How much is a kill worth?

Part 1

When it comes to playing Counter Strike, the value of your actions is a mystery. Not even delving into the many nuances of the word itself, it’s not very clear what value even is in Counter Strike. 

You might be tempted to look at the in-game economy, as the words “value” and “economy” go hand in hand, and you would be completely justified in saying that. However, while observing the economy in real time isn’t so difficult, beginning to think about quantifying how “value” works in the context of the economy is a canyon sized rabbit hole that I can’t even begin to navigate. 

And even then, there’s a greater argument to be made that favors a different mode of thinking. It’s the fact that to win a game of Counter Strike, you just need to win 13 rounds. It doesn’t matter what your economy is in those rounds, you could win all 13 of those rounds without buying anything in-game. All that matters is that your team has 13 round wins by the end of regulation.

So this is the approach we’ll take to quantifying value in Counter Strike. In this series of articles, we’ll be re-examining the statistics of Counter Strike and attempt to reformulate them in a way that both captures and expresses value in a way that will redefine how we evaluate players.

Introduction to KPR

One of the most ubiquitous Counter Strike statistics is “Kills Per Round”. It’s extremely simple to understand and calculate. As the name suggests, all you have to do to calculate it is to take total kills and divide it by total rounds played. Despite how simple it is, it’s also extremely effective. Just for an example, here are the top 5 players at big events in the first half of 2024 by KPR (min. 30 maps):

Yeah, it’s pretty good on its own merits. However, KPR has one fatal flaw. That, quite simply, is that KPR treats every kill equal to every other kill. As Counter Strike players we know, not all kills are created equal. So, let’s re-evaluate KPR. In doing so, we will express KPR in a more digestible way, weigh KPR in a way that’s proportionate to winning, and express those results in a way that showcases value added over time. 

Re-examining KPR

Earlier, we introduced KPR as taking your total number of kills and dividing it by your total number of rounds played. That’s the easiest and most straightforward way to calculate it. However, for the purposes of today we’ll need to think about KPR slightly differently. 

In any given round of Counter Strike, there are, in a basic sense, six types of contributions you can make. Those are either:

Getting 0 kills, 

getting 1 kill, 

getting 2 kills, 

getting 3 kills, 

getting 4 kills, 

or getting 5 kills. 

So, if we want to break down KPR in terms of “round contributions”, KPR is calculated as so:

(0 * 0K) + (1 * 1K) + (2 * 2K) + (3 * 3K) + (4 * 4K) + (5 * 5K)

Divided by Rounds Played

However, as discussed earlier, these kills (and by extension round contributions), are not created equal. In other words, an ace isn’t five times as valuable as getting a single kill and a quad kill isn’t twice as valuable as a double kill. So now the question is, how do we figure out how valuable these kill contributions are in relation to one another. 

Well, it’s actually pretty simple. To find the value of these round contributions (which from here on out we’ll call “Round Value”, let’s examine the win rates of these contributions from an event and from there we can extract their value. Conveniently enough, fairly recently there was an event with a bunch of teams of similar strength that all played a shitload of matches.

These are the win rates for every type of kill contribution at the Europe RMR A:

The only big surprise here is that aces had a lower win rate than quad kills, and that’s only due to Wicadia getting the final kill of his ace on Overpass after the enemy had defused the bomb. Generally, you would expect that win rate to be higher than for 4ks.

But for the other win rates, how consistent are they between different events? Well, luckily for us, there was another event that happened right after this one that featured a similar number of matches and a similar level of competition.

And here are the win rates for Europe RMR B:

Turns out the win rates are fairly consistent for 0-2Ks between events but there’s greater variance between other contributions, which makes sense because of how small the samples are for each. 

Across both RMRs, there were only 20 aces total. There were 120 quad-kills and 719 triple-kills. With more and more events the round win rates should stabilize, but it's encouraging to see how stable they are for the lower level kills. 

What’s most important to note overall though, is that every kill you get in a round offers diminishing returns on your chances to win a round. In other words, going from a quad kill to an ace is not as valuable as going from no kills to one kill.

