The current solution is obviously way better than a reset, because all players will have to play only a few games instead of dozens for the elo to be settled in. so less unballanced games (you would have to play a ton of them with a fullblown elo-reset).
The expected end-distribution should be basically the same. no need for a reset.
Iām not sure if you are right there.
I expect a different distribution. One kinda sharp one with more people around the mean elo.
And I need to check this, but it is likely there will indeed be a net elo gain per game though it might be less than we had in the beginning cause there will also be games where elo is lost.
Which would mean that again naturally all the players who just play more games will have overall increased elos.
But itās hard to tell as in this case we have counteracting forces and itās not clear which one will be more impactful yet.
Why do you reply to everyone who is commenting game negative way? or complaing about system? You clearly dont see the problem if player x enjoys playing tg more than 1v1, why they should switch into 1v1?
Iām willing to accept that thereās a problem, but only ever complaining about it doesnāt encourage the devs to fix it, and just gets annoying. They genuinely seem to hate TGs based on how often they complain about them, so Iām sincerely confused as to why they donāt play 1v1s. They have never said anything about enjoying TGs more, only ever complained about them, so it seems like a logical solution to me.
I donāt hate teamgames, but I do hate the new calculation with a passion, because it will make it harder for me to find good teamgames - which saddens me because I like teamgames. The previous method was better, so it would be an improvement if the devs reverted to that one. That is a sensible and obvious solution.
Additionally, there have been people in the forums suggesting alternative calculations. The devs could take inspiration from those. No one asked for this new calculation and the only reason some are praising it is because the devs say it is an improvement and they are being taken at their word.
I am not talking about the one time rating adjustment - that was a good idea, though it is more than offset by the new calculation, which will worsen things.
Iām not really sure why you replied to me to be honest, I wasnāt talking about you or anything.
11
Yeah there are seemingly still some general haters in the forum.
But it has become way better also thanks to some unsung heroes in the mod team who removed some of theā¦ most emberassing subjects.
And I personaly am fine with how it currently is. Life is no picnic and to some extend you can just ignore it.
Do you play with random people?
Not usually. Why? strhxhnsty20chars
Does anyone know what the Elo adjustment formula they used was ? Am needing to retroactively update my database and want to make sure I get it right otherwise its going to cause a mountain of issues
T-West posted on Reddit a formula
Not sure where he got it
Seems like itās a simple affine transformation, approximately using the formula new_elo = 0.46134453057734565 * old_elo + 433.32528079864903
, which indeed has a fixpoint of about 804.5.
I used to lose A LOT of games and Elo while playing with random people
Probably there are also just new people who now start team playing as they think āfinally they fixed itā.
So a lot of people start at 1k that should be ranked higher.
I think this phenomenom also killed EW partially.
I argued this exact point on Reddit for quite a while, at least for me the main thing that put me off EW was how many poorly balenced games I had at the start. I really wish they had just preset peoples Elos based on their 1v1 laddar position. I mean in general Elo between laddars is highly correlated so it seems like such as easy quality of life improvement (obviously itās too late now)
EDIT: itās such a shame because I genuinely believe EW is a better more entertaining game mode. I am hoping that the upcoming redbull tournament boosts itās popularity, tbh I wouldnāt be against them doing a EW Elo reset to avoid this problem again when people start trying it after redbull
Now is like a one off season end cut off and then reset everyone elo. Ofc eventually the elo of those abandon smurf account will drop to 1000 but it take times. And the queue is still broken
Iāve been making quite a few mistakes recently so am not going to rule out a bug in my code but it appears the Elo inflation from the new formula is already kicking in:
(Note that I applied the Elo transformation to all historic matches as well so that the trend is clearer)
How, do you know? 20 characters
As in the team mean Elo line has been slowly dropping ever since they originally fix the team Elo distribution formula to make it balanced . Since the most recent change the last 3 ticks on the plot have all shot up showing a Elo inflation. This could also be explained by lower Elo players all suddenly stop playing at once but the former feels more likely to me given the new flawed formula
Indeed, I am close to that and only lost 14 elo after the first match post updateā¦
When the inflation rate is too much then they will apply one time adjusting on it. Kind of refreshing the rank in every season