Larian Studios Forums Baldur's Gate III Baldur's Gate III - Suggestions & Feedback I rolled Intimidation/Persuasion 100 times to test the RNG

I wanted to test the RNG to see if it was as wonky as I thought and it turns out... not so much. At the link below, you can see the numbers I rolled in the order I rolled them.

https://docs.google.com/spreadsheets/d/1-iI05GHtKHhccnT8bQZPen-S-VoaaPABUrRXNq24K2U/edit?usp=sharing

As a test, I used an interaction where I had both the option to Intimidate and use Persuasion. I didn't care about target number, what my stat added or subtracted, etc. I just cared about the number it ended up on.

I also didn't want to do this in combat because a) there are modifiers such as high/low ground, lighting, etc. that I have no control over and b) I can't see the actual dice roll it's making. By using a skill check, I can see the die itself and did the following.

If I passed on my first Intimidate, I'd reload.
If I failed on my first Intimidate, I'd use the re-roll option.
If I passed on my second Intimidate, I'd reload.
If I failed on my second Intimidate, I'd try Persuade.
Whether or not I passed or failed Persuasion, I'd reload.
I did this until I had 100 rolls total.

I rolled the following totals:

20s - 5
19s - 7
18s - 7
17s - 3
16s - 4
15s - 4
14s - 4
13s - 5
12s - 7
11s - 4
10s - 7
9s - 2
8s - 4
7s - 1
6s - 7
5s - 6
4s - 4
3s - 6
2s - 6
1s - 7

That's a pretty even spread. Not perfect, especially for a computer program, but about what I'd expect to roll at the table.

Also, the average roll for a d20 is 10.5. For this spread, the average was 10.35. Again, not perfect but pretty darned close.

Overall, I've got to say I'm surprised by this. I was convinced the game was rolling low or high too often and wasn't truly random. And while it still may, for my taste, the fact that the overall average is pretty much spot-on, it's taught me to just take my lumps and move on.

Oh and feel free to pick apart this highly unscientific test.

I tracked it for 40 rolls and my current average is no joke, 5.

That’s typically why 100 rolls is what’s used to track randomness. If I went by 40, my numbers would be way off.

Yeah, its more of an "as I go" thing, wasnt doing it with just 1 thing. Are you you using the dice roll or your dice roll +mods as the "roll". Edit: oh, the final. Yeah so that means its lower than 10.

Mods are substracted from target, not added to the roll

Right, but when you hover it tells you the D20 roll, modifier, skill mod etc etc and then spits it out. So I just stealthed and Rolled a 15. But what that was was a D20(12 (the actual roll)) -3 +dex(3) + stealth mod (+3) = 15 total. So I ROLLED 12, I GOT a 15 total. My average DICE roll is a 5, my average total adjusted with the mods is 10/11 but it should still be a 15 if the dice roll was rolling 10 average.

This is a snippet of what Ive been doing

I don't really understand what you are doing. You have rolls that are less than your modifier there while having rolls above 20

I applaud the amount of effort you put into this. I don't want to knock your effort, but I want to point a few things out.

Statically, 100 is pretty meaningless with 20 options. It's been far to long since I took a stats class to tell you what would be meaningful.

I always roll my eyes when I see posts claiming "omg the RNG is so buggy wtf!" It almost always comes from a place of biased pattern recognition.

I would happy to bet money that Larian has some sort of unit test that uses their d20 system, rolls a million results, and compares the result to see if it deviates from an expected range. This is done so that if any change goes into the game that breaks this system by accident it is detected before getting submitted.

Ok, I edited it. I wasnt planning on sharing it so it was in Orbax-ese. Its raw rolls, until the greens which are the actual totals that beat the DC. Youll see two 25s and that was due to natural 20s that auto-succeeded regardless of beating DC. The green number is the total output of the last roll, not just the raw. To the right of each number is the Average Output, then the yellow is the adjusted output that drops the modifiers to give raw rolls. The column of adjusted average is showing just the raw dice roll. Its crap, but I understand it, hadnt really cleaned it up yet haha

Are you doing a chi-squared test or just reporting the average? The link you included assumes you're doing a chi-squared test. Otherwise, you'll need a lot more than 100 rolls

So you get the average of raw rolls then substract the modifiers? It still doesn't make any sense. If you want to do something like this, just list the raw d20 roll and calculate the average

I did, the adjusted average was me just differentiating from average output. 5 is the raw roll that was tracked. Like I said, this wasnt me setting this up to be a presentation haha, was just more of an aside and its now turning into a thing.

Yeah, it's been awhile. I think the "at least 100" is based on expected count (5%) times categories (20), but that's still the minimum. If you want some strong 99% statistical confidence I'd definitely run more.

I was mainly saying its statistically relevant and that "meaningless" might be a stretch for the analysis of the method.

All I know is that if the target is 5 or under I am guaranteed to fail lol.

+100,000,000,000