Non-Ladder Matchmaking and Smart ELO for Team Games

As AOE2 increasingly becomes a team game, the time has come to re-think ELO and matchmaking. While the win rate-based ELO system kinda-sorta works for 1v1 games, it’s clearly flawed when it comes to determining skill in team games, where the outcome is dependent on not just your skill, but that of your teammates. After all, ELO was developed for a 1v1 game (Chess), not a team game.

Lopsided matches are never fun. Most players (of both 1v1 and team games) just want to play fun games where they can try their best against players who are roughly their skill level. And our goal should be a system that can provide that consistently for the exploding number of players that primarily enjoy team games. Here is how we can do that:

Instead of focusing on game outcomes to guess a player’s skill, we should analyze how they actually play. We can start by creating a model that looks at dozens of key metrics, everything from APM, to age-up times, eco, military production, and create bands of skill levels. A players’ skill level would be determined against those models, which could be continuously updated over time.

Not only does this fix the team game matchmaking problem, we can even use it to help players improve! The game could give players tips based on how they play, like “you create fewer villages in Feudal age than other players, be sure to keep making villagers in Feudal,” or “you create military units later than other players, try making military units earlier.” The possibilities of using a machine learning model here are huge.

It will also have an impact on smurfing. Instead of 10 games of smurfing to figure out a player is much better than the ELO of their newly-created account, the system will figure out faster they are skilled by how they play.

With this new system, we can also fix another common complaint of players: ranked map pools. Because the ranked ladder is the only way to play with players of a similar level (or try, at least), many players are forced to play maps they don’t enjoy. But you need uniformity on a ladder to rank people. However, many (perhaps MOST) players don’t care about climbing the ladder. They just want balanced games against players of similar skill where they can try their best.

But now that we have a skill system that isn’t tied to a ranking, we can matchmake without a ladder! This would allow players to choose a pool of their own favorite maps (for example, we can force them to choose a minimum of 8 maps they are willing to play to keep matchmaking smooth), and just allow them to play whatever they want.

I think these two changes would dramatically improve the multiplayer and team game experience and improve the online experience for everyone.


I agree that re-thinking the ELO and matchmaking system for team games in AOE2 is important for a better multiplayer experience. Your suggestion of using machine learning models to analyze key metrics and determine skill levels could be a more accurate and fair way of matching players. It could also help players improve their gameplay and discourage smurfing. Allowing players to choose their own favorite map pool without a ladder could also make the game more enjoyable for everyone. Overall, your ideas could bring positive changes to AOE2’s multiplayer community.


Thank you! I appreciate the feedback.

1 Like

Love it! LETS GO!!!
As I suggest something similar quite a while ago I’d love to get into more detail


Wow, I definitely think you had the right idea. Obviously I think we don’t even need ELO at all, just rate people by how they play.

I don’t necessarily know what the correct metrics are. I think the devs should just run an analysis on high-level players and look at where they have spikes that can point in the right direction. Then just start to build a model from there.

1 Like

Sorry for this noob question, but does “Age Insider” mean you have access to the dev team and could give them this idea?

1 Like

HeBro, does “Age Insider” mean you have access to the dev team and could give them this idea?

1 Like

You can find the Dev Agent N°1 @FloosWorld

You can sign up for the Age Insiders on :slight_smile: through that way to get the chance to take part in beta tests!

1 Like

No, sadly it doesn’t mean that.

I think the right team to work on something like that would be the Captureage guys?
They already have done a lot of things regarding advanced statistics etc.


Thank you for the offer. I’m not really looking to join the beta group, just a way to present ideas to the dev team. I have submitted this idea a few times in support tickets but haven’t heard any feedback from them on it.

1 Like

Well, Insiders is the first step to have a more direct contact to the dev team. Beta tests is just one aspect

1 Like

Yes, given the fact that the elo is just not giving accurate information. Say 90 percentile in solo rank is 1450 elo but in team rank it is 1350. One cannot determine the skill level by direct compare their solo rank lo and team rank elo