In order to express the value of each type of multikill, it’s pretty simple. We can just take all of our percentages and subtract the win rate for 0ks (38.02%) so that a 0k is worth 0. Then, we can divide the remaining percentages by the value of a 1k to make it so a 1k is equal to 1 on our scale. If we do that, we get the following values:

The primary conclusion that can be drawn from these values is that, in a context neutral environment, multi kills aren’t as valuable as a stat like KPR would suggest. Every kill you get in a round after one offers diminishing returns in terms of increasing your team’s chances to win a round.

KPR v wKPR

Now let’s clarify, this does NOT mean that multi-kills are less valuable than single kills. The graph does show the round value of every multi-kill increases as you get more kills, which means with the more kills you get in a round, the more you increase your team’s chances of winning the round. But what this does answer is an interesting hypothetical question, which is as follows:

Let’s say you have 5 different players with the exact same KPR, but they all have distributed their kills differently. One player gets a kill every round, one player gets two kills every OTHER round, one gets three once every three rounds, etc, etc. Which player would you rather have on your team? 

Intuitively, the answer most people seem to steer towards is the player that gets one every round. Asking this hypothetical to 16 people familiar with the game, 10 correctly guessed the first option, with the other six votes being split among every option aside from four every four (poor four). 

Using the win rates we found earlier, we can use them to calculate this. Let’s say these five hypothetical players played 60 rounds, let’s break down how many rounds you would expect each player to win (with no other context given).

Now, the differences are very slight but there is a noticeable difference. In fact, for a player that gets an ace every 5 rounds, their expected round win rate would be less than 50% and would be nearly 8% less than the player who gets a kill every round. That may not seem like much, but in a normal match that’s the difference between winning 13-10 and losing in overtime, and those are players with the exact same KPR.

So while KPR is a good context-neutral statistic, we can clearly see that it has its flaws. But now that we can highlight one of these flaws, let’s examine the RMR performances of the top five players by KPR and see if we can re-contextualize their performances.

If you’re confused by KPR+, don’t be, it’s extremely simple. KPR+ is just a player’s KPR compared to the average KPR of all players multiplied by 100. The average KPR for the RMRs was 0.657, so to calculate ZywOo’s KPR+, you do 1.016 divided by 0.657 which is 1.546. Then, when multiplied by 100 and rounded to the nearest whole digit is 155. In other words, that means for the RMR, ZywOo’s KPR was roughly 55% above the average player’s KPR.

Now, we can compare these numbers to the adjusted KPR that we made and see what changes.

Unsurprisingly, these numbers aren’t that different. They’re just lower overall, which makes sense because the values for all of the multi-kills are lower than they would be in KPR. As well, this does mean the average wKPR is lower at 0.626. wKPR+ is calculated the same way as KPR+.

What perhaps is most surprising is that ZywOo goes from an 8 point lead in KPR+ to a narrow 3rd place using wKPR+. This is because of his disproportionately high numbers of quad kills and aces compared to the rest of the field. fame, who he basically ties in wKPR+ had only one such instance in comparison, leading to more of his round contributions having more value.

So while the differences between KPR and wKPR are minimal, it can still impact the perception of a player’s performance. One way we can also examine these findings is by expressing this as value added over time.

Round Wins Above Average

Earlier, we calculated expected win rates to answer a hypothetical. Now the question is, why does that have to be hypothetical? We can apply the exact same calculation to the sample we’ve collected and see how many rounds that player’s performance is worth.

Here, we can see how many rounds each player was worth cumulatively, while the wKPR+ stat from earlier is on a rate basis. That’s why, despite ranking only third in wKPR+, jkaem has the most “Round Wins”. 

What we can also do to assess individuals is look at the round wins a player had above average. This is pretty simple to calculate, because the win rate of every individual in a game is 50% because every round, 5 players win and 5 players lose. So you just take an individual’s “Round Wins” and subtract their rounds played divided by two. This gives us how many rounds the player “won” compared to the average player, otherwise known as RWAA (round wins above average).

What RWAA can be used for now is for evaluating player performance over the course of an event. This touches on the idea presented earlier of what “value” means, when specifically talking about MVP discussions. Usually when determining who the MVP of an event is, the question really is “Who made the final that has the highest HLTV rating for the event”.

