster himself, but the website is rigged to keep him at his low rating by depriving him of that opportunity.

TLDR: you are in the low tier games because you are a low tier player and those guys are similar skill/impact as you.

Also I log everything so it doesn't do you any good to lie. Take responsibility and don't do it again

[Jessup]

I'm not sure you realize how you sound so let me give you an example of how you are coming across.

A brand new player named "Jesse" starts playing chess after watching Queen's Gambit. Jesse plays for 100 hours on chess.com practicing hard day and night. Months pass by and Jesse has managed to achieve an elo rating of 635 after 1300 matches, which is a fairly average rating for a new player. Jesse finds out about a tournament where he can compete on the website to test his ability. Chess.com creates various brackets for the tournament for players to compete in to make it more fair and fun for each player. As Jesse is rated 635, he is assigned to the under 800 elo bracket wherein he will only face off against other new/casual players below 800 rating. Once Jesse finds out about this, he petitions to Chess.com that he is being discriminated against because he has just as much of a right to face the grandmasters as everyone else. After reviewing the inquiry from Jesse, a representative of Chess.com responds that it would not be fair or fun for either player to compete against each other in the tournament as the grandmaster would obviously win easily and that the brackets exist to make the overall experience more fair and enjoyable for all parties involved. Jesse then responds by saying that he is being "segregated" and refuses to compete in the tournament because he believes that if only he got a chance to face those grandmaster players that he would not be 635 rating and could take all of their points and be grandmaster himself, but the website is rigged to keep him at his low rating by depriving him of that opportunity.

TLDR: you are in the low tier games because you are a low tier player and those guys are similar skill/impact as you.

Also I log everything so it doesn't do you any good to lie. Take responsibility and don't do it again

Show the log of me rage quitting a game. Show the game you killed before too that and explain why you did that and which player " rage quit" that one and their ban. You are corrupt to the core and it is time you take responsibility for these selective illegal actions you constantly do. There was no ban on ogron and greed as both played yesterday too so you are again lying.

As far as ratings go, I was here in last 2 weeks, Then Greed threw every game he had with me or did rage quits and I also had gen x going 3-10 every game on me team too so I drastically fell due to bs trolling by you guys. You are so corrupt beam.. what a disgrace .

[qan]

The ability to use !mtt when there are only enough players for a single game is a bug. It will now default to the standard algorithm when there's only one team pair to be made. (From looking at the code, it shouldn't affect things substantially, but it does seem to.)

Tiered games don't cause corruption of ratings data. The point of this sort of system is to allow any team to play any other team and still have ratings adjust accordingly, based on statistical modeling.

There are cases in other games where a player can be "carried" by always playing with more-skilled teammates. Remember that ratings are simply an estimation of the average performance of a player, though. If a player always squads up, their average performance is literally how they perform inside that squad. It might look like inflation, but that's misunderstanding how ratings work: it's actually quite accurate. If they then start solo-queuing, that rating isn't a good measurement of their solo performance. They'll probably experience a string of losses that will cause their rating to slide.

After some code review, when not making tiered teams, bench code should still be active. Players shouldn't be benched in back-to-back games of the same league type, in most cases. Please let me know if this is happening often in non-tiered games. I'm working on a fix for tiered, and have added a tweak that should make it less likely.

In tiered games, I found out why sometimes lower-rated players are added to the higher tier. This has to do with "playability" ratings. "Pre"-benched players (for the current match) are still included in the player pool used to find the optimum balance. I need to refactor this and have written down some notes about how to do this. The tiered algorithm was an add-on that doesn't always play nice with some of the others, basically.

As always, if you have any reports of staff abuse, please send them to twcomplaints [at] gmail [dot] com.

We haven't really set up regulations for ragequitters, but as it entirely ruins a match with the current state of the code, I'd recommend starting bans on at least a week. It can't become a pattern.