alert notification image

Age of Empires IV launches October 28th, available for Pre-order Now!

Hello Community!

A major part of online play is online matchmaking. How the game decides who plays together can make or break the fun and challenging experience Age of Empires II: Definitive Edition players have come to expect from competitive matches.

That’s why today we’re making some changes to the way ranked ELO works in Team Games (TG) in Age II: DE. This change is specific to how ELO is awarded after team games, but we want to be clear: this is just the first of several changes we’re investigating to improve the competitive experience. Your feedback, as always, is vital, and we will continue to listen to your voice on our forums, Discord, and social channels to find new ways to give you great matches.

We’ve been tracking the issues you’re seeing in imbalanced Ranked Team Games, which has its root in how ELO wins/losses are currently calculated and distributed. The system can get pretty complicated, and an undesired ELO distribution can be more or less of a problem depending on the makeup in a given match. Part of how we wanted to address this was by simplifying both the solution and the conversation– here’s a simplified summary:

Ranked RM TG ELO – HERE’S HOW IT WORKS ALREADY:

Currently, if a team has a wide range of skill levels between players, the system awards ELO ranks based on the highest-level opponent on the other team. In the example below, Team 1 (with a wide ELO spread: one player at 2000 and the other at 1000) matches up against Team 2 (who have similar ELO to each other). There are several ways this could play out:

  • Team 1:
    • Player-A at 2000 ELO
    • Player-B at 1000 ELO
  • Team 2:
    • Player-C at 1500 ELO
    • Player-D at 1500 ELO
  • If Team 1 wins:
    • Player-A gains 1 ELO (as if a 2000 ELO player beat a 1.5k ELO player)
    • Player-B gains 100 ELO (as if a 1000 ELO player beat a 1.5k ELO player on their own)
    • Team 2’s loss is calculated against Player-A only, so both players lose only 1 ELO (as if losing to a 2000 ELO player)
  • If Team 2 wins:
    • Player-A loses 100 ELO (as if a 2000 ELO player loses to a 1500 ELO player)
    • Player-B loses 1 ELO (as if a 1000 ELO player loses to a 1500 ELO player)
    • Team 2’s win is calculated against Player A only, so both players gain 100 ELO (as if they’d each beaten a 2000 ELO player by themselves.)

In this example, a win for Team 2 will cause their ELO to rise more than it should… that might sound great for players who are trying to rush to an ELO of 2000, but in practice this means that Team 2 will face much tougher matchups than they should at their skill level, meaning they can look forward to a lot of losses in their future! If Team 1 wins, that’s still bad news… for Player-B, who probably isn’t ready to play against 1100 ELO players on their own yet! This causes bad matchups, and we have heard from some players that they’re hesitant to play ranked TG at all because it may inflate their ELO beyond their actual skill level.

So, how do we fix this? In our new system, ELO is awarded based on the average skill level of both teams, which produces much more appropriate adjustments. Here’s an example using the same players as before:

Ranked RM TG ELO – HERE’S AN EXAMPLE OF THE FIX:

  • If Team 1 wins:
    • Both players on Team 1 gain 16 ELO (as if a 1.5k ELO player beat another 1.5k ELO player)
    • Both players on Team 2 lose 16 ELO

In the example above, ELO wins/losses are based on the appropriate average ELO of each Team and then calculated accordingly, avoiding the inflated ELO that results in matching less experienced players with more experienced ones. Over an extended period of time, this will have the effect of normalizing ELO ratings naturally, though we’re still discussing solutions for inflated TG ELO and may take additional steps to resolve that in the future.

For now, we’re thankful for all of your feedback and some really impressive analysis on this issue, and we’re glad to be delivering this fix today. Please keep sharing your thoughts!

