The new equations for rating points are bad, if there is no change in matchmaking also:
It will again result in lower level players having a negative winrate by playing a lot of games (and get a higher rating than they should have)
It will cause (especially high level) players to be put in a small range of ratings, not accounting for their actual skill distribution. Rank boosting might be reduced, but the small range of rating doesn’t refelect actual player skill.
It’s not made for people who just queue up ladder randomly, but also play with premade teams once in a while. A lower level player will get a lot of points queueing up with higher level players. And then he is boosted, if he queues up ladder alone.
Let’s see how it goes.
Can’t be worse than before, with all the smurfing/multiusing.
I can’t believe I say this. But you are abolutely right and I have to like your post, though I personally dislike you because you are usually completely out of reality.
If the game is stopped beeing monetized and serviced, I don’t see why releasing the source would be unrealistic. It doesn’t mean giving up copyrights, and the engine/reusability doesn’t seem like a good asset to me.
I am still sure, that with a different directive, there would be a different software quality.
Totally agree. The new system is not awesome but it cant be worse than before, at least some kind of progress.
Lowbies teaming up with their high level players will have inflated, boosted elos and will get crushed later when playing solo. But thats how it is in most other games. Also thats how it should be…
The new elo system is inclined to toxicate the community. As (relative to the team) bad/low elo players will be branded as parasitic elo soakers. And “losers” nobody wants to have in their team.
If devs don’t see that and change it now before it’s too late, TG ranked will be completely destroyed.
I’m sick of this metaphrasing narratives we have all over the place these times. Sometimes the reality is just different than what we talk.
The old one only allowed a very small group of system abusers (cheesers and exploiters) to get absurd elos, the new one affects ALL team games and players with that toxicity.
Even if people play purely solo queue teamgames, they get assigned a rating from a distribtuion with a super small radius for all players. Resulting in the existing matchmaker not beeing able to gives good games.
A small radius is not desired as rating distribution.
Also the equations setting the ratings will break the winrate for an adjusted player:
Lets say a high level player gets/loses 2/8 points and for a low level player 8/2:
The low level player wins 1/5 games to maintain rating
The high level player wins 4/5 games to maintain rating
Apologies I,m not sure I follow would you mind explaining this more? Why is the radio us smaller? Why would the gain / loss be different on average for a solo queerer? Not saying your wrong, just that I don’t understand.
because of that. in the same team the player with the higher elo usually will lose elo gradually with the same winrate and the one with the lower elo will gain.
In the example I made the player with 2 k elo will lose elo on average despite a 67.8 % winrate.
If you team with people higher than you, you will get ~8 points on a win and lose ~2 points on a loss.
Ideally the matchmaker give you games with 50% winrate.
So you will rise in rating, not according to your level, in few games.
If you solo queue again, your good teammates are missing, and you will be rank boosted.
The opposite works if you play premade games with lower level palyers.
Sticking with the case of someone who queued purely solo I can see why they’d have a smaller dispersion under the new system:
+2, -30 if they are the highest Elo
+16 - 16 if they are middle Elo
+30, -2 if they are low Elo
Compared to just +16, -16 for the previous system regardless of Elo.
(Win / losses assuming balenced Elo teams)
But I’m not sure I follow why this is a bad thing, if anything it’s a good thing as it provides stability to your Elo meaning you aren’t guaranteed a loss just because by chance you won a few games in a row when you were top Elo.
Don’t get me wrong I understand the issues if you are switching between games with friends Vs solo queueing but I mean this in the context of someone who only solo queues?
The problem with this system when teaming with a player of lower elo you need to have an absurd winrate to “hold” your elo.
Then ofc nobody wants to team up with people of lower elo anymore.
Your account is adjusted, and you are supposed to get 50% winrate
But if you are a low player, you will let’s say win/lose 10/5 points for each match on average.
So to maintain your rating, you need to have 2 losses for 1 win.
It was known why the equations on release of the game were broken.
It was known why the queations since the last year are broken
It is completly obvious why the new equations won’t work either.
Vivi/Tim/Jibatong/Dragonstar/HGB/Fury don’t even need a smurf account in their team to bash newbs soon.
Matchmaking will be auto bash newbs if the rating distribution is set.
Isn’t this only true for players on the extreme ends of the Elo distribution, for like 98% of players wouldn’t you have an equal number of games being top and bottom Elo?
I guess it depends on the skew of the distribution of grouped players in a match on average but that is a function of their match making formula which I don’t think we have access to. Even then I would be surprised if the skew was significant for the vast majority of players
Agree. Also, it is better for noob players to get out of fake 1000 elo sooner. I use to play with my brother and some friends, and my friends still are arround 1000 elo despite being worde than I, because they only play with me while I play with ny brother sometimes.
So, everytime I play with them, they are considered the best player of the team… We stillnwin sometimes when we are matched against other fake 1000 elos.
Do we know the proportion of premade teams in teamgames?
Edit: i have looked into different aoe2 APIs, and it seems that the players are coded every match with a different token, so I cannot check that information from open data. The only alternative is web scrappimg websites with user information, which I dont know how to do