Testing Methodology
In an effort to promote transparency in the research process and to encourage others to perform their own testing, I’m sharing my testing methodology here:
For Golem 2*
- Golem 2* was selected for testing due to its high HP pool (190,000 HP), relatively low damage output, and attacks that do not have a debuff component to them.
- Each unit action was performed independently as to avoid chaining. No buffs such as Cheer or Focus were used.
- Max 4* lvl 60 Krile, total 110 MAG, used dual cast with a neutral elemental spell (Thunder in this case) every turn.
- Max 5* lvl 80 Lasswell, total 175 ATK, unarmed, used normal attack every turn. Lasswell’s normal attack only hits once.
- Healing was performed as necessary as well as use of Ethers to restore MP. All other units defended.
- 10 independent runs were recorded with the same setup resulting in over 400 samples for Lasswell and over 800 samples for Krile.
Observations
With both sets of samples from Krile and Lasswell, I noticed that the same 15 damage values (outside of critical hits) came up over and over again. Recall that Unarmed Damage and magic do not have any Weapon Type Damage Variance. The data collected debunks the whole Damage Variance theory of +-10%. Instead, these values support a variance of 86% to 100% of calculated damage per the damage formula.
| Damage Variance | Lasswell (175 ATK, Unarmed) | Krile (110 MAG, Thunder) |
|---|---|---|
| 86% | 1580 | 1331 |
| 87% | 1598 | 1347 |
| 88% | 1617 | 1362 |
| 89% | 1635 | 1378 |
| 90% | 1653 | 1393 |
| 91% | 1672 | 1409 |
| 92% | 1690 | 1424 |
| 93% | 1708 | 1440 |
| 94% | 1727 | 1455 |
| 95% | 1745 | 1471 |
| 96% | 1764 | 1486 |
| 97% | 1782 | 1502 |
| 98% | 1800 | 1517 |
| 99% | 1819 | 1533 |
| 100% | 1837 | 1548 |
Using the data collected during my runs, I also solved for the DEF and SPR values of Golem 2*. Golem 2*’s stats were calculated to be: DEF = 300, SPR = 150
Sample Video – Data Collection
For those interested, here’s a sample run. WARNING – this video is close to 14 minutes long and is super not exciting at all but is being made available for transparency reasons.
References and Credits:
While some of the information was obtained from various sources such as the FFBE Reddit, Exviuswiki, and other Japanese gaming sites such as Altema and Famitsu, the majority of the research, data collection, examples, and theorycrafting is completely my own work. If you feel that I may have missed crediting a specific author or individual, please drop a comment below. Thanks 🙂