The Age of Empires Team

♦  ♦  ♦

Discuss of 32 comments

  • naruto-5 June 3, 2021
    what is the point of changing how elo works wihtout reset elo?!! no new account win so few points while all other got their points by much worse win/lose balance
    Reply
  • diligentsloth June 2, 2021
    Thanks so much for the change! It's perfect for playing with my friends differ much in ELO. Now, everyone will get their own TG rating and we can find balanced matches no matter who plays. In the old system, we lost every game whenever one of our better players was missing.
    Reply
  • rogganwololo May 19, 2021
    https://aoe2.net/#profile-208337 How you manage to break an already broken system even more is simply beyond me. Why not just use a system that has already been proven to work correctly (i.e. the one used in Voobly)?
    Reply
    • rogganwololo May 19, 2021
      https://forums.ageofempires.com/t/random-map-team-game-elo-calculation/77045 My original post on this issue, dated 2nd April 2020.
    • rogganwololo May 19, 2021
      Correction: Break an already broken system even more *repeatedly* I'm sorry to have to use harsh language, since I had first posted about this issue only a year ago, only to receive zero response from anyone in the community management team, and the devs only "fixed" it now, by breaking it even more.
  • iron-churry May 19, 2021
    You have ruined people that play every day and wants to climb normaly. I have played a 4v4 match. My team were all 2100 players and the other team has two 2700, one 2400 and one 900. Obviously we lost due to the 2700 players were too good. After that suffering (due to it was not match at all) it took me 11 points of ELO. So, people that player normaly just for fun, will be ruined by people that is ELO boosting a friend. Thanks AOE2, you have ruined your game.
    Reply
  • brummer123 May 19, 2021
    First of all thanks for doing something. A few things are fixed now. Sadly smurfing and elo boosting is off the roof. Despite the people with 5000elo in 2 weeks, the good teams with a smurf in between are now allowed to kill noobs for 50 matches and not 10, because the smurf just gets around +16 elo each win. Instantly encoutered a Team in 4vs4 with three 2300elo players and one 1070elo player. got rekt very hard and the low elo one was even the top player.
    Reply
  • nc-jaysquared May 19, 2021
    I'm a bit confused, why the devs made this short sighted change. Many comments pointed out the potential for abusing smurfs and low rated accounts, in order to boost one account to arbitrary heights. The fact that less than one day after implementing the change, the #1 TG spot is already taken by someone abusing the system and gaining 300 points in <24 hours, goes to show how easily ELO can be manipulated now. The devs mentioned that this is just the first change of many, so I hope that they adjust the system as quickly as possible. My suggestion would be to keep the change that all players in a team gain the same amount of points, but instead of using the average ELO for finding opponents and caluculating points we should keep using the highest rated player in a stack as the rating of a team. That way people will be discouraged from teaming with high elo differences as is the case in most competetive games, and in order to gain points you can no longer just team with smurfs or newplayers. Example: 3000 + 2000 vs 2300 + 2800 Team A wins: Team A both get ~4 points, since their highest rated player is 200 points above Team B highest player. Team B both lose ~ 4 points, for the same reason. Team B wins: Team A both lose ~ 20 points, Team B both win ~20 points
    Reply
    • nc-jaysquared May 19, 2021
      An alternative (and in my opinion better) system would be use the average Elo for matchmaking and point distribution among the teams (as is the case in the newly implemented system), but instead of awarding/subtracting every player in a team the same amount of points, you give the higher rated players a much smaller portin for a win, and take away a bigger portion in case of a loss. Example: A (3000) + B (2000) vs C (2300) + D (2800). If AB win: A+0, B+18, C-2. D-16. If CD win: A-14, B-0, C+2, D+12. This way you have no Elo inflation and you cannot boost high players even higher.
  • iamdalv May 19, 2021
    I think a reset of the TG ladder should also be done.
    Reply
  • skialp May 19, 2021
    It was a necessary fix. @Age DE Team But since now the ranks are high, what about a ELO recalculation based on match history?
    Reply
  • mists May 19, 2021
    Have you ever thought what would happen in high elo tg with the new system? If I got a 3k2 account, and queue up with 3 1k elo smurfs, with the new system we will face 1k5 players and gain 16 points each game for all 4 accounts. After 50 games I will easily get 4k elo by beating players from 1k5 to 2k3. If this is the system you are going to try, the match making system need to be rearranged and new rules are needed for a team with extra high elo differences. Some examples from other games would be restricting the elo difference limit for a team, or calculating the team elo depanding on the highest elo only.
    Reply
    • leylthebard May 19, 2021
      edit: I missread a bit, but the point holds after some number of games, because the elo-system will correct itself.
    • leylthebard May 19, 2021
      That is absolutely not whats written there. In your example you would just get 1 point (or maybe even 0 points) for winning the game.
