Apologies if there are similar threads (maybe even by me). It seems to still be a problem.
Chess has mechanisms to combat ratings deflation, which is people with a certain skill level earning a decreasing rating as time goes on (or, the skill level of all players at rating X increasing over time). For FIDE rating, one of these mechanisms seems to be that players who go below 1k rating get unrated, and they earn a rating again by playing against rated players in a tournament, which jumps their rating up (probably significantly above 1K), injecting points into the system. Apparently, the US also uses bonus points to combat ratings deflation.
AoE2 does not appear to have any of these mechanisms. I don’t know about AoE4. Ideally, new players should be able to start playing multiplayer and within a very small number of games, find opponents against whom they are evenly matched.
AoE2 has problems that chess doesn’t: there are more people who enjoy being at an artificially low rating, which lets them win a lot of games, and players can make new accounts to get a new rating. This means that some “new players” will be very good, but might still pretend to be bad for placement.
Still, that doesn’t change the fact of ratings deflation. If new players are given an initial rating of X, and matched against established players with rating X, then X should be about the skill level that new players will stabilize at.
Should the ranked ladder have mechanisms to inject bonus points into the system to combat ratings deflation?
- Yes
- No
A thread from 2020 suggests that when players with a rating on one ladder start playing on another ladder, their rating should be based on a normalized score based on their percentile rating within the entire pool, instead of assuming that the ratings on different ladders are equivalent.
(Team games ranking was reworked and reset several patches ago, but some interesting threads from 2020: comparing 1v1 rating to inflated TG rating, and a followup thread with more discussion of the problem and consequences)
But, instead of accepting that the ratings for different ladders are not comparable, maybe actively try to make them comparable with small point adjustments each game.
But, note that each game is a prediction of the outcome. The Elo rating system is supposed to be one in which if two players of unequal but separately stabilized ranking play each other repeatedly, neither should change in rating. If one player wins 90% of the time, they should be losing 9 times as much rating for a loss as they get from a win. Team games might not work like this, especially with a simple average. It might be the case that if Hera, with ~2800 1v1 rating, teamed with a 1600 player and played against two 2200 players, the 2200 players would win 90% of the time, or lose 90% of the time.
So the matchmaking system can use data from wins and losses to refine its predictions of match outcomes, and so form more equal matches (or better assign ratings changes when it expects bias in the match’s outcome). This will allow the teamgame rating to better correspond to 1v1 rating.
Anyway: I know that a lot of the players who play against Hera in his 1v2s, 1v3s etc. have a lot of ranked team games, but I mostly watch videos from Rage Forest which is unranked.
Ideally, everyone would be honest about their skill level and not try to manipulate their ranking to get uneven matches. Then the rating system only has to worry about how to get a player with unmeasured skill to their proper rating as quickly as possible.
Then you would have two separate problems: how you could use a player’s claim about their actual skill level to accelerate the process of finding their proper rating; and how to address any concerns or motivations that might cause players to want to lie about their skill level, either to exaggerate or to understate their skill. Note that experimenting with unusual strategies that may or may not have any viability can cause a drop in rating, and that then switching to standard strategies could lead to a series of wins that seem to the opponent like ‘smurfing’ (deliberately playing at a low rating to get wins). But this is going beyond the scope of this short post.
Do you get upset when you are matched against a player who seems to have deliberately manipulated their rating to play against weaker players?
- Yes
- No
What do you think of the general idea of players being able to suggest the rating that they SHOULD be at, which accelerates rating changes towards this rating (and could be shown along with actual rating)?
- It would help
- Too many players would exploit it to get a high rating or a low rating
- Good only if it doesn’t immediately affect the system’s choice of opponents until the win/loss record verifies the claim
- The unverified claim about rating should never cause opponent to lose extra points
What do you think about allowing ranked FFA games, in which most players lose the game and some rating and only one player can gain rating?
- New players would like this option
- All players would like this option
- No one would like this option
- Even if some players like it, it isn’t appropriate for ranked play