Analyses of the ratings - Spotting the issues

Well, it still encourages or at least rewards toxic behaviour. Booming in a corner and insulting your ally until he quits, then using your lead to come back and win is a horrible thing to do. With your option 3, this strategy becomes surprisingly strong: Even with only a 33% winrate, you would still gain elo (as you gain double the points).

I think your problem is that you argue on a moral basis - “a player who resigns early does not DESERVE” instead of looking at how it will affect the game. But Elo is a matchmaking system. Not a tool for punishment.

1 Like

I think at how it will affect the game : if a player know he has not to resign to gain points, he will play until his team really loses, and i think early resign is a worse problem (at least in number) than what you talk about. The toxic behaviour you speak about has to be managed another way. You can’t build a good system with only the toxic behaviour in mind … And whatever system you create, there will be toxic behaviour …

But anyway, why don’t you just say you prefer choice 1 ?
You ONLY criticise, please feel free to suggest better ideas ! That’s not a weak defense … Just a right reaction to a weak offense …

But this is clearly false. If a player resigns, he thinks his team can’t win anymore. So your propsal does NOTHING to not make him resign.

1 Like

Is there a way to make players that have < 45% win rate lose more points and players > 70% win more?

I have returned to the game after a year hiatus and decided to play TGs to get back in shape… Its been a couple weeks now and I am still far away from “my ELO”. 2200 92% Win Rate, getting +9, +10 per victory… I believe I should be about 2600-2700. Its going to take another 30 games until I get more balanced matches.

And perhaps if people with < 45% WR lost more points, the system could adjust itself. At 2200 I keep playing people with 40-45% win rates that do not match the other players skill. In Nomad, just had a guy build the TC basically touching the water at a corner, getting to Feudal later than I got to Castle Age…

That’s what we are working on, and i think we have a solution. Just need the devs to read it.
At the moment, there is a big inflation over time on the team rating, which makes rating depend on the activity of the player (you’re in the case of a low activity, and then you are underrated), and the system i suggest solves it.

I think that people in the 45-50% range are at the correct elo. That means you are playing against evenly matched people. Your 92% win rate is the outlier in this. As you think you are just not to where you need to be.

Absolutely not. The ELO system should symetrically reward/punish players for winning/losing, so it doesnt create inflation. After so many games, the average player will have 50% win rate at the average rating (usually the starting one. i.e: 1000 ELO for 1v1), and you can then use the ELO to compare players.

Players with 55% WR after a significant number of games should only be found at the very top of the ladder. Players with 45% WR would be found at the very bottom.

The fact that the below happens, means the system is broken and these players are overrated.

Leaderboard Rank Rating Highest Rating Games Streak Wins Losses Win %
Team RM #471 2852 2874 893 -2 394 499 44%

After 900 games and a -100 games deficit, this ELO should be 49-51%. The fact that it is 44% means this player is facing stronger opponents frequently. How did he get +1841 points if he loses more than he wins? This player is clearly not at the same level as the other 2800 that have 50%+ WR and the fact that the system is not punishing when losing (but seems to be rewarding when winning) means even though he loses more than he wins (by a significant margin! 12%!!), he will continue to climb. Now we cannot compare players because the ELO is inflated and there is no baseline.

There are 76,447 players with an ELO of 1000+, and only 6,147 with sub 1000 ELO for TG. Basically if you play, you ramp up.

As for me… Yes, I am the outlier, too, because of this inflated system. I have to climb 1600 points so its going to take 100 games. On the 1v1 Ladder, as its not as inflated, I only had to climb 600 points (20-30 games).

3 Likes

2800 isnt the top for team games. Its in the 3400s. In this thread we have been talking about a way to fix the system and yes someone in the 45-55% is really close to where they need to be elo wise. It means your not dominant and not outmatched. If your lower you should be heading down the ladder to a more conformable place. There was 2 purposed fixes 2 different math formulas. The first one was the better overall idea. We are just hoping to get them implemented. If you have a super high win rate and a low elo it just sounds like you are facing people with a lower elo or you just haven’t hit your ceiling yet. Current system has 32 points available and it falls in a -200+200 window. Depending on the spread you get more for facing a higher team and less from a lower team. You also loss more from a lower team and less from a higher team. Since tg isn’t [2v2] [3v3] [4v4] its just [tg] there is a lot of variation. I can get a poor team mate and lose a few time in a row but then when my team gets together we stomp the comp.

