Method and validation

Last updated: 12 November 2026

The simulator doesn't copy a guide: it plays thousands of fights with your character and compares options by the damage they do. Here is how, what it has been checked against, and where its limits are.

How it simulates

Every fight is played event by event: white hits from both hands at their speed, the Classic attack table (miss, dodge, parry, glancing, crit; +19% white miss when dual wielding), energy regenerating continuously at 10 per second, combo points, poisons, bleeds and the target's armor (arm / (arm + 400 + 85·level)). Abilities, talents, items and racials come from the game client (build 1.60.1).

The DPS you see is the average of 3000 fights. Two options are compared with common random numbers: both see exactly the same rolls, so the difference between them is the decision, not noise. Everything runs in your browser.

Validation against real logs

Using the public report gxrgvz4yzsi2 on Forever Sixty (8 November 2026, 104 pulls, about 10,300 events from a level 25–27 Rogue), which records every hit before and after armor:

• It matches: armor reduction on a normal mob (25–28%), crits ×1.97, +19% dual-wield miss, glancing blows (×0.89; 15% against the boss), energy at 10/s, Shank's off-hand penalty, Garrote and Rupture ignoring armor, and weapon attack power (1/14 DPS per point).
• A whole fight (boss Saltspine, 45.9 s): 70.6 real DPS against 67.7 simulated with the same gear and rotation, 4% apart on a single fight.

What it corrected

• Expose Armor does not change damage today. Sunder Armor (the same mechanism) lowers the armor shown, but across 10 mob types and over 1,500 hits the damage reduction is the same with and without it. The simulator treats it as having no effect; if the game fixes it, there is a checkbox to turn it on.
• Instant attacks are not normalized: Sinister Strike uses the real weapon speed, so slow weapons are worth more.
• Instant Poison hits harder than its tooltip (17.8 on average against 13–17) and scales with attack power; some mobs are immune to poisons, and some have half the armor or much more.
• Backstab fits 150% × (weapon + bonus) on the six hits available; still marked provisional.

Not confirmed yet: Expose Armor measured directly, Serrated Blades and the armor penetration of maces, Ghostly Strike, Hemorrhage, and parry and block when attacking from the front.

How it optimizes

• Rotation. A priority list with parameters (opener, builder, combo points for each finisher, duration thresholds). They are searched with restarts from several styles, each candidate over 1,000–1,500 fights with the same rolls, and the winner is measured again over 3,000–6,000 fresh fights so a lucky candidate doesn't win.
• Talents. Local search from several seeds for your level and a target level, with the points ordered so each one is the best choice at the level it is spent (no respec needed).
• Gear. Weapons are tried one by one, because they change the rotation; everything else is ranked by its effect measured with the same simulation as your DPS.
• How much the list leaves behind. A rollout bound: at every decision in the fight, each possible action is tried and the rest is simulated. The list stays within ~1% of the best sequence found that way, which is the measurement noise.
• A generic learner. It starts from an empty list and learns rules from rollouts and search on the simulated damage. In 12 measured cases it ties the list (±0.6%) in groups and beats it by 2.4–3.5% when leveling. Its precomputed lists are used to check yours in the Rotation column.
• In progress: dynamic programming over levels to tell when a talent respec pays off, instead of a single continuous plan.

Limits

One target at a time, no group buffs or procs from other players (in the log, Windfury Totem added about 400 white damage in the boss fight), and whatever the game doesn't let us measure yet. “Best found” is the best this search finds with this model; a hand edit can beat it, and the page says so.

Sources: the game client, Forever Sixty, MetaBot.GG and ForeverDB. Have a log with Expose Armor, Backstab or Hemorrhage? Send it with “Feedback” in the simulator.