Kills/Deaths (K/D) ratio is usually recognized by people as an indicator of how good a player or a team performance was, and it is a solid indicator in around 75% of situations. However, it is faulty and unfair in other situations, so with that in mind, we wanted a more solid indicator, so we introduced the Rating.
What's wrong with K/D ratio?
K/D ratio has several downsides, mainly, it favors players who play fewer rounds and therefore have fewer deaths, giving them high K/D ratios. That is why it isn't really useful for comparing performances from different matches. For example, Player 1 had 13-6 (K/D = 2.2) in a match of 18 rounds and Player 2 had 37-17 (K/D = 2.2) in a match of 30 rounds. K/D ratio would suggest they had a similar performance, while Player1 actually had an average game with 13 kills in 18 rounds, with 5 rounds with 1 kill (1K rounds) and 4 rounds with 2 kills (2K rounds), and Player2 had a great game with 37 kills in 30 rounds of a close match with nine 2K rounds, four 3K rounds a one 4K round.  If we apply the Rating on the two performances, we get that Player1 had 1.28, while Player2 had a much higher 1.91.
Another problem can be the fact that K/D ratio doesn't have an actual limit, since it can go from 0 to infinity. That makes it hard to compare performances from a single match. For example, Player1 has 22-4 (K/D = 5.5) in 18 rounds, and in the same match Player2 has 20-9 (K/D = 2.2). K/D ratio would suggest that Player1 had a much better game than Player2, while that wasn't actually the case, as he had only 2 kills more, and 5 deaths less. Rating shows Player1 had 1.87 and Player2 had 1.56, suggesting Player1 had around 20% better performance.
How is rating calculated?
Rating is calculated by comparing a player's average stats (kills per round, survived rounds per round and a value based on number of rounds with multiple kills) to the same averages for Counter Strike in general (which we calculated over a period of time). For example, average number of kills per round is 0.679, so if a player had 22 kills in 18 rounds, or 1.22 kills per round, he would have a Kill-Rating of 1.80 (1.22/0.679), meaning he did 80% better than expected on average in the kills department. In the same way a Survival-Rating is calculated, as well as a RoundsWithMultipleKills-Rating for which you can see the formula below. These three values are then added together, with Survival-Rating participating with a 0.7 factor (as surviving a round is less important that getting a kill) and then divided by 2.7 to give a 1.0 value for an average performance.
Note: Survived rounds are used instead of deaths, as deaths are reversely proportionate to other stats (having less is better, and more is worse).
Advantages of Rating
Rating disregards the amount of rounds played in a match, as it considers average values, so it is very useful for comparing performances. It has a limited range, as it can go from 0 to 3 (although rarely over 2, which is then an amazing performance). It also has a well spread range of values that should reflect properly on how many average, good and great performances there are.

So here is the formula for calculating the Rating:
(KillRating + 0.7*SurvivalRating +RoundsWithMultipleKillsRating)/2.7

KillRating = Kill/Rounds/AverageKPR
SurvivalRating = (Rounds-Deaths)/Rounds/AverageSPR
RoundsWithMultipleKillsRating = (1K + 4*2K + 9*3K + 16*4K + 25*5K)/Rounds/AverageRMK

AverageKPR = 0.679 (average kills per round)
AverageSPR = 0.317 (average survived rounds per round)
AverageRMK = 1.277 (average value calculated from rounds with multiple kills: (1K + 4*2K + 9*3K + 16*4K + 25*5K)/Rounds)

1K = Number of rounds with 1 kill
2K = Number of rounds with 2 kill
3K = Number of rounds with 3 kill
4K = Number of rounds with 4 kill
5K = Number of rounds with 5 kill