Larian Studios Forums Baldur's Gate III Baldur's Gate III - Story & Character Discussion Balder's Gate 3 Jeopardy

"I'll have Fair RNG Not At All Biased for $600 Alex Ken!"

"A wonderful number, it is 0.0009765625."

"What is 'What are the odds of missing 75% hit chance against a minotaur 5 x in a row?' "

"Yes! Pick again."

"I'll have Fair RNG Not At All Biased for $800, Ken!"

"An even better number, it is 0.0084375"

"What is 'What are the odds of a Minotaur missing twice, hitting once, and critting twice in five rolls when I have 19 AC and it needs 11 to hit?' "

"Indeed! Pick again, you're on (an extremely bad) roll!"

"I'll have Fair RNG Not At All Biased for $1000, Ken!"

"The best number ever heard, it is 0.00000823974609375."

"What is "What are the odds of me missing a 75% hit chance 5x, and the minotaur missing twice, hitting once, and critting twice?' "

NIce! A swing with 3 independent rolls, all critical miss! 1 in 8000 is no 0.00000823974609375, I'll admit.

Using that mod which lets you set up to 10x opponents. You may have seen Morgana doing this at 100x with a special version.

Anyway, have it set to 3x and am fighting 3 bullets solo. Ok, no tricks, just straight up badass fighting. It's tough, as expected.

However, for I've savescummed 4x now. I have 2/3 of my health (fighter) and the fights have gone like this:

Bullets turns
1. 14-15 hit, I am downed
2. 1 crit miss, but since I am downed it is a critical hit
3. 3rd bullet rolls 2, again crit, dead.

This has happened 3 times now. Note the "fixup code", where it frontloads good rolls to the bad guys and bad rolls to you, to make the fights exciting. Then the engine, later in the fight, fixes up the bell curve by giving you better rolls and the enemies worse ones, once you're probably gonna win.

Anyway, it's quite blatant if you pay attention. A hit with massive damage to down me, then rolls of 1s and 2s repeatedly to finish me off, to balance out the bell curve.

5th time reloaded, I am at the same health and it's their turn. I hit space bar.

17 hit downed
4 crit <--- Fixup to restore bell curve
2 crit dead <--- Fixup to restore bell curve


5 times in a row. This is very repeatable.

6th time. Now this is interesting

19 hit badly but not quite downed
15 hit and downed <--- Requires high roll to hit me when not unconscious
2 crit <--- Now downed, emergency twisting to fix up the bell curve

7th time
18 downed
7 crit (with bonus, 5 shy of conscious AC)
1 crit


8th time
15 not downed
14 downed
9 crit (not close to AC)
2 crit dead

9th time
17 downed

10th time
18 downed

11th time
15 downed

12th time
wiffs with a 4, miracle of miracles!

I riposte with a crit
bullet 2 then crits me, downed


Come on, man.

13th time
18
18
6

14th time
something I didn't record, but a hit, not downed
20 crit
dead

15th time
19
13 (enough with bonus exceeds 19 ac, which is -2 from acid)
downed

16th time
12 (= 21, my ac), astounding damage roll, downed

17th time
12

So keep in mind these are all independent, fair rolls. We can pick them at any random order and check their distribution.

Let's just use the first bullet attacks in each savescumm.

Only one wiff, the other 16 are 12 or more, the min required to hit my AC 21 for them.




Google, With 17 d20 rolls, what are the odds 16 are 12 or more?




AI Overview

The probability depends slightly on whether you mean exactly 16 or at least 16 of the rolls are 12 or more:
• Exactly 16 rolls: 0.002644% (about a 1 in 37,826 chance)
• At least 16 rolls (16 or all 17): 0.002771% (about a 1 in 36,090 chance)

Not only is there code to frontload enemy attacks with hidden bonuses and frontload your attacks with crud, to fix up the bell curve much later, but apparently there is ActiveSaveScummDenier(tm) technology!

Let's just look at the second bullet attack, when the first didn't quite down me. There aren't many, but they are:

15
14
20
18
20
13

Again, these are the second bullet rolls when the first does not down me. There are only 6 such rolls, but all are hits. Can we glean any statistical certainty?




What are the odds of getting 13 or more on 6 d20 rolls?


2. Getting a 13 or higher on all 6 rolls

This is what happens if you need every single die to beat the target number.
• Percent Chance: 0.41%
• Dice Odds: Roughly 243 to 1 against it happening.
• The Math: (0.40)⁶ = 0.004096



It would actually be slightly lower as they just need a 12, or even 10 as some of these I now have acid or something applied from bullet attack #1, to drop my AC by 2. Worst case, which this is not, would be 1 in 2^^6 = 64.

The average of those 6 rolls is 16.666666. The average of a d20 roll is, of course, 21 / 2 = 10.5. That is one hell of a bell curve shift, not just a little bit.

The irony is this shows the difference between decision theory and game theory. Players, mini-max or not, assume the universe i.e. the dice, are honest. This is decision theory.

In game theory (which, although this is a game, would not be game theory in this context) the universe is actively working against you.

You expect the world to be against you, in the NPCs and their actions. This is game theory, or your opposition in chess. Decision theory you expect fair rolls, so you can plan accordingly to deal with the game portion.