    • ehbahsuper May 19, 2021
      "Restricting Elo difference" or "calculating the team elo depanding on the highest elo only" would encourage smurfing. So if the 3k2 player doesn't want to reach 4k because of his stupid friends who are smurfing he should juste change friends in my opinion
  • cybersteel8 May 18, 2021
    Well done for caring about the TG ranked ladder and improving it to maintain the competitive integrity of the game. I applaud the devs attention and their willingness to take action and improve the systems. Is there any information about WHEN this system will be introduced?
    Reply
    • diligentsloth June 2, 2021
      Also interested when the new system will take effect!
  • mzhao171 May 18, 2021
    To combat smurfing (someone with an under estimated elo) and account boosting (someone with an overrated elo) and make ranked team game a better place for everyone, here is my proposal: We petition Microsoft to deploy a neural network based machine learning algorithm solution that keep track of someone's true performance in both ranked and unranked games with training parameters such as elo, win rate, team contribution etc. For example, if it determines that someone should be placed at 1300 elo based on statistics and deep learning but that person's actual elo is only 1000. The elo system should in theory treat that person as a 1300 player and calculate his/her team elo based on the expected elo instead of the smurf elo. And vise versa, if someone is consistently under contributing in team games, the neural network will find an expected elo for that player based on their performance and place them accordingly. With the implementation of the machine learning algorithm, any potential smurfers will not be able to win consistently and be placed according to their performance. Ofc, no neural network is 100% accurate but it should in theory provide a better experience for all players playing online.
    Reply
  • reviewedhalo914 May 18, 2021
    As sigme pointed out the system you are proposing it's extremely risky. As he said it makes so much easier to climb the leaderboard if you play with teammates who have a very low rating since they will never reach your ELO and you will always be playing games at a lower ELO than your real ELO and this will cause an extremely new amount of smurfs accounts just to abuse this new broken system.
    Reply
  • _eaglewarrior May 18, 2021
    Why not just keep the current algo with a cap of 25 points max if team 2 wins against team 1. This new algo will incentivise smurfing which stagnated the Voobly community. Note that in your example the 2K player will look for highly rated player with smurf acc to team up with to get easy 16 points and climb the rank ladder faster. You will end up having 1000s of more active accounts but in reality it will be the same players with multiple accs. The easiest way to demotivate a new joinee to the community is to get smurfed. The community will stagnate and become venomous if you make this change. If you are going to do this still, please keep an extremely strict policy against smurfing.
    Reply
  • kingbaloni May 18, 2021
    Good change! Please do it to Age1 DE as well.
    Reply
    • joonastop May 21, 2021
      Exactly. Age 1 suffers from the same problem. Please fix that as well. Thank you.
  • sigme May 18, 2021
    Hello everyone. I wanna to explain why I think this system only solve the ELO inflaction problem, but, from my point of view, will cause a hard unbalance on the ladderboard and on the games. As I could understand on the post, now the ELO calculated is based on the average ELO in a team, and I think this is fully correct. What's more, it make the team who win, win the same ELO that lose the team who lose. But I think the problem is who this system distributes the points of ELO among the players of the same team. I think it should be proportionaly the Team ELO they have previously. Let's see an example: TEAM A: Total ELO 3000 - Average ELO 1500 - Proportions P1:2/3;P2:1/3 Player 1(2000 ELO). Player 2(1000 ELO). TEAM B: Total ELO 3000 - Average ELO 1500 - Proportions P1:1/2;P2:1/2 Player 1(1500 ELO). Player 2(1500 ELO). In case TEAM A WINS: TEAM A WIN(+16 ELO). Player 1 -> 16 * 1/3 (The inverse 1-2/3) -> +5 (Rounded) Player 2 -> 16 * 2/3 (The inverse 1-1/3) -> +11 (Rounded) TEAM B LOSE(-16 ELO). Player 1 -> 16 * 1/2 (Direct 1-1/2) -> -8 (Rounded) Player 2 -> 16 * 1/2 (Direct 1-1/2) -> -8 (Rounded) In case TEAM B WINS: TEAM A LOSE(-16 ELO). Player 1 -> 16 * 1/3 (Direct 1/3) -> -11 (Rounded) Player 2 -> 16 * 2/3 (Direct 2/3) -> -5 (Rounded) TEAM B WIN(+16 ELO). Player 1 -> 16 * 1/2 (The inverse 1-1/2) -> +8 (Rounded) Player 2 -> 16 * 1/2 (The inverse 1-1/2) -> +8 (Rounded) As you can see with this system, the responsibility for the victory or defeat of the team rests and works in proportion to the team rating that each player contributes to the team. Thus, in the long term, if these teams always play together, their team rating will end up equalizing, and representing the real level of both players when they play together (which does not have to resemble the one they have when they play individually).It type os system is similar to others used in multiplayer games with ELO system and games where faceteam vs team (LOL, Overwatch, Counter Strike, etc.). To end up, I wanna to explain way this system is better that one is proposed in this updates, and show many problems I have found in this update: 1 - Unbalance the high ELO lederboard: With the system, is easier to reach the top places of the laderboard with and accout than plays with low ELO player. What's more, many people could use it to boost accounts, playing always on team with smurfs o low ELO accounts. If you play with a low ELO account, you have easier matchups, but you win the same ELO. This makes it much easier to climb the leaderboard if you play with teammates who have a very low rating, than if you play with teammates of your level, since you earn the same points, but against a priori rivals with a lower ELO than yours. this is also very exploitable if you create how many smurft just for this purpose. 