Right now a new account starts around 1000 1v1 ELO whether you are complete beginner or an experienced veteran.
You ELO score is purely affect by either a win or a loss. (No matter whether it’s a close loss or you just go afk to lose on purpose).
The matchmaking works by searching for an opponent with similar ELO score.
This kind of matchmaking has its issues as it does not recognise secondary smurf accounts or very good / beginner players with new accounts as they need a lot of matches to find their correct rating.
Recently another game (Apex legends) introduced a new game mode which also has an ELO like ranking system (which gives you points purely on whether you win or lose a match). But they tried to fix the smurfing issue in an interesting way.
While you have your official ranking, the game also calculates a hidden skill ranking for you which not only incorporates the wins/losses of your last 10 matches but also a lot of other metrics like your K/D, damage you have done during the matches, survival time, winstreaks etc…
The matchmaking works by giving you opponents of similar skill to you (rather then similar ELO).
→ this means for example: if the game thinks you are much better then your current ELO suggests you will match with a much higher ranked opponent. If you win you will gain a lot of ELO. If you lose you will only lose little ELO (because the amount of ELO you gain / lose from winning/losing always depends on your opponent).
Do you think a similar system could be beneficial for the matchmaking in AoE 2 ?
Benefits could be:
the system detects a new account stomping opponents and going on a winstreaks → gives them tougher opponents much faster then currently.
the system detects a new account losing every game → gives them easier opponents much faster then currently
the system detects a player afking to lose a game on purpose / resigning without playing → detects that behaviour and doesn’t rate those matches (or to a lesser extent).
There is no benefit from having to seperate ratings, one visible for the player and one that is used ingame. That will only result in confusion by players in my opinion. I expect an increase of the number of thread about unfair match up, based on Elo, because some hidden value does the matching. As result i dont see any benefit of having a secondary hidden rating which is used in match making. Having the different ratings makes it needlessly complex in my opinion.
That doesnt mean you have some good points about how to detect smurfings and how to deal with them. I would suggest to change the current rating for that reason. So your skill is reflected in your visible Elo and used in the match making.
Things you mentioned are for example part of the Glicko-2 or TrueSkill. Both are rating systems that tries to improve the weaknesses of Elo. Both seem an improvement over the current Elo system and will be better in finding smurfs. Both should also be better in making balanced team games. Elo was meant for 1v1 games, where Glicko-2 and TrueSkill are better suited for team games.
The smurfing issue is also mainly a team game issue as result of the bad fix of the team game ratings from the devs. They are already looking into this issue.
This thread already discussed multiple solution in how to fix the current team game ladder. One of the current issues is that every account is weighted evenly, while higher rated players are probably able to carry the game for the team and have a bigger impact of the outcome. So there are also ideas to add more weight to the higher rated teams by matching teams. This makes adding smurf accounts to boost an elo of a main account much less worthwhile already. If you then also add something to the game to boost the elo of account that play much better then there rating suggest (based on long win streaks and stuff like that), then the total package seems like a good one to fix the team game ladder and limit smurfing.
Rating systems can, in general, benefit greatly from utilising additional information beyond simple win/loss. One problem with utilising the additional information, though, is that you’re assuming each player is trying their best, whereas people will, of course, try to exploit any such system. You mentioned an example of this with people AFKing to lose on purpose by a big margin, but people would work out other exploits. So before using any additional information, it’s essential to try to think of how people would exploit its use, and whether it is more or less exploitable than simple win/loss. Given that the devs seemed unable to imagine how easy it would be to exploit using a simple average of the ratings for team strength, I wouldn’t have much confidence in this analysis being done well.
The problem with a hidden skill rating in AoE2 is that games are snowbally on most statistics so that you wouldn’t have a good point of reference. It’s not like players have a consistent k/d ratio across their different games