Turning Ninth Age into a computer game/simulation, with AI

This site uses cookies. By continuing to browse this site, you are agreeing to our Cookie Policy.

The latest issue of the 9th Scroll is here! You can read all about it in the news.

Our beta phase is finally over. Download The Ninth Age: Fantasy Battles, 2nd Edition now!

  • Gingersmali wrote:

    work, unit are either in combat or not and there has to be some sharp boundary when these things change.
    But we have the option to make it into a probabilistic boundary, like if you are a few degrees out of arc you can roll a dice and get in with some probability.

    Or if you have a very questionable flank charge there can be intermediate results, or probabilities or something.
  • Auto2 wrote:

    Yes but I am say that that's unrealistic and makes for un-fun, slow, nitpicky gameplay, and ideally the game should change in the future to make micro-optimization of position less important.
    i´d say iti is far from unrealitic and is what makes or breaks the game for most of the players that use rules at all as a foundation of their games. Rules are there to be applied.

    I massively doubt this part of the rules will change at all in the far future.
    WTC OPEN 2020 - 25.-26.04.2020 in Amsterdam
    an official T9A event
  • i agree that movement detail is crucial to keep the game "as is", and that any simplification would give us an essentially different game. not saying it's necessarily a bad thing, tho. @Frederick, i'm really interested in your opinion here: what about a half inch resolution? imagine you can only move/wheel/pivot in half inch increments. would that drastically change the nature of movement?

    from a more general perspective, i think it would be useful to keep movement separated from the rest. if i wanted to go with AI, i would first focus on getting done calculations and decisionmaking for combat/shooting/magic. only after that, i'd work on movement. AI simulations can be complex, the game is very complex, maybe we should take one step at a time.


    click the logo and reach the KickStarter page!
  • Frederick wrote:

    Auto2 wrote:

    Yes but I am say that that's unrealistic and makes for un-fun, slow, nitpicky gameplay, and ideally the game should change in the future to make micro-optimization of position less important.
    i´d say iti is far from unrealitic and is what makes or breaks the game for most of the players that use rules at all as a foundation of their games. Rules are there to be applied.
    I massively doubt this part of the rules will change at all in the far future.
    Really? We always played with "Eh, close enough, let's say that counts". Sometimes also "That rule is silly, let's not do that.", but that was rare.
    A summary of all proposed ideas from the VS LAB brainstorm thread

    MAKE THE SWARM COWARDLY AGAIN!
    WE DEMAND TERRIBLE LEADERSHIP!
    DOWN WITH COMPETENT GENERALS!
  • piteglio wrote:

    what about a half inch resolution? imagine you can only move/wheel/pivot in half inch increments. would that drastically change the nature of movement?
    yes, this would drastically influence games. it is the mm´s that matter often to have the proper angle to show enemies your front or something for several units that would love to have one in the flank and one in the front for their charges. having only 0,5" steps would mean i will often have situations i have to choose between pestilence and cholera while with the normal movment rules i could just pivot the unit a small tad to have the perfect angles.

    Don´t change movement. It´s the heart and soul of the game and changing it ime will make the PC version end up being a flop (at least to T9A players).
    WTC OPEN 2020 - 25.-26.04.2020 in Amsterdam
    an official T9A event
  • Eldan wrote:

    Really? We always played with "Eh, close enough, let's say that counts". Sometimes also "That rule is silly, let's not do that.", but that was rare.
    agreements are the basics of the games on tournaments for sure. a computer-game will not have the option to agree on "this is out of line of sight, right?". Any misplacement will mean the system will backfire on you and ruin your day.
    WTC OPEN 2020 - 25.-26.04.2020 in Amsterdam
    an official T9A event
  • Frederick wrote:

    while with the normal movment rules i could just pivot the unit a small tad to have the perfect angles.
    yes but that is not a good thing as far as I am concerned.

    The original intent of movement in wargaming is not to beat your opponent by picking a position that's super optimized because in a real battle you would not be able to do this.

    In real life you cannot magically funnel enemies into your front facing by shuffling your feet a bit and I would say that T9As movement system is bugged because it allows this.

