I found some interesting points from these 25 matches.
In regard to 4x3:
Here are the average #s from that sample data:
Lights ------ 2.44
Mediums -- 2.64
Heavies --- 3.48
Assaults --- 3.44
Outliers. Lights are actually NON-EXISTENT in one match. That means least 3 valves where released by them alone in that match! Mediums are almost as bad, since they have 3 matches where there was only a single Medium. On the other hand, there is a match where there are FIVE Heavies, as well as a match where there are FIVE Assaults.
50% of matches where 4x3. That is decent, but makes valves being released the norm, not the exception. Russ did state that this is one of the things they want to quickly fix.
In regard to Variance:
Now, something we cannot get a full view of is the Variance (far left column). We can only see 14 out of the 25 matches. Still a decent number to look at, though.
Elo is from 0-2800. While the average difference in variance is 241.27, match #24 in the list has a HUGE 762 variance. That is 30% of the whole Elo range, meaning one team is roughly in the middle of the mid-Elo "bucket", while the other team is roughly in the middle of the LOW-Elo "bucket". I would be willing to bet that was a horrid, horrid match.
Also, 9/14 matches were in triple digit variance, 8/14 were 200+ variance, and 4/14 were 350+ variance.
What makes this even worse is that these team variance's are the average Elo of the players on that team. Not only is there a large amount of difference between teams, but you also have the very possible issue that has been brought up before: High+Low vs. Medium.
What this means is that you have a few very High Elo players offsetting a bunch of very low Elo players, balanced against a full team of average players. Even though their combined Elo averages may be "close", the skills of those individual players are horribly "not close", which is what causes a lot of stomps to begin with.
Lessons:
- Role Warfare (rewards) needs to be implemented ASAP. Trying to force players to play lighter mechs is very obviously not working, and never will work. Until players are rewarded for playing what they WANT to play, they will never play them regularly. The vast majority of players would rather wait a couple minutes than play a mech they get little to no rewards playing.
- Data Transparency. This is exactly why PGI does not like releasing specific data...
UPDATE:
Karl Berg, on 05 July 2014 - 11:34 AM, said:
Hey Cimarb, unfortunately those variance numbers are not true statistical variance values. Out of lazyness I was simply calculating highest Elo on team - lowest Elo on team, rather than a true normalized sum of differences from mean.
Those screenshots were also taken within the first hour of launch of the new matchmaker. We've gone through extensive tuning passes since, both to bring wait times vs percentage 3's games in-line with stakeholder expectations, and to significantly tighten up Elo matching as well.
Edited by Cimarb, 05 July 2014 - 11:40 AM.