Thread: Monte Carlo versus "The Math"
Tuesday, 17th November, 2009, 12:25 AM #91
Superhero (Lvl 15)
- Join Date
- Aug 2006
- Lynn, MA
ø Ignore keterys
Sfedi, you should pickup that once per encounter 2d10 attack for your soldiers there.
- EN World
- has no influence
- on advertisings
- that are displayed by
- Google Adsense
Let's see what changes:
Level 1 Dwarf Greatweapon Fighter VS. Level 1 Soldier
Level 1 Dwarf Greatweapon Fighter VS. Level 2 SoldierStatistics
Level 1 Dwarf Greatweapon Fighter VS. Level 3 SoldierStatistics
Level 1 Dwarf Greatweapon Fighter VS. Level 4 SoldierStatistics
Level 1 Dwarf Greatweapon Fighter VS. Level 5 SoldierStatistics
Magsman (Lvl 14)
- Join Date
- May 2008
- Columbus, OH
ø Ignore Mengu
Damage is too much on that soldier. It should increase every other level.
Level 1: 1d10+3
Level 2: 1d10+3
Level 3: 1d10+4
Level 4: 1d10+4
Level 5: 1d10+5
On another note, what I would like to see is if this character had Toughness instead of Dwarven Weapon Training, Comeback Strike instead of Brute Strike, and Warhammer and Shield instead of Mordenkrad, would his survivability increase or decrease against the soldier? Your Comeback Strike AI would be something like "if bloodied, after second wind has been expended, use Comeback Strike".
Going against the second level soldier we see a massive 28% death rate in the first encounter. My gut says the adjustments would lower that rate, while prolonging the fight.
Warning: This post may contain sarcasm.
Oh, and I left the "100 surges" from a previous simulation.
Here's the above but all of the above corrected:
Note that Reaping Strike deals less damage.
I've implemented Comeback Strike but didn't modify the AI because it seems the current AI covers this.
The AI chooses the more damaging power that is Encounter or more frequent.
When it becomes Bloodied, then it chooses de more damaging power that is Daily or more frequent.
So it seems this should work fine.
You can check the code posted in previous posts.
The only problem with this AI is that Comeback Strike has the same average damage as the encounter power, so it may happen that the Dwarf becomes bloodied, he hasn't used his encounter power yet (because he missed all attacks), and then ComebackStrike and Steel Serpent Strike are equal choices.
Here's the same simulation as before (in this same post) but with the more defensive build:
Note that, for the time being, Truename's simulator is much better for this analysis than mine.
As sfedi mentioned, my sim is more accurate (for now)--mine includes action points, which allows the dwarf to one-shot the soldier during most first fights. He only dies 2.8% of the time in the first fight against a level 1 soldier. When facing the level 2 soldier, he dies 9.4% of the time.
Dwarf Fighter vs. Soldier
Code:Days simulated: 10000 Avg rounds per fight (when survived): 3.0 1 (17.2%): =============================================== 2 (41.4%): =================================================================== 3 (65.1%): ================================================================= 4 (82.0%): ============================================== 5 (93.0%): ============================== 6 (97.6%): ============ 7 (99.3%): ==== 8 (99.8%): = Avg fights survived: 2.6 0 (9.4%): ========================= 1 (34.1%): =================================================================== 2 (46.4%): ================================= 3 (66.5%): ====================================================== 4 (80.4%): ===================================== 5(100.0%): ===================================================== Avg surges remaining (survivors only): 0.7 of 10 0 (72.9%): =================================================================== 1 (83.1%): ========= 2 (89.4%): ===== 3 (94.5%): ==== 4 (97.1%): == 5 (98.7%): = Survival rate: 19.6%Example of 0 fight(s) survived (9.4% chance):
Example of 1 fight(s) survived (24.7% chance):
Example of 2 fight(s) survived (12.3% chance):
Example of 3 fight(s) survived (20.1% chance):
Example of 4 fight(s) survived (13.9% chance):
Example of 5 fight(s) survived (19.6% chance):
I'll try out your more defensive build later in the week.
Novice (Lvl 1)
- Join Date
- May 2005
- odense, denmark
ø Ignore Goolpsy
Awesome idea you got.
I've got some point for you, that might make some things easier.
To simplify thigns you need to make some Guidelines, which could be:
1) Real players wouldn't make obvious bad choices
2) Real players are seldom perfect at creating their Char, even when they try.
3) The simulation is made to give a general idea, not the OPTIMAL scenario
Once you got combat figured out, with multiple PC's and Multiple monsters do something like this: Get Enworlders to make you 10 Characters of each class/role at lvl 15.
