Ranked Valorant
July 28th, 2020
I’m going to be explaining my research into the Valorant ranked system using data and statistical analysis. The main goal for this project was to understand the inner workings of the competitive mode, theorizing methods for players to rank up faster and more efficiently by focusing on statistically proven aspects of the game.
I recorded the post match data from 50 games over the past 2 weeks.
Every number was recorded using the career tab in the main menu. I did not include rank up or derank games because the exact elo lost or gained is not displayed. I’ll have the google sheet linked below if you’d like to take a look at it in more detail. Once all 50 games had been taken down, it was time to analyze how each independent variable affected the change in ELO at the end of each match.
Let’s start with the objective based information. Looking at the graphs for times the spike was planted compared to the change in ELO, you can clearly see that there is almost no direct correlation between the two.
This doesn’t mean that planting the spike isn’t important though. You are more vulnerable when planting for those 4 seconds, but planting the spike gets you an ultimate orb, which can potentially get you more kills or victories. As you’ll see later, winning rounds is the defining factor in ELO progression, and not playing the objective in a game like Valorant tends to lose you rounds which otherwise could have been won.
The same thing applies to defuses. A very important thing to do but the individual player’s ELO does not directly benefit from the action. So, my advice for those wanting to rank up as fast as possible is to hand off the spike to someone who is most likely to successfully plant, unless you only need one orb for an ultimate. This should lower your time spent in a vulnerable position, and increase your time with your weapon drawn able to get frags. It’s a selfish strategy really but it’s how the ranked system appears to be designed.
Based on these graphs, it’s safe to assume that whether or not you’re being efficient with your money doesn’t directly affect your change in rank. This doesn’t mean you should go waste your credits however. An important aspect of the game is maintaining a good economy that can support yourself and your team. Won rounds are backed with funds, but with this data you can relax about accidentally buying that extra gun.
A hot topic in the community has been a lack of love for support characters. The players that get stuck smoking, healing, or finding information for teammates often are the ones who get the least elo. This becomes obvious when looking at the correlation between assist and change in rank.
It’s tricky because you don’t want to create a system that allows players to boost their mmr simply by getting a bunch of assists, but it’s ridiculous how it barely seems to directly affect your rank if at all. Teamplay and assists are important to winning rounds, but looking at the situation from an individual's ELO, you’re better off playing frag characters that get kills.
Being the first person to get a kill each round however appears to have very little importance to riot.
Although an opening pick can win you the round, it doesn’t seem to matter when it comes to your mmr. This promotes a common theme of baiting in the competitive scene. Because if the extra risk taken to snag the first kill is not rewarded with any extra compensation, then most players are likely to let their teammates walk out first, often getting the trade and the overall elo benefit. I’m not saying that the players who get a lot of kill shouldn’t be rewarded, I just think the current system needs some refinement in how that elo is dealt out.
A number most people don’t even realize exists until they stare at the career tab while waiting in queue is the individual score earned each game.
This seems to have no direct effect on your elo because it does not take into account rounds played. A 13-0 win will often award a lower score than a 13-13 draw.
You would think Riot would punish the players who died frequently. There’s a difference between those who are entering for their team and those who simply run in alone. No one likes the dude who jet dashes over a wall without their team only to get one kill and is instantly traded. But looking at the recorded data, it becomes obvious this isn’t important to ELO.
Ideally you would see an inverse relationship here. This is somewhat true but it doesn’t seem to be a constant. I had games with two up arrows where I nearly died 20 times. I understand that someone needs to enter the site first, and oftentimes they are the ones who die. But what ends up happening is people play selfishly, making dumb descions that will get them one kill with very little benefit to the team without a hit to their ELO.
When looking at the data for kills, it looks like a shrunken version of the ELO graph. There are some exceptions, but most of the time it is pretty clear that kills stand to be one of the most important factors when ranking up. And that is certainly not a bad thing, because players should be rewarded for this. But when the players who play support roles, assisting those who get the high kill games, don’t receive similar elo progression then I think that’s a problem.
The average kills per round pretty much just matches what I’ve already said for the general kill analysis.
The average damage per round is interesting though, because with players learning brimstone molly lineups and shock dart arrows, you think this assisting damage and zone clearance would award more ELO.
But looking at the graph you can see it doesn’t really matter (in terms of direct ELO correlation) if you’re doing damage for your team or not. This system is good for punishing players who can’t finish their kills, but it also rewards those who bait out their fellow teammates by finishing weakened enemies for their personal gain.
Your combat score is defined by your damage and kills. You can see the basic relationship between these two graphs. There’s nothing really specific to focus on gameplay wise when it comes to this besides getting as many kills as possible.
And finally, the most important factor when ranking up is the Score Gap.
These two graphs almost look the exact same. This ELO system promotes winning the game by as large a margin as possible over all else. Winning rounds is the priority and Riot wanted to make it clear that you’ll rank up fastest by doing so. But as an individual this is nearly impossible to control and means you have to rely on your teammates. In 5 stacks a lot of the problems listed aren’t an issue because if your team wins by a lot you get very similar ELO, but when playing solo you often can’t trust your teammates so most people just focus on getting kills because it’s all they solely can control.
In conclusion, we’ve looked over each independent factor and how it affects one’s mmr. So, the best statistically proven strategy for ranking up based on the data collected (at least in a solo queue) should be to play a fragging character, bait your teammates for kills, never plant or defuse the spike unless necessary or selfish, don’t worry about deaths as long as you get kills, and surround yourself with suckers who will play support characters who just accept the fact they will get less ELO for it.
The fact the system is like this makes sense, because it rewards those more skillful to higher ranks, but the lack of support for those not playing frag characters is frustrating.
In addition to this report, you can watch the YouTube video to for a more cohesive and easier to follow explanation.
This was a great project when I was younger and just getting into stats. Now that I have complete my masters in Statistics I'd love to go back and do further anyalsis because just adding some ANOVA tables would go along way in terms of a well rounded anaylsis. This was also before they added actual ELO numbers instead of the arrows they used in the beta, so some more granularity in another test would be a great next step. Lastly, an automated data gather to allow for quicker, easier, and more data collection would be ideal for next time.