Training Ground guide

How to test whether a build change actually helps

The Training Ground is a controlled comparison room. It does not prove the strongest universal build; it shows whether one change helps your build under the same conditions.

Same targetOne changeThree repeat runs

Inscription comparison · 2 October 2026

Constant won the six-minute comparison

+1.79B · +43% total damage with the constant lineup in this pair of runs.

Both results show 00:06:00, 4.13M power and Auto combat against the same Deathshade II training dummy.

Constant

14:36:10 UTC+2 · 6:00

Total damage5.95B

Displayed DPS16.5M

Arcane DomainLv 61 · Gold 2
Pause FieldLv 43 · Gold 3
Stasis FieldLv 42 · Gold 2
Slow FieldLv 43 · Gold 4
Winterless ArtLv 42 · Gold 4

Result screenshot ↗Lineup screenshot ↗

Boost

14:28:44 UTC+2 · 6:00

Total damage4.16B

Displayed DPS11.5M

Vengeance DomainLv 61 · Gold 4
Dreamforge ArtLv 43 · Gold 4
Quick FieldLv 42 · Gold 4
Slow FieldLv 43 · Gold 4
Winterless ArtLv 42 · Gold 4

Result screenshot ↗Lineup screenshot ↗

Damage entries visible in both captures
Skill	Constant	Boost
Avatar of Doom	2.48B	1.45B
Twin Chaos	1.53B	1.19B
Reality Collapse	680M	588M
Endrain	361M	297M

Avatar of Doom accounts for 1.03B of the 1.79B total gap. Other fully visible entries are Shadow Trail: 255M (4%) on constant, and Fester: 194M (5%) on boost. Lower rows are clipped.

Use constant for six-minute fights based on this pair. The +24.60% permanent bonus sum is not a prediction of the difference between complete builds. Several inscriptions changed together; these captures do not isolate Arcane’s contribution or explain the activation timing behind the gap.

There is one recorded run per build. The screenshots confirm the target, duration, power and Auto mode; the remaining loadout was not re-captured. They do not measure the highest short burst or establish a universal optimum. To confirm repeatability, alternate both lineups with the rest of the setup fixed and compare repeated six-minute totals.

B = billion; M = million. Totals and DPS are copied as displayed; rounding explains small differences when dividing by 360 seconds.

Player-tested · 29 September 2026

29 September stone tests: 10.8M DPS

Total damage1.31B

Duration120s

TargetDeathshade II

Avatar of Doom641M · 49%

The timer reads 00:02:00: two minutes. Values below are the game’s displayed totals and DPS; rounded totals may not divide exactly into displayed DPS.

Higher observed damage, with variable activation. Rodrigo confirmed that the stronger comparison used Abyssal Call — Enhanced. Longer transformation alone did not establish higher damage. The two new-Moon captures show 9.95M and 10.8M DPS: neither is a guaranteed average, and the captures do not isolate the Moon’s contribution or measure Abyssal Call’s activation probability.

Capture 1 · 12:15 UTC+2

9.60M DPS

1.15B · 120s

Avatar of Doom: 475M · 41%

Abyssal Call recipe shown in the paired tree

View result

Capture 2 · 12:24 UTC+2

7.84M DPS

944M · 120s

Avatar of Doom: 266M · 28%

Avatar of Doom recipe shown in the paired tree

View result

Capture 3 · 12:35 UTC+2

10.5M DPS

1.32B · 125s

Avatar of Doom: 570M · 43%

Same-Origin Moon comparison; longer run

View result

Capture 4 · 12:52 UTC+2

9.13M DPS

1.10B · 120s

Avatar of Doom: 483M · 44%

Further stone comparison; full package not isolated

View result

Capture 5 · 13:29 UTC+2

9.95M DPS

1.19B · 120s

Avatar of Doom: 525M · 44%

First supplied result after the new Moon

View result

Capture 6 · 13:40 UTC+2

10.8M DPS

1.31B · 120s

Avatar of Doom: 641M · 49%

Latest result with the current stone package

View result

Latest damage breakdown

Avatar of Doom: 641M · 49%
Twin Chaos: 328M · 25%
Endrain: 108M · 8%
Soulbolt: 69.5M · 5%
Shadow Trail: 65.8M · 5%

Five visible leading entries; remaining damage is not omitted from the total.

What the comparison can support

10.8M is about 16.0% above the 21 September 9.31M record, using displayed DPS. That is a whole-build observation across dates, not a 16% Moon-stone gain. The 1.32B run lasted 125 seconds, so its larger total does not beat the latest 10.8M DPS result.

To measure reliability: keep target, 120-second duration, skills, pets, inscriptions and inputs fixed; alternate the two recipes and record every run, plus Abyssal Call activations. Compare medians and ranges. These captures alone do not establish the gold cost per sustained DPS gained.

Attack, Crit or Penetration?

For Rishi’s recorded boss setup, the higher-Attack package is the useful comparison point. The tests do not establish one stat order for every Warlock or activity.

Read the stat labels correctly

Attack · Crit · Hit · Penetration: Attack, Crit, Hit and Penetration are separate point values on the stat sheet. Compare each before and after a change.
Précision / Hit Rating: On this site, French “Précision” translates the English Hit / Hit Rating label. It is not Penetration.
Attack Bonus · Crit Bonus · Penetration Bonus: Compare these percentage entries separately from flat points. A displayed Crit Bonus percentage is not a displayed critical-hit probability.
Damage Bonus / DPS: Damage Bonus is a percentage stat. Total damage and DPS are measured combat results. Use the results to decide whether a whole equipment change helped.

Choosing weapon, armour and talent passives

Use the exact tooltip and roll size, rather than the stat name alone. Save your current preset, change one passive or one item, and compare three equal-duration runs on the same target with the same skills, pets, inscriptions and buffs. Compare median damage and survival; repeat in the activity you actually play.