Running the simulation you will get a ranking at which Build is best (ignoring a possible synergy 2 different roles/builds between). Here i suggest using the 75% best choice, bad on the guidelines above:
below 50% might be 'Noticeably bad build' and would be used as much(1).
The mean of 51% to 100% is 75%, making sure our "General idea of combat"(3) isn't based on 'perfect' Chars(2).
Once you've got the Builds selected, extend the level range and run the party at different levels.
I further suggest you neglect Movement. Unless you have a clearly advantageous/Disadvantageous terrain, or the battlefiend is very large, Movement won't make much of a difference (IMO) - This greatly reduces simulation time and complexity.
Note: Flanking factors could be added in as a percentage based on the monster to PC ratio.
Further: Instead of running 5vs5 combat, i suggest an encounter table with 10-20 different encounters to choose from. (Again im hopeful that the Enworlders will assist, if it needs to be made over many levels). These encounters should vary in monster types and numbers.
Note: Since you do not need all the monster stats, when you get to this point, just write which Stats are to be implementet. ->Will reduce implementation time and encounter development time for the helpful souls
Last point for now: "What if the monsters focuses a particulaly team member?"
Just randomize possible Focus patterns between not focusing and focusing a random member.
As for PC AI, just run some simulations and see what Monster types/roles are best or (75% best) to focus and implement according responses. (I.E is it best to kill of controllers or minions first? and in what order)
Mess With The Best... Die Like The Rest
Waghalter (Lvl 7)
- Join Date
- May 2007
ø Ignore eamon
Given the even/odd fight split you're likely to get overall higher survivability if you delay the action point when the fight is going (extremely?) well - reducing variability should favor the stronger party; clearly the PC here...
4e balanced random loot system / 3.5e death&dying variant
Spellbinder (Lvl 16)
- Join Date
- Jul 2004
- A Nation's Capitol
ø Ignore TerraDave
Greatweapon Fighter vs. Guardian Fighter
Mengu suggested comparing a Guardian Fighter to the Greatweapon Fighter we've been using:
I had some time to work on the sim today, so I coded it up. The results impressed me. The Guardian Fighter and Greatweapon Fighter are nearly perfectly balanced against each other. My sim showed the Greatweapon fighter surviving 60.2% of the fights. The Guardian fighter survived 58.5%. The results were so close, I had to run them again at 10x the resolution (100,000 simulations rather than 10,000), even though that takes over ten minutes each to run.On another note, what I would like to see is if this character had Toughness instead of Dwarven Weapon Training, Comeback Strike instead of Brute Strike, and Warhammer and Shield instead of Mordenkrad, would his survivability increase or decrease against the soldier?
The final results: Greatweapon, 60.3%; Guardian, 58.0%.
I'm impressed. Although they use a lot of the same powers, the two builds are pretty different. The Greatweapon fighter averages 33 damage in the first round when he hits and has an action point. This often allows him to mop up the soldier before he gets a hit in. I see it all the time when I look at the combat examples. The Guardian fighter, on the other hand, averages 25 damage under the same circumstances. I just don't see the rapid mop-up with the Guardian.
If you look at the example combats, the two builds play out very differently... but their survivability is practically the same. Bravo WoTC.
One other thing of note: although the survivability is the same, the Greatweapon build plays a lot faster than the Guardian build. Half of the Greatweapon's fights are over in two rounds or less. The Guardian's fights take twice as long (and take twice as long to simulate, annoyingly). The unsurprising moral? If everything else is as well balanced, the key to eliminating grind is to deal lots of damage.
Dwarf Greatweapon Fighter vs. Soldier
Dwarf Guardian Fighter vs. Soldier
(Note: in some of my initial trials of this match-up, I saw that both combatants were running through too many surges. I dialed back the AI so that it's only willing to waste half of a surge during a short rest (rather than 3/4 of it), and that increased survivability for both builds. It also narrowed the gap--the Greatweapon fighter's survivability was previously about 3% higher than the Guardian fighter's.
Next, I'll see if this balance holds true against higher-level foes.
By RyvenCedrylle in forum D&D and PathfinderReplies: 73Last Post: Saturday, 5th October, 2013, 04:38 AM
By innerdude in forum RPGs & Tabletop Gaming DiscussionReplies: 10Last Post: Monday, 8th March, 2010, 01:02 PM
By Dark Jezter in forum Miscellaneous Geek Talk & Media LoungeReplies: 27Last Post: Tuesday, 10th May, 2005, 11:50 PM
By old school 1E in forum D&D and PathfinderReplies: 7Last Post: Thursday, 27th January, 2005, 05:00 PM
By der_kluge in forum RPGs & Tabletop Gaming DiscussionReplies: 41Last Post: Wednesday, 22nd December, 2004, 04:01 AM