# How the Elemental Chance Value is Measured

As the title says, I’d like to know how those values get measured.

I have a few fire weapons and tested them out on the target practice at Marcus’s Shop. I was under the impression the chance was regarding the shots fired (i.e. out of 100 shots, youll have about 16 cause the elemental effect if the chance is 15.6…at least that’s the way I understood it from the wiki). I tested this out with a Fervid Provacateur (lvl 9 with a 16.5%) with all modifiers off (BARs and no skill points), but the values I got (after 10 trials with 50 bullets each) do not agree with that percent. Could anyone explain the actual system used?

It’s pure RNG. You may see less, you may see more, but with huge enough sample size, you should see a close percentage.

Then there’s elemental resistances and weaknesses which IIRC also affects the chance of elemental effects.

Every single shot has X % chance to cause elemental effect. Every shot is independent of the other shots. It’s the same thing as when you roll a dice. It doesn’t matter what the last roll was, you will still have a 1/6 chance for each roll.

2 Likes

Right, but because each is an independent event, and (the specific event happening)/(total events) is the RATIO which also happens to be the probability (without multiplying by 100 to make a percent), the probability value will be constant. I could have initially chosen (and actually did) just one mag size (of 25) and the approximate occurrence would have been (and was because i did two trials) ~25% which were around 6 (plus or minus 1) shots per magazine. Putting them together with the other trials for a bigger sample size gave a total of 250 with ~ 60 occurrences which is still ~25%.

What I’m saying is that BECAUSE they are independent events from each other, the probability (chance frequency or whatever you want to call it) will be inherent from the ratio (like the 1/6). They wont exactly be that value but they will be very close to it (since it depends on sample size as well).

Ah , the age old problem… chance v probability!

1 Like

Wow ok so I guess it did come to this. Forgive me, I am not intending to be rude, but you’re the 5th person to tell me this about sample size: it is NOT the sample size. I actually created a Matlab script that simulates the basic principles of an RNG where the system will choose a discrete random value and verify whether its below the threshold value. If it is, then the action is true, if it isnt, then the action is false. I simulated the sample size and trials to what I mentioned, and it was very and i mean VERY rare to get above 20% (whereas the values I got from the ‘physical’ test were on average 25%). The average for this test run was 15.8 (plus or minus ~ 0.5)%. I created the script in such a way that it includes the number of trials so 1 run of the script simulates an independent run simulating what I did in the game (of 10 trials with sample size 50). I ran the simulation about 20 times and only once did I get 20% but never above. Running the script multiple times reflects a total sample size and trial number that is WAY over what i would ever do myself.

What I am trying to say is that the sample size and trial number IS big enough to zero in enough into the percent displayed on the info card. It wont be the exact percent value, but definitely close enough to conclude the value concurs.

BTW I had to use a trial size of 2 to get over 20% (and that was only about 2 out of another 20 runs).

hah, I won’t argue since it’s not my field. If I got to tell your finds to someone who told me about my post, I will.

Anyway I’m missing the point. I was trying to say what @Ronnie_Rayburn says. Ignore my reply.

Haha it’s all good. I have this same question on Steam but not much help there either. I will say this about the script. It does NOT simulate the exact code used in the game, but I AM sure the basic RNG principal is within it. If my code/testing technique is lacking a factor, I would like to know what that is to better understand the system I enjoy very much lol.

Don’t things like your level and the targets level affect even elemental chances , as in higher level enemies will be less likely to pick up dot?
I dont know if that can maybe skew some results?

yea IIRC higher level enemies also has less chances to catch elemental effects.

Yea they do but I was a level 9 with no used skill points (i manually reset them) and I turned off my badass ranking points. I was also shooting at the marcus shop target practice.

I think the system as described is accurate, but there may be modifiers on your target? The other thing (to address your later post describing your Matlab simulations) is that you should calculate your confidence intervals on those outcomes. You would need to use the correct function for a binomial distribution but, if you know how to use Matlab, you likely know how to do that too! Anyway, inferring rates from experimental observations is tricky stuff, and it’s surprising how many more observations you actually need to get narrow confidence intervals.

1 Like

What type(s) of stuff do I need to narrow onto? I agree, I did a VERY crude simulation but was hoping to encompass the basic principles.

I also want to point out that I am not trying to show that the system or information is wrong. On the contrary, I believe its accurate and showing what is supposed to, but I am just trying to figure out the mechanism/factors (in a more mathematical way i suppose?) of how it works at its most elemental form (no modifiers). Not so elemental that I have to recreate the code behind this (god no, my skills wouldnt even reach that level), but just enough that I know what is being modified at different instances. I hope that makes sense?

Id imagine you would get pretty accurate results from that then.
Unless there’s other sneaky factors we’re not accounting for

1 Like

Precisely those sneaky factors are what Im looking for, if there are any.

Actually, if I understand correctly, the elemental resistances would not affect proc chance but damage done? Not sure exactly how that works. The damage and splash damage formulas should include all relevant modifiers, but I think that’s applied after the proc chance is calculated? I actually suspect that the reason you’ve been seeing lower numbers is simply a sampling issue; it takes a surprising number of runs to get the confidence interval on observed occurrence rates down to the 1-2% level.