I've already touched this issue on other places here but I feel this topic deserves its own thread. Some of you did read the reddit post with quick analysis of the HP vs AC issue. I've taken this a little bit further and incorporate proper fight simulation with ciritcal hits and misses and with advantages and disadvantages. My goal was to compare RAW (Rules As Written) DnD with modified BG3.

The testing scenario contains 1st level immortal STR 16 Figher vs 1 million of Goblins. For simulation I did use standard Monte Carlo method with Mersenne Twister random generator. I did some tests with even more random cryptographic generators but results for rolling dice were roughly the same.

For modded BG3 I used Goblins with AC 8 and HP 14. Also 80% of attacks were made with advantage and 20% were made normally or with disadvantage to simulate backstab and high ground rule. For RAW DnD I used Goblin with AC 15 and HP 7. 80% of attacks were made normally while 20% have either advantage or disadvantage. This simulates no backstab nor high ground advantage.

Code
Modded BG3   Rounds    Mean   Min  Med  Max   StdDev  Round>4
-------------------------------------------------------------------
Dagger       3085444   3,1    2    3    7     0,5     10688 (1%)
Shortsword   2677084   2,7    2    3    8     0,6     4588 (0%)
Longsword    2402956   2,4    2    2    8     0,6     2312 (0%)
Greatsword   2000084   2,0    1    2    6     0,4     208 (0%)

RAW DnD      Rounds    Mean   Min  Med  Max   StdDev
-------------------------------------------------------------------
Dagger       3184108   3,2    1    3    21    1,8     192228 (19%)
Shortsword   2729368   2,7    1    2    19    1,7     141336 (14%)
Longsword    2497792   2,5    1    2    19    1,7     116516 (12%)
Greatsword   1968412   2,0    1    1    19    1,4     57424 (6%)


[Linked Image]

Results are surprising. Total time for killing 1M of Goblin is roughly the same in BG3 as in RAW DnD. Larian manages to keep same average numbers while significantly reduce the long miss streaks. Those horrific max values of round-to-kill-a-goblin for RAW DnD are very rare but number of goblins which takes more then 4 round to kill in RAW DnD are significantly high. BG3 is more consistent (see the standard deviation) than RAW while possibly a little bit slower judging from minimal value and median. Kudos for them for changing martial combat to feel better. But this change has very deep consequences (see e.g. this thread), including very big indirect nerf of:

  • damaging spells and abilities vs DC (no advantage for those spell, enemies have more HP but same DC)
  • spells and abilities which grant advantage or impose disadvantage (because backstab or high ground is easier and cost no resource and everybody can use it, no team cooperation is needed)
  • spells or abilities requiring concentration (concentration needs to be hold for a little bit longer due to higher median but concentration checks are much more frequent due to damage from surface effect)


If we are telling Larian to discard AP/HP changes and backstab/high ground rule I think we should present an alternative system which gets rid of those long miss streaks or at least the bad feeling associated with them. My proposals so far:

Do not interper all misses as misses

This is certainly not a silver bullet but it can help especially with bad feelings. Depending on how close was attack roll to AC, there could be a hit with no damage accompanied with an animation (sword hit a shield or helm, sparks, dust, chips) or sound (taunts, shouts). For example we could have some uniqe shout after 4 misses in a row.

Pseudo-random die roll distribution

This is implemented especially in competetives games like Dota 2 (see their wiki). In this implementation the event's chance increases every time it does not occur, but is lower in the first place as compensation. Results are more consistent but statistically the overall probability is same as true random generation.

Simultaneous enemy AI

This is another indirect method. The miss itself is not that bad if we can strike again as soon as possible. This may be a reason while some people still advocates for RTwP. In RTwP everything is happening simultaneously and you need to wait for next attack roll 6 seconds in max. But in turn-base mode, when every enemy takes their turn in series you may wait a whole minute! Since RTwP is off the table let at least grouped enemies play their turn simultaneously.

Larger party and smaller groups of enemies

This is controversial proposal and may go either way, but since we are brainstorming ideas here lets include it. The general idea is to reduce the ratio of player controlled characters vs number of enemies to allow player simply play more often. And also elevate the party concept of BG. Maybe your 3 warriors missed this turn and your main wizzard is out of spell slots, but your Arcane Trickster still has a Magic Missile to not risk another miss.

And that is all I have at the moment. What are you thought about those solutions and about the analisys itself? Did I make a mistake maybe?


Last edited by Zahur; 02/11/20 12:38 PM.