If a player feels that they can no longer contribute to the game, they will surely resign. It is the job of their teammates to boost morale to not give up rather than the job of the rating system.

In my preference choice 1 is better option. The other choices have all sorts of unwanted behaviour like resigning just before a victory to avoid gaining elo. In voobly, you used to get more rating if one of your teammates resigned. This promoted the other kind of behaviour where you don’t care about your teammate resigning. (this can be fixed by reducing the total points gained for winning players instead of increasing it.)

I also like Mercy9545’s suggestion for Team Elo calculation better than gnarfk’s. They have better explanation for why they use it and I have some intuition that a log scale is a better way to represent player skill levels. On the other hand std. deviation feels completely unrelated. It kind of achieves the same goals but seems like a crude way to get things done. Also the higher rated players will usually have a bigger impact on the final result. Which team is more likely to win (1100, 2000) or (1300,1900). Often in such games, the result is entirely decided by the performance of the stronger players.

Good idea. Definitely a good idea. Sometimes can be impossible if you don’t have your own team. When pairing up with randoms, you might not even share a common language. Some players also just don’t pull their weight, and them resigning is not really a loss, but sometimes that player resigning can cause others in the team to ragequit or panic resign. These problems tend to be more relevant for low elos though.

So, very recently we surpassed the milestone of the playerbase having an average TG ELO of 1500.

That is an inflation of 500 ELO points in one and a half years. Since there are more games being played this year then at the start of DE we might be able to break the 2000ELO threshold by the end of the year.

3 Likes

For the love of God, fix this bug already

3 Likes

It’s not even a bug it’s just a horrible design of the rating system that they intentionally chose and will never be fix because they give no fucks if the ratings are worth less than toxic waste.

I literally didn’t know anyone was playing team games with randoms outside of the custom game lobby 11

45-48% is actually pretty low in the context of TGs given there is 1/2/3 other players (who should be on average, average) factoring in. [Illustrating this is the fact that there are only 6 civs with below 48% WRs in 1650+ TGs https://aoestats.io/stats/RM_TEAM/1650+ ) If you had 4x 45-48% players on a team together, their WR wouldn’t be 45-48%, it would be much, much lower.

Separate note:
Earlier in the thread I said “what is the harm in the inflation”? I know now that there definitely is major harm. First new but strong players (and smurfs) take way too long to get to their real Elo and 2nd, weak players (and rage quitters) get inflated, creating unbalanced games.

Inflation causes elo manipulation on different ladders cause of the initial boost, as it was mentioned several months ago.

Elo points distribution for multiplayer games gives usually way less points to prevent inflation, it was like that on zone, igzones,voobly and game park, idk why DE devs thought that giving 17 points each player for a 4x4 was a good idea, on 1x1 you can’t stack but on team games you can, manipulating once more the ladder.

1 Like

I really really really hope the devs will fix this issue in the next patch… It will also solve part of the Alt+F4 reasons from players.

@OMGBoxShark

3 Likes

Now we also get EW, it is really time that the devs fix this. Otherwise the EW 1v1 and TG ladders will start as a mess as well, like the current TG ladder. I posted the following in the announcement thread of ranked EW:

I would even advice to just fix the formula (this look easy, just replace the max by the average) and then reset all ladders. That is the most easy fix with the least effort for the dev team. If you dont reset the ladders, then TG ladder will still be a mess. And also the newly EW will become still a mess, since they will use the ratings of other ladders as starting point. Instead of sorting out their mess, the most easy solution will be a full reset of the ladders. But that might be a unpopular decision…

Edit: Looks like @ChristheCo is tracking the bugs. I see his name pop up in the bug report section on behave of the dev team. So i will tag him here too. Hopefully he can make sure this issue will also be tracked quickly. I started this thread 1 year ago, but it is still not part of the Known Issues & Solutions – Age of Empires Support section?!

4 Likes

Fix the formula for tg rating! Reset the ladder!

As long as this doesnt show up in known/tracked bugs I can only conclude that this forum has absolutely no purpose when it comes to reporting bugs. How can this be ignored for more than one year?
The only logic explanation is that the devs have no intention of actually fixing their game and this is just a place to confort the players.

2 Likes