I wanted to investigate and see if there were any issues with the RNG.
So I recorded all of my player combat rolls on a new playthrough. I had a hunch advantage/disadvantage may be skewing the distribution, so I notated each advantage and disadvantage roll. During the playthrough I tried to get as many normal attack rolls as possible. Here are the findings.
For all rolls recorded |Frequency| column; a roll of 17 was further than two standard deviations from the average. The sample from my playthrough looked like rolls may be biased towards the player.
- N=176
- St.dev 3.3
- Average 8.8
Note: if one of the 17s had been any other outcome this sample would have been completely normal.
After stratifying for normal rolls (Frequency Adv/Dis Removed); all outcomes were normal.
- N=117
- St.dev 2.6
- Average 5.85
Advantage and disadvantage are so common in Patch 3.0 that d20 rolls can be skewed for or against the player. Overall, I don't think the RNG is flawed.
DataRoll_|Frequency|_(Frequency Adv/Dis removed)
01---------|05|----------(03)
02---------|08|----------(07)
03---------|10|----------(05)
04---------|09|----------(06)
05---------|14|----------(11)
06---------|04|----------(03)
07---------|05|----------(03)
08---------|06|----------(05)
09---------|06|----------(04)
10---------|14|----------(10)
11---------|12|----------(07)
12---------|07|----------(03)
13---------|09|----------(07)
14---------|10|----------(06)
15---------|06|----------(04)
16---------|08|----------(05)
17---------|16|----------(11)
18---------|11|----------(08)
19---------|07|----------(03)
20---------|09|----------(06)
EDIT: While I am saying the ratio of rolls isn't flawed. There is still a possible case to be made that pseudo-RNG might be better. On the Nautiloid I started out with a long streak of 5s, 10s, and 17s. A 5 will usually hit on the Nautiloid, so I wasn't bothered. That was a streaky start to the game and I can understand that if a player streaked with 2 instead, that could be an annoying experience.
Pseudo RNG that avoids long streaks may be better for the game. Each session, the player only experiences an hour to a four hour window of the game, which will always be a small sample size.