Compare stat changes

Player-tested · 21 September 2026

High Attack won the four-build boss comparison.

Every recorded package ran for two minutes against Deathshade II. The final package reached 9.31M DPS and 1.12B total damage, establishing the previous measured baseline. The current account presets are recorded separately on My Build.

21 SEP BASELINE9.31MDPS

RUN 1

High Penetration

7.25M DPS

Attack: 23,622
Crit: 12,876.2
Penetration: 12,874.5
Total: 871M

Highest displayed Penetration, but the lowest measured damage in this comparison.

RUN 2

Balanced Attack

7.68M DPS

Attack: 25,535
Crit: 13,464.8
Penetration: 10,329.4
Total: 921M

More Attack and Crit beat the higher-Penetration package despite a slightly lower Attack Rating.

RUN 3

High Attack

9.02M DPS

Attack: 26,781
Crit: 13,464.8
Penetration: 9,688
Total: 1.08B

The first high-Attack pass established the new performance band.

RUN 4WINNER

Final High Attack

9.31M DPS

Attack: 27,450
Crit: 13,464.8
Penetration: 9,688
Total: 1.12B

Best of the four 21 September packages. Preserved as a historical whole-build baseline.

FINAL RECORDED PANEL

27,450 Attack · 13,464.8 Crit · 9,688 Penetration

Attack Rating100,734.3Attack Bonus218.3%Crit Bonus210.2%Damage Bonus84.8%

Keep this complete package as the historical comparison baseline. It beat the high-Penetration run by 28.4%.

Final 1.12B breakdown

Twin Chaos358M32%Avatar of Doom358M32%Shadow Trail149M13%Hellrain93.1M8%Soulbolt89.5M8%

Final Warlock stat panel showing 27,450 Attack, 13,464.8 Crit and 9,688 Penetration — Final recorded stat panel

Two-minute Deathshade II boss result showing 1.12B total damage and 9.31M damage per second — Two-minute Boss result · Deathshade II

Evidence boundary: the four runs prove which complete recorded package won. The last step also changed Attack while the Moon of Tenacity potential changed, so the 3.2% increase from 9.02M to 9.31M cannot be assigned to Doomsday Embers alone.

THE RULE

Change one thing. Keep everything else the same.

That is the whole message. If the target, position, timer, Auto setting or another build part changes, the result cannot tell you what caused the difference.

01
Lock the conditionsSame dummy, position, timer and Auto setting.
02
Change one partOne item, Talent, Pet, Emblem board or other system.
03
Repeat both packagesRun A three times, run B three times, then compare the pattern.

Full method + worksheetOpen the complete controlled test

Pumpkin-headed Training Ground practice dummy with a target on its chest — The dummy is the fixed reference point. Use the same target for every comparison.

CONTROL THE ROOM

Keep everything fixed except the change being tested.

The Training Ground removes much of the noise from a real fight. The comparison becomes useful only when these conditions stay the same.

TargetSame dummy and difficulty
PositionSame range and facing
TimerReset meter · continuous 61s
InputAuto enabled for both packages

A FAIR A/B TEST

Compare the build you have with one controlled change.

Package A is your current setup. Package B is the same setup with one slot or system changed. Do not rebuild several parts at once.

PACKAGE A

Current setup

Same equipment slot
Current stats and passive effects
Three repeated runs

PACKAGE B

One controlled swap

One item, Pet, Emblem board or talent change
Every changed stat and passive effect
The same three repeated runs

Repeat the recorded routine for both packages.

Reset before every attempt. Save total damage, DPS and the complete source breakdown after each run.

1
Record the starting stats, equipped systems, temporary buffs and anything that cannot be switched off.
2
Use the Training Ground damage dummy and stand in the same fixed position beside it. Keep other players off the target.
3
Reset the damage meter, cooldowns and combat resource, then switch Auto on.
4
Run Package A for 61 uninterrupted seconds. Save total damage, DPS and the complete source breakdown; repeat three times.
5
Change one slot or system for Package B, reset everything and run the same 61-second Auto test three times.
6
Compare the middle DPS result, the spread between runs and the damage-source breakdown. Rerun interruptions.

Important: if one gear swap changes Attack, Penetration and a conditional damage effect together, the result belongs to the whole item package. It does not prove which single stat caused the gain.

Historical Mage example · preserved baseline · 18 September 2026

5.70M DPS

Training Ground · Damage dummy · 348M displayed total · 61 seconds

This older Mage run remains a worked example for the controlled method. It is not the current Warlock benchmark.

Open Mage archive

Complete recorded breakdown

Reality Collapse83.0M24%Frostspread82.2M24%Snowstorm66.2M19%Lightning Lash62.9M18%Concentrated Exploding Star49.4M14%Equipment Effect4.44M1%

READ THE RESULT

Decide from the pattern across the runs.

Clearer improvement

Package B is higher across repeated runs, and the intended damage source improves without a much wider spread.

No clear winner yet

The two result ranges overlap heavily. Extend the timer or add more matched runs before deciding.

Weak evidence

Only the best run improved, a rare effect triggered once, or another condition changed during the test.

THEN VALIDATE IN REAL CONTENT

The dummy isolates the change. A real fight tests whether it survives movement.

After a Training Ground win, use both packages against the same boss or mode. Keep those real-combat results separate: movement, mechanics, survival and target downtime answer a different question from the dummy.

YOUR RESULT SHEET

Save the same eight fields after every run.

1Target and difficulty
2Slot or system tested
3Package A: stats and effects
4Package B: stats and effects
5Position, range and active buffs
6Duration and input method
7Three DPS results per package
8Median, spread and damage share

If any condition changes, start a new comparison instead of mixing the numbers.