Vainsocial stats meaning?

Hi there, in the Vainsocial webpage, there are these stats of “Impact Score” in a game, “Mastery” of heroes and “Weakness” against certain heroes. I didn’t see any explanation about how they work in that webpage, so I thought somebody here might know how these stats are calculated. I understand what they mean overall, but what exactly does that calculate to come to that conclusion i.e. do i get a low “Impact Score” for not getting many kills or assists, or is that because i might have done little damage to the objectives and therefore contributed just a little to the game?
Any info would be appreciated.

@shutterfly might know a thing or two. :wink:

Re: Impact, here’s something I asked him a little while back:

From this thread:


@shutterfly should be able to give an official explanation …

Lol, @HipsterSkaarf … replied same instant. Yours was much more helpful, though.


Impact and trueskills in VainSocial is suck! Don’t believe them

In one match, I used Gwen CP and fought against Celeste CP. I won that. In VainSocial, they gave Celeste -39 Impact Score, just because the captain never place a ward and jungler never cover her, since she throw skill shot pretty good (never miss a B).

Those other people in her team just have more Impact score because support Ardan shielded himself, while WP fortress just bit me a little bit and then run away in a fast speed.


Want some statistic?

The impact score is a little questionable, considering a lot has changed since they developed the algorithm for it. But TrueSkill? Do you understand how that works? It’s very accurate, actually.

The Math Behind TrueSkill

Rather than simply trashing a web site (which is rather highly regarded among Vainglory’s analysts and coaches), how about elaborating on your reasons?

1 Like


This is my True Skill, but I am just talking about my nearest 3 matches.
The first match, I played Samuel, and I literally carry the whole team to victory. Steal jungle, kill heroes, gank frequently, etc.

The second match, I played Skaarf, because I played so much I felt asleep, and even I am main Skaarf, I fed like a noob.

The third match, I played Adagio, I never roam, I never ganked, build full damage and just… push. THe team win because Gwen carried us.

What I look in the graph here, it is totally like: I win, I have more True Skill, I lost, my True Skill drop.

Wow, just like elo! And we know about its accuracy already!

Hi, I’m one of the devs!

Mastery is calculated as the sum of impact score * (5% for a win, 2% for a loss) * (1% reduction for every day in past) * (0 if 60 days old). That means for a high score you need to play a hero regularly and win often with a good KDA (which is what impact score rates). Weakness is using the same score for the opposing team, so if you lose very often against a certain hero, the weakness is very high.

I will add an entry to our FaQ this weekend :slightly_smiling_face:

There have been some ideas for an improved scoring algorithm, but y’know, real life slows development down so our site always has a few rough edges. :wink: We’re usually chatting on Discord about ideas and progress, so if you want to contribute, you can just join us:


So, in the conclusion, the impact score depends on KDA and frequenly play. Right? Then why did you guys name it: Impact Score? All I thought is “Impact score will show that how much participation you make in your team victory, including pushing turret, killing hereos, etc.”

Moreover, I want you to look at this.

In here, we have my last match history. Taka left game after his first death and he had 44 Impact Score. Gwen had 80 Impact Score even she was the person who carried the team (both pushing and killing). And me just staying in lane, came back home for shopping for the whole 5 mins, teleported to lane to find out that… Gwen had pushed my mid to the enemy’s base! And I had … 106 Impact Score. So this mean, although we all win the game, Taka and Gwen had lower because… they not frequently play?

But this is Gwen’s match history.

100% sure the person played Gwen played her more than me playing Adagio.

Impact score is only on a match level — it is a value like KDA ratio. Mastery is on a player level, like win rate. Sorry for the confusion! Since you’re not the first one to ask, I’ll look how we can make it clearer.

What you are talking about is a general issue of rating microplay… we do not have the algorithms to do that yet. We can only look at post match statistics — kills, deaths, farm, gold, and a few extra attributes, and make a judgement based on that. It is statistics, so it is not accurate 100% the time, but fits close enough 90% of the time.
In your example, Taka had a significantly lower score and you and Gwen had good scores, which fits your general feeling about the match.
But it is never perfect, especially in unconventional situations like with an AFK player, so the difference between you and Gwen feels wrong because it maybe is.

Then how can you calculate and judge the Impact Score?

Now base on the match log, my KDA is (9+8)/1 = 17, while Gwen is (20+4)/1.
My damage is 31.6k, while her is 47.8K.
My farm is 82, her farm is 57, but since she is a jungler, we have to, 57 x 2 = 114. Get 11% of it (because there are cases jungler took farm from laners), we have around 100.
Her tier is lower to mine, so judging from the statistic, she must be the one who has high Impact Score than me.

There are so many reasons (statistic aspect) to prove Gwen is better than me.I don’t think Taka made the algorithm went wrong so much like that.

As an analyst for one of the top teams in SEA I use both.
VGpro I use for a quick and dirty match history - our scouting guy uses it to build a profile up on the teams that we play against - for example by looking at what heroes they typically ban and which heroes they play poorly on so we can try to manipulate draft to put them on the backfoot. So for example we noted that OfficialHein from Elite8’s winrate on assassin’s is much lower than his winrate on other junglers - so we would try to bait an assassin pick out of them.

Vainsocial however I use much more extensively because it gives much more useful stats about the meta, hero winrates and so on. So that tends (along with player opinions and our collective views about playstyle and hero synergy) to help us hone in on the heroes that are strong in the current meta and thus direct our drafts/bans to ensure we pick up or prevent the enemy having those heroes. We also use the builds data they give when theorycrafting potential build paths during the matches.


This is how impact score is distributed. Ideally, this should be a bell curve, because in nature, skills are modeled in a gaussian distribution most of the time.
X is the impact score and Y is the number of players in a match with that score.

This is the difference (y) between the average win rate and the average impact score (x) a player has. Ideally, it would be a flat line y = 0.

As you can see, you can predict the likelihood of a player winning a match very accurately using the impact score.
The score is calculated as a * kills + b * deaths + c * farm + d with a b c d being a constant factor for every role that we determined using a linear regression. It is more than 90% accurate.

The charts data is from a few months ago from 3v3 standard matches.