    Also having a discrete set of possible movements helps to simulate the reality of command and control - you cannot tell a real unit of swordsmen to make the angle of their right foot with the enemy exactly equal to 44.99 degrees.
  • well, i disagree that such degree of realism is what the project aimed for when creating the game.

    i will not stop you from doing your PC game - i just want to raise my concerns about this not being successful when the actual rules are not 100% reflected in it.

    As an example: would you play a version of a beloved boardgame online, when suddenly the PC game would change the rules drastically?


    For T9A the PC version should be seen as an addition to play practice games for the real games with your toy soldiers. To have this making sense the PC version should be ideally exactly like a game at a real table.
    WTC OPEN 2020 - 25.-26.04.2020 in Amsterdam
    an official T9A event
  • Frederick wrote:

    raise my concerns about this not being successful when the actual rules are not 100% reflected in it.
    The actual rules are going to cause problems for a computer game though. Despite what some people have said I still think that the set of allowed moves needs to be finite rather than infinite or else various problems will arise.
  • @Frederick

    Also, I'm thinking about what the core rules for the next major edition for the game should be.

    Maybe it would be a good idea to make the game less about micropositioning and more about strategy, moving through turns more quickly and executing a plan rather than spending 5-10 minutes every movement phase with a laser line and a ruler.

    The post was edited 1 time, last by Auto2 ().

  • Auto2 wrote:

    Despite what some people have said I still think that the set of allowed moves needs to be finite rather than infinite or else various problems will arise.
    Its not infinite in RL as well. Its common misconception but in fact human can play only up to certain detail. If you have differences in placement of roughly 1mm they are sometimes extremely difficult to measure ending up with two players not being able to tell the difference. Even if it sometimes means games being lost we do not have in RL infinitely many possibilities of placement. Simply its not half an inch but more like a millimeter. Its not too much of a hassle thought for most software solutions to represent that. Even if we drive it to 0.5mm we end up with biggest distance on the battlefield (from corner to opposite corner) to measure 4400 steps of 0.5mm. Measuring angles from that and checking whether particular point in such grind is occupied or not is rather not going to cause precision errors for regular applications i guess. And really making it 0.1 mm is rather not going to break the system as well.

    The post was edited 2 times, last by slivek ().

  • You should use continuous variables for the state space. Each unit's state would then be xy for posicion of center, orientation angle, and holding the width/length of unit. But in order to take any decision, you must take into account all the other states of all the other units in the battlefield. This isn't a limiting factor, and the actions by themselves could be represented by the set of positions you can actually go to (in case you move) or charge or whatever. However, you'd need some AI planning because more than often you move your units thinking 2-3 turns ahead, and some policy for the action decision-making part, which should consider some notion of a reward function based os some belief of a certain gain in terms of advantage for winning.

    The problem is not state representation, it is making it learn to do something meaningful.
    Building reward functions is very tricky, you often end up getting an output you didnt expect but that yields a better reward than the supposed one.

    If anyone wants to test things, id like to see how it goes.

    BH Army Community Support

    Mathemagician

    Stop SE favouritism! we want green nametags for everyone!
  • slivek wrote:

    we do not have in RL infinitely many possibilities of placement. Simply its not half an inch but more like a millimeter. Its not too much of a hassle thought for most software solutions to represent that.
    Yes, this is true and you could make a computer version of a game that used +- 1mm for positions of the center of a unit.

    You will probably have some issues that need to be dealt with to do with the angle of a unit, how to discretize angles and how to represent the set of points on your chosen grid that a unit at a certain angle occupies, how to adjudicate contact between units, how to avoid abusive scenarios where it seems like a unit can touch another but actually there is a 1mm positional discrepancy somewhere that prevents combat (which I am pretty sure will occur and be hard to debug)

    These is also the issue that you have made the state-space of the game explode compared to what it really needs to be which will complicate any attempt to use AI. And if you actually make the AI understand this huge space you have the danger of the AI learning how to abuse micropositioning better than human players can, which segues nicely with what @goatman says here:

    goatman wrote:

    often end up getting an output you didnt expect but that yields a better reward than the supposed one.
    Yes, the AI might learn a very advanced version of something abusive. If you can get it to learn anything at all!