So let’s see what Round Wins Above Average would say the most valuable player for each event is and see how it compares to HLTV’s MVPs and EVPs. Note that the round win percentages for these events were not calculated, I’m just using the RMR win rates for the purposes of this but in a perfect world I would use the sample size of all of these events. Sadly I don’t have that time right now. However it should be similar enough to where it isn’t a huge deal, as shown earlier with the differences between the RMRs.

Well they agree pretty decisively here, donk was head and shoulders above everyone else. I mean, this is just insane. donk was twice as valuable as m0NESY in less rounds played, his performance was truly once in a lifetime. 

It’s also interesting to note that while frozen and hades had very similar HLTV Ratings, hades was able to accrue as much RWAA as frozen in 100 less rounds than him. This is primarily due to the fact that RWAA, as it stands, only looks at kills and HLTV Rating takes into account kills and deaths, and frozen is somewhat of a notorious KD warrior. 

jL did narrowly edge out m0NESY in terms of RWAA, and it only took him 90 more rounds played. Also shout out to donk who was eliminated in the quarters and still earned 8.1 RWAA.

Our first disagreement! Well, sort of. HLTV gave the MVP to broky, while the player which had the highest RWAA for the event was m0NESY, and he did it in 28 less rounds than broky. With that being said however, it’s still understandable why broky was awarded the MVP given that he both won the tournament and played more maps than m0NESY in the playoffs. Still the disagreement is noteworthy.

Not much to say here, ZywOo ran away with it.

This is the first event where the EVPs were listed in the same order as RWAA, so that's nice.

Not much disagreement here aside from mezii, who is mainly here for his playoff performances. Although his performance is not too dissimilar to some other EVPs listed earlier.

Funnily enough, in a version I wrote before the EVP article was released, I was able to guess the correct 6 that were in the article, however I had ZywOo as second, sh1ro as third, flameZ as fifth, and zont1x as sixth. The order listed here is the same as is listed in the EVP article. 

Also just to note, I checked device and broky’s stats just out of curiosity and in 125 and 175 rounds played, they accrued 3.9 and 3.0 WRAA respectively.

And finally a leaderboard of the fifteen most valuable players at Big Events (HLTV filter) this year up until the summer player break:

I’m actually surprised by how well this works. If I had to make a top five right now for 2024, the top five this has pretty much nailed it for me. It goes to show how impressive donk is, that even with a 400 round deficit compared to ZywOo, he’s still accrued (slightly) more RWAA than him. 

The name that’s perhaps the most surprising here is TeSeS, but that’s just more the fact HEROIC has been flying under the radar this year. Also props to xertioN who has risen to bona fide star status, being 5th in HLTV rating and ranked 6th so far by RWAA.

Also just for fun, here’s the bottom 5 (of those that have played 30 maps at big events this year):

Yeah this checks out. Aleksib hasn’t quite had the reputation of the other IGLs here but HLTV rating very much agrees with the sentiment. nexa being 5th is perhaps a bit surprising, but he’s also tied for the most rounds played at big events this year. The next lowest IGL is advent at 6th worst with -15.5 (in only 282 rounds!!!) and the next lowest non-IGL is floppy at -11.5.

Closing thoughts

So this article kind of turned into something completely different than the initial premise. I had created the premise and compiled all of the RMR data back in March but I got distracted with stuff and only now did I get around to finishing this article. I had finished writing about my RMR research and ultimately found the conclusion to what I was wondering, which is just that the more kills you get in a round, the less impact each subsequent kill has on winning the round.

However, I think RWAA is a very promising concept that I’m going to continue to refine. Even though its only factor right now is multi-kills (and I probably messed up the math somewhere) it has some pretty promising results. With that being said, there are still so many ways to improve it that I’m excited to continue developing this idea. 

Just to name a few things, it doesn’t take into account deaths, opening kills, maps, sides, economy, or other things I haven’t even thought about. Not only are these a question of math, some are a philosophical question of whether they should even be accounted for in a stat like this. 

I will attempt to answer those questions in due time, but for now, this is what I’ve got. This is just part 1, and hopefully there will be many more parts to come! If you have any questions (or want to point out any mistakes I made), feel free to reach out.

Thank you for reading and I hope you have a good day.