Monte Carlo versus "The Math"

[Incendax]

Stereotypes exist for a reason, and I believe that a short term goal for this project should be calculating a party of four (Controller, Defender, Leader, Striker) using commonly selected powers with commonly selected feats fighting their way through four encounters per day. Each encounter should be randomly pulled from an encounter pool of maybe 20 different scenarios.

Run a couple million times we should be able to discern survivability at various keystone levels for what would be considered a typical party.

 

[Infiniti2000]

...at levels 15, 20, and 25. Simplify it by just sticking with PHB classes, feats, and items and with MM monsters. 1-on-1 fights don't interest me. They'll never be a real part of the game and are completely useless as a test case. If I can't see how the simulation works with a full party of FIVE (not four), with some reasonable terrain mixtures, then you're just pissing in the wind.

We'll see you sometime next summer, 2012.

 

[sfedi]

Great thread.

I'm a little worried, however, about the example we're using.

A Dwarf fighter and a Soldier are both non-average creatures (PC and Monster respectively).

We should be comparing a PC with no race (but assuming primary stats between 16 and 18, as the game does)
vs
Skirmisher, which has all the average numbers.

The Dwarf has a racial ability that is EXTREMELY useful in survivability: second wind as a minor action, this essentially adds 25% to it's hit point total, so the results are going to be skewed.

On the matter of how to model a group, and account for tactics, there are several things that could be done:
- Define best tactic (concentrate damage on one foe at a time)
- Define worse tactic (spread damage as much as possible)

And measure the two.

We could add a third one: damage enemies randomly, to have a "control case" for the above)

Another way of accounting for tactics is to add bonus to attacks, penalties, make combatants loose turns, deny focus fire, provoque OAs, etc and add those bonuses/penalties to simulate good or bad tactics.

Good job so far.

 

[keterys]

The dwarf's not so bad for modeling _someone_ in the party having a heal available.

 

[Jonathan Drain]

I'm particularly interested in seeing:

1. A party of four or five against various encounters (real D&D isn't fought solo, 4E's fighter is a damage-taker and not a damage-dealer, fighters are expected to have the support of a cleric, and so on)

2. Analysis of the duration of combat as one increases in level, and in comparison to 3E

3. Analysis of the real-time duration of a combat encounter, and comparison to 3E: at a given level, does a 4E combat take more time than a 3E combat?

 

[Truename]

I'm particularly interested in seeing:

1. A party of four or five against various encounters (real D&D isn't fought solo, 4E's fighter is a damage-taker and not a damage-dealer, fighters are expected to have the support of a cleric, and so on)

It's high on my priority list. First I need to finish the basic fight mechanics... I still need to add initiative, monster encounter powers, and missing on a natural 1.

2. Analysis of the duration of combat as one increases in level, and in comparison to 3E

I'm already generating charts that show how many rounds combat takes, and I plan to keep doing so. I won't be doing any 3e comparisons.

3. Analysis of the real-time duration of a combat encounter, and comparison to 3E: at a given level, does a 4E combat take more time than a 3E combat?

I might do time duration estimates, but if I don't, it's pretty easy to get a feel for times based on the number of rounds.

 

[ExploderWizard]

And, thus we learn how WOTC gets others to design thier software free of charge. Carry on. This is interesting.

 

[Truename]

If the Dwarf goes into every fight fresh, then Elric is right, the data should show the most results at zero. Is the dwarf at the start of the second encounter impaired in any way he's not at the start?

He must be. The code's not designed to support hypothetical rules like "infinite surges"--I just hacked it in, and I must have missed something. I'm not going to try to find the bug, though, since an analysis of a dwarf with infinite surges isn't where I was planning on going with this.

 

[keterys]

Infinite surges isn't part of the system, but nor is infinite fights. It's often the case that there's no reason not to just spend the surge to be at full hp.

 

[Storminator]

I think your results agree with the DPR analysis more than those early posts indicate. DPR should give you the MEDIAN number of rounds, which is the point when you win 50% of the fights. For example, in your first graph, the one with DPR giving 4.6 rounds and the MC giving 5.5, 57% of the fights are over by round 5.

It makes sense that the mean is skewed high. Variation from the average can only shorten the fight so much - it is going to take at least 2 rounds to kill that soldier. But on the long end, the fight can go long as 13 rounds. The 13 is going to pull things high a lot more than the 2 pulls things low. This same effect is why home prices are usually listed as median home price. One millionaire throws the mean out of whack.

Interesting project!

PS