We need a pathfinder combat simulator

[Mok]

Thinking about the other thread about statistics on rolling for hit points it reminded me of the 90's when I used to play a lot of Axis and Allies via PBEM.

One tool that I had back then which was very handy was a combat simulator. It let you plug in the forces for a battle and then it would run the combat for you, except it would run 10,000 instances of that combat, average up the results, and then give you a percentage chance of success.

Because it was PBEM, you had plenty of time to fuss over your turn, and so this simulator was a great way of min-maxing your forces to get the best odds you needed for either attack or defense.

Now, A&A combat is trivial compared to Pathfinder combat. The program was easy to make (I didn't make it, but I imagine it was easy to make) because the decisions are pretty much no brainers.

Still, in this day and age, if you can model nuclear explosions accurately through computer modeling then I assume that given enough money and time then a Pathfinder simulator.

But why would you want it? Right now the underlying math of the game is at best some spreadsheet calculations, but for the most part eyeballing it. If you have a simulator that can run through thousands and thousands of instances of a battle, and several variations of a battle, then you'd get a vast amount of data on how the game actually plays.

Right now with playtesting on these boards the devs are looking for data on actual combats. They are using crowdsourcing to replicate a combat simulator, using our collective brains as a bunch of computers to perform and report data. It's awfully inefficient, but better than nothing.

So from a design perspective a combat simulator would be incredibly valuable. You'd really hone in on numbers. You'd be able to have it output real DPR values across all levels and with loads of different character builds, all in the blink of an eye. CR values could really be dialed in, and allow for far more granularity. The endlessly controversial class tiers theory could be put to the test.

People might ask, "what about the AI?" At first you'd likely be using canned scenarios to establish particular benchmarks, but then it just comes down to money, time and technology. AI development is advancing, and now we even have programs that are out thinking scientists.

I think all of this is doable, it's just an issue of how much time and effort is needed to do it? All of the data of the game is already digitized, and much of it is already in table formats. If there was a way to extract all the data from the Hero Labs files you'd likely have the bulk of the game already in a digital format to start being coded for use.

"We can rebuild it. We have the technology. We have the capability to build the world's first Pathfinder RPG Simulator. It will be better than it was before. Better, stronger, faster."

[Bobson]

I definitely like the idea, but I'm not sure it's anywhere near feasible.

First of all, you'd have to have a way to import and/or specify fully fleshed out characters. This is doable - you just need to program to read .pcg files and whatever format HeroLab saves in.

Next, you'd need some sort of AI, even if it was just "choose a random option from all possible options". But even that becomes problematic. A character as low as a 5th level wizard has 10 spells prepared before counting bonus spells. Plus they have whatever abilities their school focus gives them. Plus they may have a actual weapon. And each thing they do affects and is limited by their positioning, their available actions, and what they've done previously. Their opponent is likely to have a similar set of abilities to choose from.

You'd also have to actually model what each spell and ability can do, as opposed to just listing it like the chargen programs do. If a fighter has a one-handed weapon and a buckler, each attack he makes can be either one-handed or two-handed (with the effect on his AC and attack rolls), and if he has power attack, that's now 4 options... Not as many as the wizard, but he has the same set of options next turn too.

You could certainly build something that pits two randomly-choosing characters against each other, but it wouldn't really prove much of anything.

-----------------

All that being said, I will certainly not deny that it is possible to do. We model the weather, with some degree of accuracy. We have easily-available chess-playing programs that can beat most people, and go-playing programs which can do a fairly decent job. We have simulators for WoW characters and for traffic flow. We even have software that could grow a reasonable AI using iterative "genetic" logic, where the best ones got to spawn new ones. Given sufficient computer power and comp-sci time, it would be possible to model a combat. But it's sadly not feasible to do anything more than DPR calculations.