The Mystic Realms of Valorant.
Embark on a journey through a fantasy map that captures the essence of the 2023 VCT Season, highlighting the diverse game-styles of each player.
Valorant is a game with clearly defined roles, yet they don’t dictate a player’s playstyle. In the 2023 Season, we’ve seen a lot of different ways to play the game, such as the fast and aggressive system shown by Paper Rex or the innovative and adaptable method proposed by Evil Geniuses. Inspired by NER0cs’s article “The four spheres of riflers in Counter-Strike”, I’ve analized the performance of the VCT players and conducted a cluster group analysis to create a fantasy map. Its goal is to learn more about the individual game-styles of the players throughout 2023 in a amusing and mystic manner.
For the creation of the map, I gathered much individual data and applied a nonlinear dimensionality reduction technique to embed the data into a 2-dimensional space, enabling visualization in a graph. The realms were defined using a clustering algorithm. The last section will provide detailed information about this process.
The resulting Valorant fanstasy map:
The players are categorized into six distinct realms, each distinguished by a different color. Let’s hence proceed to visit each realm identifying which statistics distinguish them from the rest, so we can create an average player profile.
Realm 1. Situated in the southern region of the map, this realm predominantly houses duelists. Their role in the round is to entry the site and make space for the team. Consequently, these players have a higher entry rate than the players in the other realms. The distinguishing stats for them include the following (with the average of all players in parentheses):
- 0.17 first Kills per Round and 0.14 true first Kills per Round (0.09 and 0.08), 0.13 Operator Kills per Round (0.02), 0.20 Kills under 30s per Round (0.14)
- 0.22 assists per round (0.30), 0.08 non-damaging assisst (0.12), 0.02 Plants per Round (0.07)
Demon1 resides alone in his unique castle separated by the Skill Gap and located amidst the Aimer’s Sea.
Realm 2. To the east and following the Aimer’s Sea, one can find the next kingdom in a distinguishable white land. This realm shares some space with the realm 6 and houses some Sentinels in the Stronghold Archipielago such as Alfajer or SUYGETSU. It is known because of their Skill at the rifle fights, in particular with Vandal (region with the highest % Usage).
- 55% Rifle Win Rate (50%), 0.65 Deaths per Round (0.69)
- 85 seconds until first Aggression (79s), 150 Damage per Round (133)
- 88% Vandal Usage (66%).
Realm 3. Heading northward reveals the next realm located in a unique island, the Phantom Abyss Isle. This island harbors the vast mayority of Phantom players and features the Initiator’s Grove which divides the right edge of the island where Leo resides in solitude.
- 61% Phantom Usage (34%), 39% Vandal Usage (66%)
- 20% Clutch Rate (13%), 0.38 Assists per Round (0.3), 0.2 non-damaging Assists per Round (0.12)
Players in this realm are recognized for their high clutch success rates and for a good ability usage providing non-damaging assists to their teammates. Hence, there are a lot of initiators living in this realm such as Leo, stax, d4v41 or crashies.
Get Pablolopezarauzo’s stories in your inbox
Join Medium for free to get updates from this writer.
Realm 4. The next kingdom is located in the northern part of the map and is the most populated one. It is known for the big mountain located on the middle of the region, the Planting Hill, since this realm houses the majority of planters in VCT such as Ethan, Chronicle, S0M, rhyme, Zyppan or cauanzin. In the northerneast side of the island one can find the Flex Coast. Some of their distinguishing stats are the following:
- 0.11 Plants per Round (0.07), 0.024 Clutches per Round (0.016)
Realm 5. Crossing the Strategist Sea brings us to the fifth region, situated to the east of the map. The inhabitants of this region reside around a big cavern, known as the IGL’s Depths. Many IGL’s such as ANGE1, Saadhak or Bonecold live in this isle. Hence, their statistics are known to be worse than those of players in other regions:
- 0.35 Assists per Round (0.3), 0.18 non-dam. Assists per Round (0.12)
- 45% Rifle Win Rate (50%), 7% Clutch Rate (13%), 0.009 Clutches per Round (0.016)
Realm 6. The last realm is located in the center of the map, surrounded by the other five kingdoms. Here you can find the house of Controllers (e.g. pAncada), Sentinels (e.g. Rb or Shyy) and more flexible players (e.g. jawgemo or Boostio). It is separated from the northern region by the Controller’s Channel and from the southern realm by the Aggression Gulf. The players in this island adopt a more passive game-style:
- Average death after 60s (56s), 83s until first Aggression (78s)
- 0.72 Deaths per Round (0.69), 0.12 Kills under 30s per Round (0.14)
This fantasy map, featuring six regions, allows us to gain a better understanding of the diverse game styles in Valorant and provides a player classification based on their individual performances.
The Process of obtaining the 2-D visualization and the clusters.
Initially, the process involved gathering data from VCT players who had played more than 500 rounds during the 2023 Season, amounting to approximately 106 players in total.
Since I wanted to represent the diverse game-styles in a 2-D plot, I decided to obtain the folowing statistics for each player:
- First Kills, First Deaths, True First Kills, True First Deaths
- Kills under 30s, time until first Aggresion, Average Death Time
- Assists, non-damaging Assists
- Vandal and Phatom Usage, Operator Kills
- Clutches done, Clutch Success Rate
- Average Damage, Average Deaths
- Spike Plants done
- Win Rate in rifle fights
After obtaining (from rib.gg) and standardizing the data (a preprocessing technique to normalize the data and ensure equal importance), I assigned a higher weight to the Plant and Rifle Win Rate features since I wanted them to be a slighty more relevant.
Having the high-dimensional data, I used t-SNE to reduce the dimensionality such that each player is represented by 2 single values. The new data can be hence portrayed in a 2-dimensional plot where closer points indicate a higher degree of similarity. The resulting graphic:
The final step is to define the regions in the map. For that, I employed the well-known K-Means algorithm, a clustering method designed to classify the data into distinct clusters based on similarities and patterns. With 6 as the number of clusters, the algorithm classified the data as following (the crosses represent the cluster centers/centroids of the groups):
Overall, this was my version of a Valorant fantasy map. The process of creating it was immensely enjoyable, and it provided me with insights into the players and their unique performances.
I hope you can enjoy the result as well. Feedback is greatly appreciated!
My socials: Twitter.
