Hi there from VainSocial 👋


#1

Hello forum people!

I’m shutterfly in EU and on Discord, I do a lot of research with API data and build https://vainsocial.com. I’ve never been active on the old forums because I didn’t like the
software much. After the rebirth, I can’t wait to discuss our data and our findings with you! :slightly_smiling_face:
In real life, I study informatics and occasionally dance as a hobby.

If you have questions about statistics, you can ask me and I’ll do my best to answer :nerd_face:


Vainsocial stats meaning?
#2

Love the site. I see peaks and valleys across the True Skill. What all goes into the metric (aside from play well and don’t die)?


#3

It’s only win/loss and the ratings of all players in the match :smiley: It’s a Matchmaker rating from Microsoft Research: https://www.microsoft.com/en-us/research/project/trueskill-ranking-system/


#4

Hello from your Discord server, it’s nice to see you here!


#5

Interesting - I’d love to know more about this has played out in the VG player pool over time. I was under the impression that roam mains generally have lower true skill ratings than carry mains (assumption 1, which may or may not be true), and that the true skill system might be using stats other than win/loss to make its ratings (assumption 2). I think I was wrong! Do you have numbers on average ratings by role? Say, players with >70% X role v. players >70% Y role? And ideally, solo queue versions of those populations?


#6

Do you think it’ll be possible to calculate what heroes typically win/lose against in match ups and be able to create a counters list based off the data? An example of this is when CP Taka is in a match against CP Petal, Petal wins most of the time, therefore CP Petal is a counter against CP Taka. Additionally, do you know if you have access to what talents were used in brawl matches? If so, that would make brawl mode matches feel more complete.


#7

Under the assumption that a “main” is someone who plays a role at least 2/3 ranked matches, only looking at a sample of players with more than 20 matches analyzed by us:

  • On average, a captain main has a TrueSkill of 1 547.63 (765 players)
  • On average, a jungle main has a TrueSkill of 1 599.33 (533 players)
  • On average, a carry main has a TrueSkill of 1 895.96 (388 players)

The same analysis with players at least at skill tier 7 in their last match:

  • captain: 1 906.60 (454 players)
  • jungler: 2 242.03 (244 players)
  • carry: 2 615.47 (209 players)

We do not get any party information (see https://github.com/gamelocker/vainglory-assets/issues/194#issuecomment-320777656).

The statistics confirm that captain mains have a lower TrueSkill rating on average.
TrueSkill is relative and based on nothing but win and loss. If you let all captain mains play against all carry mains and fill the other positions with players who are perfectly, equally skilled, the carry mains will win slightly more matches than the captains.
There can be multiple reasons for this - just theorizing a few possibilities:

  • Most captain mains just fill, unable to play their true strengths.
  • There are two large groups of players: “flex” damage mains and captain mains. Only the most dedicated damage mains stick to one of carry or jungle.
  • The captain role appeals to players who lack mechanical skill, which, in 3v3, is the key to success for 95% of the players.
  • The carry role has the most impact on the game. While a captain is necessary, games are decided by damage dealers.
  • It is harder to play the captain role to its full potential with two strangers. The carry and the jungler need to synergize with the captain. A good carry can play with a bad captain better than a good captain can play with a bad carry.
  • VainSocial is only being visited by good carry and jungle mains.

Analysis: Role "mains"
I think I need to be a god in SA to hit POA again?
#8

Yes, we have that data. SovereignChats (https://twitter.com/SovChats) is actively using it and we are working on a smart draft helper.

We do not have Talent data but we have requested it. Maybe it happens this year.


#9

Awesome, thanks for running those numbers for me! It’s always really interesting to see some meta-data about the community at-large.

I think I had Trueskill and Impact confused in my mind. Could you explain what Impact is measuring, and what you think it’s limitations might be?


Vainsocial stats meaning?
#10

I wrote more stuff. Analysis: Role "mains"

Impact is a formula we developed that rates KDA and farm. Each role has its own formula. It is the mathematical version of “12 / 1 / 3 is a good KDA”. We have not adjusted it for 5v5 yet… but on the other hand, I am doubting whether it is that useful at all. Most of our visitors turned out to be experienced enough to not need a machine to help them interpret a KDA.


Vainsocial stats meaning?