We're pleased to introduce a new upgrade to our site in form of Rating 2.0, an updated formula for the Rating which we use to quickly assess player performance. It now incorporates data like damage dealt, opening kills, 1onX wins, traded deaths and more, and thus better encapsulates the different ways players contribute in the game.
We recently updated our entire website's design and introduced a number of new features, which as we pointed out was just the beginning of our new era and a foundation for future improvements. One of those improvements is Rating 2.0, a new version of our nowadays commonly used Rating.
What is Rating?
For those who are unfamiliar with Rating, it is a relatively simple concept based on statistics that we introduced back in 2010 to improve upon the then commonly used K/D Ratio with the hope that a quick glance at Rating gives us a better idea about a player's performance. It basically tells us if the player put up above or below average numbers, with 1.00 being the average.
And while it has been doing a good job across both CS 1.6 and CS:GO over the years, there was always room for improvement and we never stopped looking for ways to do that. Over the past year, we overcame some obstacles in data collection and processing, with the biggest step forward being the addition of ADR (average damage per round, introduced to our site in early 2016).
You also may have noticed the addition of KAST (percentage of rounds with a kill, assist, survival or traded death) to our site last month, which is a stat that is best described as round-to-round consistency. It helps us notice players who might not put up the big numbers but often contribute to their team in some fashion (a good example of this would be a player like Epitacio "TACO" de Melo).
And thanks to this various new data we've been able to create Rating 2.0 which we hope will give you an even better idea about a player's performance, whether in a match, event or across a certain time period.
What's new in Rating 2.0?
Rating 2.0 is based on the same principle as Rating 1.0 – we take the expected value (average) for certain statistics (like kills per round) and check how much above or below that expected value a certain player is.
The first difference is that we now calculate these values separately for CT and Terrorist sides, as expected values differ that way. The second is that instead of previously dividing by the average to find how much a player stands out (or not), we check how many standard deviations the player is above or below average, which produces more accurate results.
The average value for Rating 2.0 will still be 1.00, and it will behave exactly the same way as before – good performances will give ratings higher than 1.00, and bad ones below 1.00.
The biggest difference is that there are many more factors going into the calculation of Rating 2.0, including the above-mentioned ADR and KAST and much more. The exact formula won't be public this time, but here is how it compares to Rating 1.0.
It includes ten components, five per side (compared to three in total previously), which are:
Those components were chosen through research because they provide a good balance between favoring statistics that more directly lead to winning rounds (such as multi-kills, damage and just kills in general) and statistics that are a result of less impactful but consistent contribution (such as KAST and surviving).
Some of the five components have additional parts to them, such as Kill Rating being less rewarding towards kills that dealt less than 60 damage (assisted kills), or Survival Rating being less punishing towards deaths which were traded.
How does it look in practice?
The main goal in moving to Rating 2.0 was to provide a more accurate representation of a player's performance, especially to discontinue underrating players who contribute in less impactful ways while continuing to reward those who have high impact on the game.
And during our testing Rating 2.0 has done exactly that. Here are only a few examples, such as the match between Immortals and Astralis at IEM Katowice where Lincoln "fnx" Lau had an 18:20 scoreline and a 0.83 Rating. What the old Rating missed out on was that fnx also had 8 assists, dealt the most damage in the match (109.6 ADR) and contributed in match-high 79% of rounds, all of which describe that his contribution extended beyond just getting kills, which was also evident during the eye test.
And as you can see in the above picture, Rating 2.0 gives him a 1.14 mark, good for the 2nd highest in the team compared to 5th previously. And while the newly included data increased his rating, it still wasn't increased far enough to be above Lucas "steel" Lopes, who was the obvious Player of the Match.
Another example would be Andreas "Xyp9x" Højsleth's performance on Overpass during the grand final of ELEAGUE Major, where he was by most accounts the clear standout player of the map, but according to the old Rating his 1.25 ranked behind Janusz "Snax" Pogorzelski's 1.33.
That was in part because his 1vs3 clutch, 4 opening kills, and map-high ADR and KAST weren't included. With Rating 2.0, he is more accurately rated at 1.44, the highest in the map, whereas Snax remains around the same level with 1.31.
Of course, these are only limited examples to give you the idea about what Rating 2.0 is supposed to help with, while you'll have to discover yourself over time whether this is a notable change and an improvement or not.
Where can you find Rating 2.0?
We have updated all of our stats pages where we have the data needed for Rating 2.0, and those places are clearly marked as the 2.0 version.
That includes match pages as well, where you will additionally be able to find this graph in the "Player highlight" section, outlining the components that go into the Rating 2.0 calculation and how they compare to the average:
On the other hand, on stats pages containing data older than February 2016, as well as players' career stats which go that far back, we cannot calculate Rating 2.0 so we will display Rating 1.0.
Keep in mind that statistics and our Rating are just one possible interpretation of what happens on the server, and that for various reasons (like anti-eco rounds for example, or different importance of kills within a given round) they are not telling the full story, so it's always best to also watch the matches yourself for the most accurate assessment of a player's performance.
Having said that, we'll continue working on providing you as many as possible stats to help you digest the crazy world of competitive CS:GO, starting with the addition of match economy-related numbers in the near future.
As always, you can use the comments below to let us know if you find a bug related to the new Rating or if you think of new features you would like to see added in the future.