After spending quite some time investigating Silkroad's damage calculations, I wanted to share my findings and the source code with other developers.
I started with the damage formulas that have been circulating in the community for years. They helped us understand the basics, but several parts were still unclear, especially damage randomization, mastery scaling, level differences, and how Imbue damage is actually added.
Instead of continuing to guess where the missing constants 1.27 1.28 come from
Thanks to @ we went directly into the original ''32-bit SR_GameServer.exe'' and analyzed its assembly code with ghidra.
What we found
I. Damage randomization
The GameServer doesn't use a simple uniform random damage roll. We identified a system involving three random samples, their minimum, and an additional random branch that affects damage distribution.
II. Level damage multiplier
The level difference bonus isn't a universal fixed multiplier like 1.28 or 1.30.
The recovered calculation increases damage by approximately 3% per level difference, capped at 30% for a higher-level attacker.
III. Mastery and damage calculations
We traced how mastery affects attack power and identified the separate physical and magical calculation paths, including balance, defense, absorption, and modifier order.
IV. Skill damage and Imbue
One of the most interesting findings was the secondary damage calculation.
The GameServer stores separate percentages for a skill's main damage and its secondary contribution.
For example, Flame Wave can use 250% for its main magical damage while applying 83% to the secondary Imbue contribution.
This explains why simply adding Physical and Magical damage with the same skill multiplier isn't always correct.
V. Critical damage
Critical hits double the physical damage component, not the magical component.
We also identified where integer truncation and floating-point calculations affect the final result.
In-game validation
I compared our calculations against actual in-game damage using Normal Attacks, Physical Skills, Fire Nukes, Cold Nukes, Lightning Skills, and Imbues.
The results were extremely close, with many comparisons differing by only 1–4 damage points. The secondary Imbue contributions matched within 0–1 damage in our three elemental tests.
Source Code & Technical Documentation
I'm making the research and implementation available so other developers can inspect, reproduce, and improve it.
Recovered Damage Engine (JavaScript):**
Original GameServer Analysis (assembly references, offsets, formulas and constants):
Hybrid Damage Implementation:
Full GitHub Repository:
Hopefully, this research will save others some time and help us get closer to a fully understood Silkroad systems to improve.
I tried long ago to make the maths on how damage is calculated to know what's the best output, the break point on damage between bow hybrid builds, but I wasn't able to do it.
Here is my maths, I was able to calculate exactly hp/mp/phy and mag balance. But the most important, the damage didn't matched at all in game, I give up.
Edit: Oh, and to answer your question, yeah, skill should be flat added to the damage of the weapon based on the percentage it says at the end between parenthesis for each hit.
1 hit (+%100) = x1
3 hit (+%80) = x0.8 + x0.8 + x0.8 so each hit is calculated separately with different dice/roll.
I tried long ago to make the maths on how damage is calculated to know what's the best output, the break point on damage between bow hybrid builds, but I wasn't able to do it.
Here is my maths, I was able to calculate exactly hp/mp/phy and mag balance. But the most important, the damage didn't matched at all in game, I give up.
Edit: Oh, and to answer your question, yeah, skill should be flat added to the damage of the weapon based on the percentage it says at the end between parenthesis for each hit.
1 hit (+%100) = x1
3 hit (+%80) = x0.8 + x0.8 + x0.8 so each hit is calculated separately with different dice/roll.
thanks bro you save me a lot of time on hp mp phy mag balance,ı cooked some damage formulas but didnt tested yet when its finished i share the code here
I did a very deep research and I've got this result
At the moment I don't have the access to my computer, but as soon I will, I will upload the dmg formula, it's reversed from the original GameServer binary... I will update this post, asap
I did a very deep research and I've got this result
At the moment I don't have the access to my computer, but as soon I will, I will upload the dmg formula, it's reversed from the original GameServer binary... I will update this post, asap
Interesting.Since you said it was reversed from the original GameServer, could you also share the
GameServer version,
function address/RVA,
small Ghidra/IDA snippet
constants used by the calculation
when you post the formula? That would make it much easier to verify.
Interesting.Since you said it was reversed from the original GameServer, could you also share the
GameServer version,
function address/RVA,
small Ghidra/IDA snippet
constants used by the calculation
when you post the formula? That would make it much easier to verify.
I wrote it in a .md file , try to open it via a program that supports the .md, looks nicer, easier to understand, I also uploaded the gameserver version that I used to extract it.
