The only changes that were made were to the new user experience, and Chemie not being a new user and presumeably of an above average Elo should remain unaffected (I assume!).
With the way the MM rewards a player for their team win or loss I do not see how the MM is specifically and intentionally trending players towards a 50:50 win loss rather than a 50:50 win loss being the resultant effect of MM mechanics at work for the average player.
The generally accepted notion that the playerbase as a whole will have a certain average of "skill" will have a large proportion of the playerbase falling in or around a 50:50 win loss without MM help in any event over time. We can see this later on.
That is to say that seeing a cohort of players with a 50:50 win loss is not to say the MM is currently working as intended if the vast population of the server is proficient at playing, but has only a tiny % of godly players.
If we take a look at the bell curve (see below) that was released showing the spread of players Elo, we can extrapolate that there are indeed only a small % of players on the playerbase who are to be considered as having a considerably better chance of impacting a match than the average player (the high Elo players).
Then consider that prior to the Elo adjustments, when the amount of games played impacted your Elo, it was still considered at that time that the Elo "was a perfect bell curve".
Bryan Ekman, on 08 April 2013 - 11:36 AM, said:
Ask the Devs #35
cjmurphy87: Would you consider releasing some data on elo? In particular I'd love to see a frequency distribution of the various elo scores, I'm curios whether the community has formed a bell curve, or if there are bands centered around several different modes.
A: There is a perfect bell curve. We are considering however, dropping the entry point ELO value for new players. This will help new players start at an easier play bracket and allow them to quickly move into their appropriate ELO bracket.
Obviously, we now know that the bell curve itself as it apparently existed at that time could only highlight what % of players had played the most games - presumeably not even looking like the perfect bell curve when applied to the Elo mechanic at all or there was a missing metric for it at that time.
As per the adjustments made we then have this thread:
http://mwomercs.com/...ost__p__2265319
Coming to today, we can see that now the vast server majority at the time of this graph has an Elo score below that of the entry level of 1300 (adjusted to 1350 now but 1300 at the time) - i'd guestimate 70% but someone else can do the actual counting.
This means that unless there is a significant portion of players online within the top 30% then any player within that bracket is having a disproportionate effect on the win loss ratio of any team he is then dropped on made up of the bottom 70%.
I suspect a lot of the top 30% drop in premades, which might explain the beating down of the bottom 70% below the entry level Elo point.
As should be obvious, only 50% of the server population can be considered below average so we can see from the graph that the difference beyond the point of average and that of highly proficient is quite large, as is the difference in the number of games the top 30% play compared to the bottom 70%; tis what I assume is pushing the Elo down from a starting point.
If it was the intention with MM to push players towards a 50:50 win ratio then this graph shows the wait time has to be pushed up for the top 30%.
As w/l in this current arena game really means very little without something to play for, I suspect top Elo playered don't mind it not being fiddled with so long as they're getting a game.
However, what they do mind is dropping vs trials who, generally speaking, will have a very small % of winning or even understanding what it is that's going on when they are getting pummeled by a mech that doesn't seem to shut down.
It's a bad new user experience too.
Edited by Exoth3rmic, 10 June 2013 - 10:36 AM.
3
I KID!