2 - Unbanlancen on long term: In the long term, as mentioned, if these two teammates always keep playing together and improving together, the player with the lowest ELO will never reach the first one, and his team ELO will never reflect his real level, since it will to always play with players better than him due to the average ELO. On the other hand, the player with the most team ELO will always have an ELO higher than his real level, since the match works by average level, and this never reaches an equilibrium. The system that I propose solves both problems, since it distributes the rating proportionally among the players that make up the team. Thus, the player with the highest rating wins less and loses more, and vice versa. This means that if you play with worse mates, and in games with an average rating lower than yours, you will earn fewer points, since your level is supposed to be higher. In addition, in the long term, this system causes both teammates who always play together to tend to equal their rating, and in turn this approaches the average rating of both, which will correctly reflect thereal rating of this team, within the classification and as individual players. This balance the team ELO, and makes the match making more fair. Thank you for read. SiGMe.
    Reply
    • resp2ct May 18, 2021
      be*
    • resp2ct May 18, 2021
      this is very smart. I hope the Age DE team will interested by your suggestion.
    • achaean-2 May 18, 2021
      This this this this this. Sounds almost perfect. The only thing id change id that maybe the way you split elo shouldnt be % based, because I believe a team composed of a 1100 player and a 1000 player should split elo de same way a 2100 player and a 2000 player do. The difference between elos should decide it, the same way its the difference that decides how much total elo is awarded
    • kauzenrocks May 18, 2021
      yes. If you just compare players to the average of the opposing team (not the maximum as before), and make sure that you round correctly, both teams will allways lose/win the same amount of total points (which is essential for a elo-based system). But distributed in a way, that makes playing only with low-ranked smurfs as teammates impossible.
    • povilaspanda May 18, 2021
      This is a very reasonable suggestion. It addresses the both reasons why originally the team matchmaking wasn't done based solely on average team rating. It's important to be able to catch up a bit (we can't have 1000 ELO difference between 2000 and 1000 always playing on the same team 2v2 games), and also, to not be able to boost your rating easily (by playing with weaker players on your team_
  • sigme May 18, 2021
    Hello everyone. I wanna to explain why I think this system only solve the ELO inflaction problem, but, from my point of view, will cause a hard unbalance on the ladderboard and on the games. As I could understand on the post, now the ELO calculated is based on the average ELO in a team, and I think this is fully correct. What's more, it make the team who win, win the same ELO that lose the team who lose. But I think the problem is who this system distributes the points of ELO among the players of the same team. I think it should be proportionaly the Team ELO they have previously. Let's see an example: TEAM A: Total ELO 3000 - Average ELO 1500 - Proportions P1:2/3;P2:1/3 Player 1(2000 ELO). Player 2(1000 ELO). TEAM B: Total ELO 3000 - Average ELO 1500 - Proportions P1:1/2;P2:1/2 Player 1(1500 ELO). Player 2(1500 ELO). In case TEAM A WINS: TEAM A WIN(+16 ELO). Player 1 -> 16 * 1/3 (The inverse 1-2/3) -> +5 (Rounded) Player 2 -> 16 * 2/3 (The inverse 1-1/3) -> +11 (Rounded) TEAM B LOSE(-16 ELO). Player 1 -> 16 * 1/2 (Direct 1-1/2) -> -8 (Rounded) Player 2 -> 16 * 1/2 (Direct 1-1/2) -> -8 (Rounded) In case TEAM B WINS: TEAM A LOSE(-16 ELO). Player 1 -> 16 * 1/3 (Direct 1/3) -> -11 (Rounded) Player 2 -> 16 * 2/3 (Direct 2/3) -> -5 (Rounded) TEAM B WIN(+16 ELO). Player 1 -> 16 * 1/2 (The inverse 1-1/2) -> +8 (Rounded) Player 2 -> 16 * 1/2 (The inverse 1-1/2) -> +8 (Rounded) As you can see with this system, the responsibility for the victory or defeat of the team rests and works in proportion to the team rating that each player contributes to the team. Thus, in the long term, if these teams always play together, their team rating will end up equalizing, and representing the real level of both players when they play together (which does not have to resemble the one they have when they play individually).It type os system is similar to others used in multiplayer games with ELO system and games where face team vs team (LOL, Overwatch, Counter Strike, etc.). To end up, I wanna to explain way this system is better that one is proposed in this updates, and show many problems I have found in this update: 1 - Unbalance the high ELO lederboard: With the system, is easier to reach the top places of the laderboard with and accout than plays with low ELO player. What's more, many people could use it to boost accounts, playing always on team with smurfs o low ELO accounts. If you play with a low ELO account, you have easier matchups, but you win the same ELO. This makes it much easier to climb the leaderboard if you play with teammates who have a very low rating, than if you play with teammates of your level, since you earn the same points, but against a priori rivals with a lower ELO than yours. this is also very exploitable if you create how many smurft just for this purpose. 