Raided Lost Ark

TL;DR

Click here to see the charts. We rank classes based on their DPS. The boxes represent a reasonable range for the typical DPS of each build. Compare classes primarily based on these boxes, farther right is better. The right whisker provides a reasonable ceiling, but do not expect most players to reach this. There are many factors we do not account for such as gear, player skill, and supports.

How to read

In general, you should compare classes based on their box (the colored part), as that's where a majority of players will fall, in terms of performance. Note that we do not account for differences in gear outside of ilevel (which means tripods, gems, elixirs, and transcendence will affect this). Indeed player skill matters quite a bit. If you would like to compare performance with more similar gear, you can try to reduce the ilevel ranges in the advanced options to cut out certain progression systems. Be careful of the low sample sizes when doing this.

If you would like to look at a "reasonable ceiling" performance for each class, the right whisker provides a reasonable estimate. Note that this "reasonable ceiling" does not mean you cannot perform above it, this is exemplified by the stars that show the literal best performance given the filters.

Method

We scrape data from Faust's Lost Ark logs every night and classify each player in the logs based on their build. We produce box plots using this data. The boxplots follow the typical procedures. It should be noted that the whiskers [Q1 - 1.5 * IQR, Q3 + 1.5 * IQR] are truncated to the minimum and maximum values for that build to guarantee that it represents a possible value. The line in the middle represents the median and the dot in the bar represents the mean.

The default filters, in general, seek to provide a representative dataset for on ilevel players in the typical situations in the current balance patch.

To ensure we do not use data that is not representative of the typical situation, we classify logs as "Weird" if they do not have the expected number of players ( 4 or 8, depending on content), if they have a Princess Gunlancer, if they do not have the expected number of supports (either none or more than 1 per party), if we fail to classify a player's build, or if we detect any Sidereal weapons. Additonally, we filter out players that were dead when they cleared. These filters are on by default but can be toggled in the advanced options.

There is no guarantee that we classify builds correctly but we do our best with the following procedures:

Build classification criteria Most builds are binary under a class, so if a player doesn't meet the criteria, they're the other build. All builds (except for Princess Maker) are named after their main engraving.
  • Berserker: Checking for the Mayhem buff
  • Destroyer: Checking for the special Gravity Training weapon attack
  • Gunlancer: If they do less than 5% damage, they're Princess Maker; if they have Nightmare or Hallucination set they're Combat Readiness
  • Paladin: If they do more than 10% of the team damage, they're Judgment.
  • Slayer: Checking for the Predator buff
  • Arcanist: If they did damage with the card Emperor
  • Summoner: If they did damage with the skill Kelsion (Communication Overflow)
  • Bard: If they do more than 10% of the team damage, they're True Courage
  • Sorceress: Checking for the Igniter buff
  • Wardancer: Checking for Esoteric Skills
  • Scrapper: Checking for the unique Shock Training buff
  • Soulfist: Checking for the unique Robust Spirit buff
  • Glaivier: Checking for the Pinnacle buff
  • Striker: Checking for the skill Call of the Wind God (Esoteric Flurry)
  • Breaker: Checking for the special Asura weapon attack.
  • Deathblade: Checking for the unique Remaining Energy Death Trance doing more than 10% damage
  • Shadowhunter: Checking for Demonic Impulse buff
  • Reaper: Checking for the Lunar Voice buff
  • Souleater: Checking for the Soul Snatch buff (Night's Edge)
  • Sharpshooter: Checking for the Loyal Companion buff
  • Deadeye: Checking for the Enhanced Weapon buff
  • Artillerist: Checking for the Barrage: Focus Fire skill doing more than 10% damage
  • Machinist: Checking for the Evolutionary Legacy buff
  • Gunslinger: Checking for Sharpshooter skill (Peacemaker)
  • Artist: If they do more than 10% of the team damage, they're Recurrence.
  • Aeromancer: Checking for the Sunshower synergy buff on their own Sunshower skill (Wind Fury)

Whenever scraping methods are updated, it is unlikely previous logs will be rescraped with new methods. Check the changelogs to see when scraping procedures have changed (e.g., Sidereal weapons being classified as weird was added in June 2024).

The data scraping is written in Python and the site/visualization is written with Javascript in the Observable Framework. Data scraping and deployment is automated through Github Actions and Github Pages. Observable Framework builds the static files to serve the site with useful features such as responsivity, interaction, light/dark modes, and mobile layouts. Repos are available for the site and the scraper, contributions are welcomed.

Limitations

All data we get is from Faust's fork of Lost Ark Logs. This means that the data we have is a subset of a subset of a subset. This may not be a perfectly representative sample of the population.

Build classification (and weird logs classification) may be incorrect as we are only inferring what we can using simple rules given the available data. Future improvements to the data source (Faust's logs) may improve accuracy of the classifications.

Especially for the latest content early in the balance patch cycles, we have relatively low sample sizes meaning the estimates of performance for uncommon builds may not be precise. We do not attempt to account for all possible causes of performance differences, such as tripods, gems, cards, engravings, quality, elixirs, transcendence, or player skill. Additionally, the support in the party of each player will largely affect performance. In all, with sufficient sample size, we should be able to estimate a reasonable range of performance (where the true median build performance is) given the variance from factors other than the build itself. However, it is difficult to determine the exact cause of performance differences without more detailed data. This means that the data should be used as a rough guide of relative balance and not an exact representation of the true ranks of each build.

Acknowledgements

Thanks to Faust for providing the logs and the API. Thanks to Snow for creating Lost Ark Logs. Thanks to PoisonSandwich and Kennethnyu for inspiration and early feedback.

Contact evilandrex on Discord for any questions or feedback for this site specifically.

Daily Log Scrape Deploy Observable Framework to Pages