Larian Banner
Previous Thread
Next Thread
Print Thread
Joined: Oct 2020
I
stranger
OP Offline
stranger
I
Joined: Oct 2020
I have done a lot of fighting especially the gnolls and the Tyr priests and even though I did a lot of reloads every time the enemies do have natural 20 very often while I struggle with 1s and 2s. I would like to show you some console copies but I can't copy them.

I don't know if it is related but same is for ability checks where I often get 9 or lower while 10 and and above seems to be rare.

I have done multiple analysis in (other) random number generators and it did show me often these number generators aren't very well in generate unified distributed numbers. Especially mean value and variance do often have errors and do shift from their theoretical values.

Please do some research and check if this is the case here too.

Thank you!

Last edited by Into Oblivion; 12/10/20 07:54 PM.
Joined: Oct 2020
Location: Italy
enthusiast
Offline
enthusiast
Joined: Oct 2020
Location: Italy
+1

I missed 6 attacks in a row with my pet at 81% hit chance. Not impossible, I know, but suspiciously unlikely.

Joined: Oct 2020
S
addict
Online Content
addict
S
Joined: Oct 2020
What concerns me even more is that sometimes the enemy crits, but when I look into the combat log it says they rolled some number under 20, NOT a natural 20. Happened multiple times with the gnolls. This could be a bug of course.

Last edited by Sigi98; 12/10/20 09:22 PM.
Joined: Oct 2020
V
member
Offline
member
V
Joined: Oct 2020
For skillchecks there might be mali at play here that are not displayed. Like I had a +4 Persuasion role, yet rolled a 4. There were only a couple skillchecks that I failed multiple times and I would wager that it is because some modifiers are hidden from you, which I do not think should be the case.

I have not seen something similar in combat though and I do keep rough track of things like this and for me they add up so far. Especially since they give you a breakdown of how these percentages come to play in the bottom left corner. But there might be some hidden stuff there as well, or things you simply miss. Like debuffs, buffs etc. and this might affect the dice after the roll as well. So the log shows a 19, but there is a +1 bonus from whatever, hence a crit happens. This might not show up in log or does so separately, or hidden or somehow obscure.

Joined: Oct 2020
S
addict
Online Content
addict
S
Joined: Oct 2020
Well, its just that according to DnD 5e rules, you only crit if you roll a NATURAL 20, so without any modifiers. thats a 5% chance if you neither have disadvantage or advantage.

Joined: Oct 2020
V
member
Offline
member
V
Joined: Oct 2020
Originally Posted by Sigi98
Well, its just that according to DnD 5e rules, you only crit if you roll a NATURAL 20, so without any modifiers. thats a 5% chance if you neither have disadvantage or advantage.


Yeah I do not know the system well enough, so I was mostly talking out of my arse here. I just figured through skillchecks that there are modifiers working both ways that are not always transparent to the player.

Joined: Apr 2014
stranger
Offline
stranger
Joined: Apr 2014
Yeah, the constant missises by party members with supposedly 80%+ chances to hit are VERY suspicious.

On the other hand I've yet to see the AI controlled MOBs hit less then 75% of the time.

The other possibility is Larian set AI armor classes and to-hit ratings too high. Thus you can't hit them and they rarely miss you.

Last edited by Suhiira; 12/10/20 10:04 PM.
Joined: Nov 2019
apprentice
Offline
apprentice
Joined: Nov 2019
Originally Posted by Into Oblivion


I have done multiple analysis in (other) random number generators and it did show me often these number generators aren't very well in generate unified distributed numbers. Especially mean value and variance do often have errors and do shift from their theoretical values.

Please do some research and check if this is the case here too.

Thank you!



I haven't had any big problems myself with the rolls and have missed plenty >80% but also hit several of the (few) <20%. What I know from these discussions from other games with random chances to hit is that many players get a scewed perspective of outcomes as they are very unlikely to take a shot at the low chance ones at all, hence they almost never experience the "lucky" hits.

However, for random chances in games where you tend to save and reload encounters a lot one of the first things that is interesting is; when is a new seed generated and how/for how long is a seed stored. If a seed is stored in a save and a new one not generated in some interval then there is a big chance you will see similar results to acions when you reload.

Joined: Oct 2020
I
stranger
OP Offline
stranger
I
Joined: Oct 2020
Originally Posted by Waeress


I haven't had any big problems myself with the rolls and have missed plenty >80% but also hit several of the (few) <20%. What I know from these discussions from other games with random chances to hit is that many players get a screwed perspective of outcomes as they are very unlikely to take a shot at the low chance ones at all, hence they almost never experience the "lucky" hits.


That is not the point. A random number generator by definition generates random numbers resulting in this use case in lucky hits / missed chances. But if you realize that there are patterns in the actual realized random numbers there need to be a check in the code.

Originally Posted by Waeress

However, for random chances in games where you tend to save and reload encounters a lot one of the first things that is interesting is; when is a new seed generated and how/for how long is a seed stored. If a seed is stored in a save and a new one not generated in some interval then there is a big chance you will see similar results to actions when you reload.


You can fix this seed problem with frequently generating random seeds, but without knowing if there is an issue and what is the reason for it, everything else is just speculation. I just want to highlight that there are patterns and someone with access to the code should have a look/ do some analysis.


Link Copied to Clipboard
Powered by UBB.threads™ PHP Forum Software 7.7.5