Why K/D Ratio is Deeply Flawed, and How We Can Make it Better.

What K/D Tells Us, and How We Can Better Measure It

In Call of Duty, for years a players KD Ratio is how we have determined who the best players are. It's not hard to figure out why it is popular. Does a player kill more players than he dies? That is what K/D ratio answers. On the surface level it seems like it would be a very productive stat, but in terms of proving who is a better Call of Duty player, it does not paint an accurate picture.

So why is KD Ratio flawed? Simply, there are an endless amount of different variables that go into how well a player performs when it comes to kills and deaths. Is the player running an AR or a Sub? Is the player on a winning team or a losing team? Is the players engagements low or high? Is the player always rotating for spawns, the first to break a hill? Is the player always soaking up hill-time?

The best skilled player in the game could be running a sub, always pushing for new spawns, playing on a losing team and even though he may be the most skilled. His KD ratio is never going to come close to someone who may be running an AR, playing on a winning team, and engaging at a low frequency around the map.

So how can we make K/D ratio more accurately reflect who the best slayers are in the game, while still keeping the simplicity that makes it such a popular and great stat?

When looking at the issue of figuring out a better way to evaluate a players kills and deaths. We have to take all the data that is available. For myself, it is tracking what gun every player is using, hill-time, engagements, and team success. For now, this is the data and variables we have to work with, and the most I can track on my own. As we get deeper, and we hopefully get access to more data (For example: Engagements on rotation) we can make this stat more and more accurate.

The first step when working with the baseline data is to make it better. All the numbers I work with in my advanced stats are first adjusted to the map that is being played (certain maps average more engagements, etc. than others --- more to come on this in a later article), as well as opponent (obviously not all opponents are equal). Finally the numbers are broken down on a per minute basis (not all Hardpoint game type's are the same length).

With this data, we can now take a players gun they use, engagements per minute, hill-time (hardpoint), and the net win/loss of their team and run a linear regression to find a players projected Kills Per Minute and an expected Deaths Per Minute. We can then average out the results to get every players projected K/D vs a league average opponent on a neutral map for the entire season. 

Now that we have a more accurate idea of what every players K/D SHOULD be for the role they are playing, we can then look at what they have ACTUALLY produced on the season in terms of Kills per Minute and Deaths per Minute. (Adjusted for map and opponent).

By using this formula: (Adjusted Actual KD - Adjusted Projected KD) + 1. We get a more true reflection of who the best players are when it comes to slaying. Example: Player A has a 1.04 projected KD and a 1.12 Actual KD. We take (1.12-1.04) + 1 to get an adjusted KD ratio of 1.08.

While this is not an end-all be-all stat, it allows us to look at and reflect on a players slaying power in a more accurate way. Obviously, this is still going to reflect poor on players who are always doing the dirty work. Ex: always in tough engagements, first to a hill, pushing for spawns, etc. But as more data becomes available/tracked, the stat can improve and become better.