Quote:

Originally Posted by

**Zizoz**
I strongly disagree with the assertion that we don't care if dodgers' Elo increases due to dodging. It results in a game where the optimal strategy for gaining Elo involves as much waiting as playing, which IMO is a problem. I don't want to spend half an hour or more waiting to queue because I dodged every time I don't get my best position/get countered/etc, but I equally don't want to play knowing that I could have 100 more elo if I just had the patience to dodge in those cases.

So the short answer is that the assumption that creates the result that people simply dodge if they think their chance of winning at character select is <50% and so it takes ages to play assumes

*there is no penalty to dodging*
But there is a penalty for dodging, and its decently effective in normal. Since the penalty is extended now we can expect fewer(or a similar amount) of people will dodge. As far as i can tell, if you dodge twice in a row you're not going to be waiting 30 minutes for queue, you'll be waiting even longer.

Quote:

I'm not sure what exactly you're getting at in the second part, so could you explain more clearly how dodgers «break» the system? I posited in my earlier post that they would simply cause deflation in elo for everyone, which I agree is undesirable but which I suggested is more fair than loss of elo only for non-dodgers. It also is not the only source of elo inflation/deflation: loss-forgiven games result in net increases in elo, while games where new players to ranked (whose elo fluctuates more per game) are matched up against more experienced players likely results in a net loss in elo, as the experienced players are more likely to win. If the elo being introduced into the system is equal to that being removed, then there is no problem at all that I see. If there is net elo deflation, then IMO the best solution is probably to compensate for it by introducing elo elsewhere.

Let us assume that you do not dodge. The distribution of dodgers on your team is binomial (p,4) and the distribution of dodgers on the other team is binomial (p,5) where p is the probability a player systematically dodges.

We note that E[binomial (p,5)] > E[binomial (p,4)]. Which is to say that for any probability we expect more dodgers on the other team assume we do not dodge.

We also note that f = E[binomial (p,5)] - E[binomial (p,4)] is a function of p. If p increases then f increases, if p decreases then f decreases.

A "general reduction of everyone in ELO with no changes to the probability of winning given your skill ELO and listed ELO" will occur if p is constant and the size of the consistent reduction from dodging is also constant [I don't have a proof for this but it should not be that difficult to show. Note that if its not true then your proposition is false].

If this is the case it should also be the case the ELO loss on dodge behaves exactly like no ELO loss on dodge given that there is no penalty from dodging, since the equilibrium point for dodgers is when their lower skill fails to compensate for them leaving unfavorable champion matchups

If p increases as listed ELO falls, or if the size of the consistent reduction from dodging increases as ELO falls then there is no guarantee that this is the case. Instead there will be section where losing listed ELO will not increase your real chance of winning a game.

If losing ELO does not increase your real chance of winning a game[assuming that you're not dodging] then our matchmaking isn't working and ELO hell can exist. To give an example: there could be people who stabilize at 1000 even though they could also stabilize at 1100 since the chance to win a 1000 ELO game is the same chance to win a 1100 ELO game given that your real skill level should put you at 1050 or 1100.

Figuring out exactly what the chance of winning a game is depending on your actual skill level would require a bit more information about the matchmaker [and frankly a bunch of time I am not willing to spend] but this is the general idea and problem that exists with ELO loss on dodge.

edit: And we have a reason to believe that p is not constant because consistent dodgers are doing so in reaction to the games they're playing, which means that if the ELO of non-dodgers all falls, they're going to respond by dodging more and getting even more favorable matchups