So as it comes to year end, I wanted to refresh my interest in presenting lovely new stats and numbers to the people of Vlr. What better way to do this than by inspiring controversy and discussion about the validity of stats and about who the best player in North America is. Hint, hint BATCHEST Tenz BATCHEST.
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
[First of all I need to clear up some stuff that might be unclear or seem strange.]
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
- I only included the stats of players since their team's last roster change because that seemed the fairest given the end goal of showing the most statistically impressive player. This does unfortunately mean that teams like Sentinels have loads of data whereas a team like Cloud9 or XSET may have less which may affect the rankings. Honestly hard to know...
- The above also means that it became incredibly difficult to work 100 Thieves players into the stats because they played so few matches with b0i and I found it hard to justify working around the time period where they played with steel. This does unfortunately mean that Asuna is not included but I have no doubt in my mind that he would have been pretty high up and you can just imagine him in the final rankings.
- THESE RANKINGS DO NOT TAKE INTO ACCOUNT INTANGIBLES OR THE ROLE OF IGL, OR EVEN REALLY THE ROLE THAT THE PLAYER IN QUESTION IS FILLING FOR THE TEAM. SHUTUP, EAT THE NUMBERS AND DON'T COMPLAIN ABOUT HOW stats don't tell the whole story BECAUSE I ALREADY KNOW THAT
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Now, I tried to calculate all of my stats using readily available numbers, partly because I couldn't be bothered to source hard-to-find stats and partly so that this method could be replicated by somebody far more experienced or educated in the field then I am.
With all the waffle aside, let's crunch some numbers
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
The first stat I decided I was going to use I have affectionately called "Statistical Importance Rating" i.e. to what extent do a given player's stats make up the majority of their team's stats.
To do this, I calculated some "engagement" values which I primitively calculated as being:
Kills + Deaths
There is 100% a better way to do this but I worked with what I had.
With that calculated I calculated Kills per engagement, deaths per engagement and assists per engagement values for each of the players and then averages for the team.
The formula for unadjusted Statistical Importance Rating is as follows:
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
uaSIR = ( ( 1 / Rounds Played) x ( 0.5 x ( ( First Kills / Rounds Played) ) + ( ( ( Kills per Engagement / Team Average Kills per Engagement) + ( Assists per Engagement / Team Average Assists per Engagement ) x 0.5 ) - ( Deaths per Engagement / Team Average Deaths per Engagement)) x Engagements per Round) ) ) x 100000
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Then calculate a team adjusted SIR value by dividing the output of the above by the average of the team's unadjusted SIR values and multiply it by 10 just to give a nicer number
To make this value comparable across all teams I had to calculate engagement rates or paces for each of the team's I was calculating for and then adjust the team adjusted SIR values so that they reflected what the players would be like if every team were functioning at the same engagement pace. Just calculate each team's engagement pace as a percentage of the region average and then multiply team adjusted values by that percentage divided by 100.
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Consistency of Performance
The consistency of a player is a large part of what makes someone statistically impressive and I decided, therefore, that was going to help me decide.
To calculate this, I simply calculated the standard deviation of both the ACS and the ADR of a player and took an average of the two values. This output was once again region adjusted based on how far above the average value it was.
I then combined a quarter of this value with the Statistical Importance Rating to provide my first metric for my calculation.
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
First Kill Demon Rating
I found that the first metric I had calculated found a way to ever-so-slightly favour supportive roles due to its assigned values on deaths and assists alongside the fact that players on these supportive roles do tend to be more consistent.
With this in mind, I decided that a metric favouring the demonic duellists wouldn't be too overpowering for the overall rankings.
This calculation is relatively simple:
FKDR - ( First Kills / First Deaths ) x ( ( SUM ( First Kills, First Deaths ) / Engagements ) ) x 10
I didn't make any adjustments to this rating as it is one of the most individual metrics and ranked it accordingly
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
ACS Correlation to Victory
The final metric has literally the worst name. If anyone can think of a snappier name to give it that would be great.
For this metric I simply used ACS values, which I genuinely believe is a pretty good metric already for solely statistical impressiveness, and conducted a Spearman's Rank correlation test against the wins and losses of the team in that game. This, in a very primitive way, tests the impact that a single player's performance has on the outcome of a match.
I then calculated a score for the metric by doing the following
Score = ( 1 - ( Spearman's Rank coefficient for 0.001 for number of values in respective dataset - Spearman Correlation value for individual player ) ) x 100
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
The results of all of these metrics will be in the next part
Sorry, I was worried about the character limit
Link to part 2...
https://www.vlr.gg/61671/the-most-statistically-impressive-players-in-na-2
Link to new and improved rankings
https://www.vlr.gg/61824/the-most-statistically-impressive-players-na-2-0
Discord