-
December 4th, 2020, 13:28 #21
- Join Date
- Aug 2019
- Posts
- 2,025
I suggest everyone who is concerned about weighted 3D dice to use the modded extension. We can neither prove nor disprove what practical impact 3D dice have on actual game-play, but this extension allows us to get more randomly spread dice results to begin with.
At least you get a choice here.
-
December 4th, 2020, 14:33 #22
- Join Date
- Aug 2019
- Posts
- 2,025
One clarification: With Instant Dice you can still drag rolls to the chat and onto tokens and it will still do 3D rolls then. Just all those click-rolls and automated rolls are replaced with a simple random number generator.
-
December 15th, 2020, 09:24 #23
- Join Date
- Aug 2019
- Posts
- 2,025
Curiously the very first roll after a campaign start is not handled by Instant Dice, so the first roll is a 3D roll. Yesterday we also had several single instances during the campaign where 3D dice were rolled instead of ID making the roll, despite the roll being a click-roll (no drag).
-
December 16th, 2020, 15:29 #24
- Join Date
- Aug 2019
- Posts
- 2,025
Distribution of 1 mio. D20 results using Instant Dice (modded) vs. Random.org (real random numbers):
Expected: 5%, standard deviation: 0.021%
Expected: 5%, standard deviation: 0.023%
For comparison, standard deviation of FGU 3D: 0.248%, FGC 3D: 0.137%
Doubles (tripples, quadruples, etc. count as multiple doubles):
Expected: 0.25%, standard deviation: 0.006%
Expected: 0.25%, standard deviation: 0.006%
For comparison, standard deviation of FGU 3D: 0.027%, FGC 3D: 0.017%
-
December 17th, 2020, 12:18 #25
- Join Date
- Aug 2020
- Posts
- 94
Wow, great work visualizing this numbers. As far as I know Lua uses Mersenne Twister for RNG and unless you do cryptography or quantum mechanics simulations there is practically no difference to true RNG. The random seed must be properly randomized of course.
Big thanks to Weissrolf for pointing this out, did not expect FG to fix the seed...
I will update first post and incorporate feedback in the coming days.Last edited by xelab; December 17th, 2020 at 12:21.
-
December 17th, 2020, 12:25 #26
- Join Date
- Aug 2019
- Posts
- 2,025
AFAIK Fantasy Grounds uses Lua 5.1, which in turn uses the old C Rand instead of Mersenne Twister. I tried to implement some other RNG, but this was not possible with the stripped down Lua of FG. C Rand is implemented differently on each OS platform, so the results are only (fully) valid for my Windows 10 installation.
Keep in mind that I only showed distribution here. Consecutive rolls of 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20 ,1,2,3... would also be uniformly distributed, but not random.
I did show better random capabilities from Instant Dice (math.random) compared to 3D dice in another (locked) thread, though. It's still rather weak, especially compared to something like PCG32. But likely better than rolling tumbled dice.Last edited by Weissrolf; December 17th, 2020 at 12:42.
-
December 18th, 2020, 20:01 #27
- Join Date
- Aug 2019
- Posts
- 2,025
It's worth mentioning that for D8 rolls FG Classic 3D dice create measurably better randomness than Instant Dice (math.org = C Rand) on my W10 system.
If FG Unity would use Classic's dice physics then its results were far better.The latest 4.0.5 changes seem to improve results,though, despite looking rather funny (like dice doing backflips short before stopping). Still checking...Last edited by Weissrolf; December 19th, 2020 at 00:17.
-
February 15th, 2021, 07:35 #28
- Join Date
- May 2019
- Posts
- 339
-
February 15th, 2021, 07:52 #29
- Join Date
- Aug 2019
- Posts
- 2,025
I removed the download link by request. Was this extension improved since its original inception. Currently it keeps rolling 3D dice for the very first role of each player and even some rolls in-between. Not much of a real problem, just wondering.
-
February 15th, 2021, 08:01 #30
- Join Date
- Aug 2020
- Posts
- 94
Must be an incompatibility with some other extension. This never happened for me on either FGU or FGC, and is not reproducible by me.
The extension has indeed been improved (due to your feedback Weissrolf) incorporate proper random seed setting at initialization.
Thread Information
Users Browsing this Thread
There are currently 1 users browsing this thread. (0 members and 1 guests)
Bookmarks