On the test server for the Winter Update 2025, the Tamed Frazzlemaw creature was bugged for a few days, and was unkillable - your attacks produced damage numbers, but the creture remained on full hp. Additionally, it had neutral resistances, 0 armor and 0 mitigation. It was effectively a target dummy, with the one restriction that damage was capped at 1000.
I realised this was the perfect opportunity to do some tests. My setup was:
- Level 876 with 12 bonus damage from the wheel for a total of 174 flat damage
- Sword fighting 132
- Bright Sword (atk 36) with a T3 crit imbuement
- I had an attack value of 382
Over a couple of days I managed to record 10496 auto-attacks on the tamed frazzlemaw: 9609 normal hits and 1086 crits. Here is a summary.
| non crit | crit |
| count | 9609 | 1086 |
| unique | 147 | 51 |
| min | 278 | 731 |
| max | 585 | 887 |
| avg | 381.3439484 | 808.5911602 |
I also produced two histograms as shown below.
Some points of note about these two plots:
- The game seems to be using multiple random number rolls to choose the damage, which results in a Binomial distribution curve.
- We also see that for non-crit hits, high damage rolls are rare. This is only true for single target auto-attacks. Spells have a completely flat distribution.
- Another noteworthy point is that damage falls into distinct buckets. Judging by my maximum observed crit damage of 887, we can infer that it would have come from a non-crit of 591 damage, but because the highest damage rolls are so rare, we never actually saw a 591 hit happen, even across 10,000+ tries. Extrapolating from all of my critical hits, we can infer that I am missing 572, 587, 589, 591 in the non-crit dataset, so we would have 151 non-crit buckets in total.
You might think I've made a mistake by finding 151 instead of a nice round number like 150, but from my previous research, I have found that all spells have an odd number of buckets too. This is because there is always an average value, and then an even number of buckets either side of it, resulting in an odd number in total.
Now we can answer your question. We have 51 buckets of crit damage and 151 buckets of non-crit damage, which suggests that a critical hit takes from the top 1/3 of the non-crit damage distribution. From another perspective, the minimum damage crit is 731, which would have come from a 487 non-crit hit. This is 2/3 (or 66.666%) into my non-crit damage range.
To summarise: all single target critical hits are high rolls that take from the top 1/3 of the non-crit damage distribution. Additionally, non-crit hits in this top 1/3 are quite rare, which results in single target auto-attack critical hits doing considerably more damage than you would expect.