This is the base calculation.
This document covers damage amounts and the immediately relevant caller modifiers. It does not claim a complete probability model for accuracy, avoidance, critical chance, blocking, abnormal-status application, aggro, or skill-target selection.
--The .exe I used to extract the formula is ISRO-R GameServer, RIGID files, I don't know if whoever had it before reaching me, edited something related to the dmg formula, which I doubt... so whatever .md file details contains, it's from that compiled version of GameServer, also mind one thing, VSRO version might be different, I wouldn't say very different, from my understanding. SROR introduced dmg 1, when the calculation reaches a certain coefficient.
I wrote it in a .md file , try to open it via a program that supports the .md, looks nicer, easier to understand, I also uploaded the gameserver version that I used to extract it.
This is the base calculation.
This document covers damage amounts and the immediately relevant caller modifiers. It does not claim a complete probability model for accuracy, avoidance, critical chance, blocking, abnormal-status application, aggro, or skill-target selection.
--The .exe I used to extract the formula is ISRO-R GameServer, RIGID files, I don't know if whoever had it before reaching me, edited something related to the dmg formula, which I doubt... so whatever .md file details contains, it's from that compiled version of GameServer, also mind one thing, VSRO version might be different, I wouldn't say very different, from my understanding. SROR introduced dmg 1, when the calculation reaches a certain coefficient.
Thanks a lot, We had already worked out most of the damage formula, but your files helped us finally understand the 1.30 / 1.28 constants and the Attack Rate part. Those were two of the missing pieces in our research.
I’ll post our cleaned-up formula and findings once I tested new formula in game.
Thanks a lot, We had already worked out most of the damage formula, but your files helped us finally understand the 1.30 / 1.28 constants and the Attack Rate part. Those were two of the missing pieces in our research.
I’ll post our cleaned-up formula and findings once I tested new formula in game.
Lovely. I will update my post once I have extra information!
Do you also have the server data/config/database files used with this SR_GameServer.exe? I’m specifically trying to recover the runtime values for character parameters 174, 175, 178, 180 and 259, plus the basic attack descriptor value.
Yea i made big progress.But i must confess there is 3687 skill data and each one have different parameters to effect end damage for example.Fire Tree Nukes
Element Skill Count MainDamage(Param4) Imbue Rate(Param7)
Fire A 18 %250 %83
Fire B 18 %300 %100
Fire C 18 %300 / 315 / 330 %100
Fire D 22 %263 %88
Fire E 19 %315 %105
Fire F 13 %330 %110
Fire G 7 %263 %88
The damage engine and formula works %100 but if you try to add imbue and skill together you will need that skills imbue rate at database.So my plan giving ai this dataset and make it work with damage engine.
[Request] Attack Player Damage Formula 04/03/2021 - CO2 Private Server - 3 Replies Hello Guys,
If anyone have some good formula on how to get the perfect / balance damage without calculating the battle potency.
I'm thinking of plain damage no Battle Power add on. I'm thinking of the calculation of late conquer something like a version of (4427).
What could be a nice formula for this action:
1. Melee / Physical Damage.
2. Ranged Damage.
3. Magic Damage.
The Silkroad Damage Formula 02/16/2020 - Silkroad Online - 18 Replies Copied from: http://sro.mmosite.com/content/2009-09-11/20090911 002742992.shtml
Posted on 11 September, 2009. Looks like a very accurate formula that seems to work every single time. Whoever figured this out really is going to help out emulator developers.
Check the link to see the screenshots!
Plus / Damage Formula 04/19/2019 - Silkroad Online - 9 Replies hi all. i'm working on an alchemy simulator and want to be precise about these values without typing all of them separately. does anybody know the alchemy formula that determines the min/max phy. and mag. damage of a weapon one plus above?
lv.90 bow for example:
+0 phy. atk. 1047~1283
+1 phy. atk. 1083~1319
+2 phy. atk. 1120~1356
+0 mag. atk. 1674~2052
+1 mag. atk. 1732~2110
Plus / Damage Formula 04/15/2019 - SRO Private Server - 0 Replies wasn't meant to open here, please delete
Damage Damage Damage 02/28/2012 - Mabinogi Hacks, Bots, Cheats & Exploits - 3 Replies Because of poor English grammar may be a bit strange
May I ask what damage accumulated or MOD