2 - Unbanlancen on long term: In the long term, as mentioned, if these two teammates always keep playing together and improving together, the player with the lowest ELO will never reach the first one, and his team ELO will never reflect his real level, since it will to always play with players better than him due to the average ELO. On the other hand, the player with the most team ELO will always have an ELO higher than his real level, since the match works by average level, and this never reaches an equilibrium. The system that I propose solves both problems, since it distributes the rating proportionally among the players that make up the team. Thus, the player with the highest rating wins less and loses more, and vice versa. This means that if you play with worse mates, and in games with an average rating lower than yours, you will earn fewer points, since your level is supposed to be higher. In addition, in the long term, this system causes both teammates who always play together to tend to equal their rating, and in turn this approaches the average rating of both, which will correctly reflect the real rating of this team, within the classification and as individual players. This balance the team ELO, and makes the match making more fair. Thank you for read. SiGMe.
    Reply
    • sigme May 18, 2021
      Duplicated, sorry i can't delete the second :(
  • startedgorgon94 May 17, 2021
    So there should definitely be conservation of points, and that's a great change - awarding based on averag elo. The way points are split, however, encourages smurfing, so I think that's not a great change. I am therefore going to smurf until that change is reversed.
    Reply
  • fffdis May 17, 2021
    This affects AOE1:DE too right??
    Reply
  • viridbirch77415 May 17, 2021
    I have been saying that since DE came out! Thank you for "actively listening". Keep congratulating yourselves but by being so slow (and continuing to not listen) many friends and me lost all interest in the game.
    Reply
  • clemol45 May 17, 2021
    Amazing news ! It will encourage higher seeded players to play with everyone, even if they play with way lower ranked ones. They won't "fear" anymore to lose all their points on 1 game. It should prevent people from creating smurf accounts, which was even worse. Good job guys !
    Reply
  • arck May 17, 2021
    Thanks to the dev team for this extremely long-overdue fix. One more reason to play and enjoy ranked TGs again. Rather than recalculate existing elos using the match history, which may be janky due to so many matchups that never should have happened, not to mention potentially error-prone, I'd be happy to see the elos renormalized. I.e. calculate everyone's percentile, and assign them a new elo that gives them that percentile in a normally-distributed elo centered at 1000.
    Reply
  • hectornauta May 17, 2021
    That's some awesome news. Nice job devs.
    Reply
  • rustyiesty May 17, 2021
    Credit to Will and the team for fixing this after I posted about it last week on the EW article. Hopefully Elo could now be recalculated from 18 months of match history to give a correct Elo for everyone who has ever played a ranked game.
    Reply
  • rob-5 May 17, 2021
    Hooray! Took a while but great to see the change being implemented.
    Reply
  • zbyszekde May 17, 2021
    Btw I find it kinda weird that you are announcing this in a blog post just hours before this fix hits live. I mean its probably the single biggest change in the last 4 months.
    Reply
  • tocaraca May 17, 2021
    Very happy that this is finally being addressed.
    Reply
  • fadedfriend2906 May 17, 2021
    you guys are doing great work! thanks for putting in all the work! keep it up!
    Reply
  • tchesgos-2 May 17, 2021
    if player disconnect -1 min game should be invalid.
    Reply
    • heedfulfive7 May 19, 2021
      This used to be the case. I don't know why they changed it.
  • moveable35 May 17, 2021
    But what happens if some one is say 1700 elo and invites some one else that's ether 1500 (or lower) or 2000 elo. What then?
    Reply
    • tempires May 18, 2021
      (1700+2000)/2= 1 850 elo which would be base for elo distribution
  • davahr0103 May 17, 2021
    Great work devs! This is huge!
    Reply
  • riskierdeer2624 May 17, 2021
    With this change a reset of the TG-ladder/elos would be a wise thing to do in order to erase the elo-inflation that happend over the past 1.5 years.
    Reply
  • xuanzue May 17, 2021
    There is also the problem with premades. Premades tend to have different penalties in different games, because they have more coordination/discord. For example, a platinum premade 2v2 in starcraft tend to play with 2 solo diamonds.
    Reply

of 32 comments