All Jobs Damage Simulator And Gear Sets

Weather bonuses heavily outweigh acuity belt even at highest content level, eg Sortie basement.

Exception is when spells are not landing at all without the extra macc+15, which is unlikely unless at ML0 or something.

It's not even necessarily SC bonus(that is a factor if intentionally closing skillchains sure, but sc damage also simply increases from the actual wsd caused).

It's more about how Ankusa helm and Nukumi body work together and will always win likely provided you have the correct killer trait. The combination at a minimum is always 7.5%(rounded down likely but still) increase to wsd but this damage increase is also applied to all hits(as best I can tell anyways) So even in multi hit ws's its likely to be best. Add Killer Instinct to this or a circle effect and its damage increase is even further from Nukumi Body. I have been hard pressed to find any situation that even Nyame body r30 outweighs Empy body.

This is very easy to do. I tried including TP sets when I first posted this thread, but it turns out that players are very particular about their TP sets, especially since the automatic set finder does not consider Magic Evasion. I'm now on holiday for ~2 weeks, but I'll add 50% DT and glass cannon sets for all jobs in January.


You may have misunderstood that part of my post. The discussion was comparing Skrymir, Acuity, and Sacro for situations when you have no weather or day bonus. I imagine that most end-game players use gearswap (or similar) and include automated checks for weather, day, distance from target, and even time of day for deciding special swaps for their sets. In the case of no weather/day, I think Acuity Belt +1 would be the best general nuking waist option in most situations.

It's a terrible suggestion, but you might want to include a set that is 50DT, doesn't have a subjob, and includes warder's charm and shadow ring for odyssey.

maybe put in all caps "ODY BIS TP SET" too, might have less DDs dying to v20/25s

but bro.... that's just a bad browhm who obviously isn't bro enough to keep up with the brodeeps.

There is an "error" with code for Zanhasso not adding the Ikishoten bonus. Function is invoked without True for zanshin parameter, which is False by default.

actions.py

lines 620 and 631:


requires change to:

Thanks for catching this. It looks like this only affected the DPS simulation part of the code such that each successful Zanshin hit was returning 150 less TP than it should.

This has been fixed, but the fix is to use player.main_job=="sam" instead of "True", since only SAM main job can benefit from Ikishoten merits.


You may have an outdated or modified version of the code. The affected lines were 868 and 887. Lines 620 and 631 are both inside of a main-hand OA3 check in the most up to date version of the actions.py file.

Horos Toe Shoes may need to be updated, I don't think their STP value reflects 5/5 Closed Position merits.

The prerequisite is situational though, not sure how that would be coded in because it's not simply +15 STP. But regardless, +15STP would likely win if you can 1. Safely face the mob, 2. Afford the defensive loss.

Well I assume you would just uncheck them when simming DNC attacks from the rear or flank.

I simply added a new checkbox for "Closed Position", similar to Overwhelm. This new checkbox adds Accuracy+15 and Evasion+15 by default. Selecting the "Closed Position" checkbox with "Horos Toe Shoes +3" equipped additionally provides "Store TP"+15.

This was not a change to the gear.py file, so you'll need to update the main code (version 2025 January 14b) from the GitHub page to see the changes.

Hi man. I was wondering if you can add the new Null equipment to this, especially the Null Shawl? I am struggling to add it manually, given that the cape section of the gear file is made entirely for Ambuscade. Also, I noticed an issue with Sanctity Necklace not having any accuracy. Thank you.

It is added in the latest version.

Thanks for this! Updating now.

anyway you can add some of the more niche ws to it? like Retribution for brd ect?

It would be helpful if there was an option to run multiple trials, and output the mean and standard deviation.
Would also like the Turms +1 set to be added, since it's used in some RUN hybrid for parry.

I think BRD Retribution and similar are too niche for me to add to the BG Wiki sets. I encourage users to play with the code themselves to determine such sets.

This is one of the original features of the code. You'll want to click the "Create weapon skill distribution plot" button on the "Simulations" tab. This will simulate 10,000 of your selected WS using the equipped WS set (on the Simulations tab) against your input Enemy Stats. Be sure to disable "Verbose" and "Very Verbose" DPS in the Settings menu before running the simulations or the code will spam the log with the damage dealt.

However, the mean and standard deviation are only statistically useful if you have close to a Gaussian distribution. This is not the case for many weapon skills, especially FTP transferring weapon skills.

I've added Turms+1 to the gear.py file.

That's mostly what I'm looking for, but it would be helpful if I could do the whole DPS simulation, and not just weaponskills. It's hard to say anything meaningful about TP sets from just one trial, and putting the results of multiple trials into excel is pretty tedious.

Have you checked that application? You can do full DPS sim, you can do multiple of them too. You can even have all of them on the same graph. If you are willing to change the code, then you can even manipulate duration and other things (depends on your Python skills).

I've been doing the full dps simulation, and I see how to graph weaponskill damage, but what I want is to run 1000 trials of the full dps simulation and get the mean and standard deviation of dps.

What exactly is your goal in doing so? Each trial already simulates a 2 hour long period. If you want the mean of 1000 trials then you could change the code to simulate a 2000 hour period instead, but I very much doubt that's going to be instructive. Even upping it to a 10 hour period would have little impact. Your standard deviation is also likely somewhere around a quarter of a percent or less of your total DPS and isn't going to vary between different sets in a way that's informative.

How else am I supposed to compare sets that are close, or understand the variance of the dps? You need multiple trials to draw conclusions.

Why?

If you run a simulation for 24 hours, what makes you think that running 5, 10, or more 24 hours simulations will produce a different result? There's variance in a 15 minute fight sure, but in a 24 hour sample?

Or, as was suggested above, just increase the sample size of a single test to whatever range you think is appropriate.

In my experience, running the same test a few times in a row will result in a TINY difference in the actual results, completely negligible.

Increasing the sample size is not the same as running multiple trials. I'm seeing variation of around 2%, and would like a better way to compare sets that fall within that range.

How does that work, mathematically?


You're running the exact same set of gear twice and seeing a 2% increase in DPS one time over the other? I've run some sets ~5 times in a row and never seen more than a ~.2% difference.

Longer trial duration is a perfectly valid method to reduce variance. Run longer tests, vary your parameters to evaluate your sets in multiple environments, and consider whether there are any additional factors of merit to differentiate the two sets. If you still can't differentiate them by that point, then you can conclude there's no meaningful difference.

Now if you wanted additional tools to dig into that, sure. Distribution of TP gain per round and rounds/time to target TP, etc. That's potentially valuable information the sim doesn't readily show right now (afaik). But you're fixating on the mean while rejecting the value of a longer trial as if the latter doesn't naturally converge towards the true value in this case, which is untrue.

If you make 2h sim (I think that's the default duration) using multi hit WS with FTP transfer with moderate amount of multi hit in WS and TP set and uncapped attack with cratio around 2.0 (pdif range is much wider then), then it's very possible to have results even further apart each other than 2%.

Still if someone wants to narrow the results. Just change the code to simulate much longer "fight".

I am all for figuring out optimal dps, but ultimately in my opinion you have to work within the parameters of the game to be realistic. Running multiple samples is fine, but duration doesn't mean much when your actual encounters are not even remotely as long.

Doesn't matter if a particular set proves to be better over an hour or 24 hours of fighting vs another set that is better if your encounter is actually 1-10 minutes.