Multi player before single player?

That is pretty cool. But can you trust an AI to come up with all the tactics that a human could? Or will it lead to such incredibly strong AI that it is absolutely undefeatable? In any case an AI would have to be developed first. In order to do so I think it would be best to start with human players.

Yes, I said that. But the NPC AI is detached from the player AI. And I think the player AI is far more complex as it has to appear to be human.
And if multi player has already been decided upon then might as well get it out of the way first. At least you'll be sure you have it. And the game would then be ready for market in it's first incarnation.

Personally as long as they are not an attacking force or interacting directly with me I am not that concerned with what I see and what the other player sees. AS long as the meta is equal from what they see and I see I am OK if I see something slightly diffrent. I.E. I do not care if I see an enemy peep leaving one building and he sees it leaving another building.

Humans have a unique ability that is difficult to emulate or program into AI, currently, and that is our chaotic propensity at creatively ignoring the logical or intuitive solutions and doing the unpredictable irrational thing. Basically, we change the game by not playing fair, or by making new rules.

Ok! I apologize for the delay here, I wanted to speak with tech people and clarify some things first. Preface - I'm not an AI Programmer but I'll try to explain the things told to me on the subject and also apply the design considerations we've made so far.



Firstly, Ba'al you're actually right in the gains we get from not including an adversary AI God over having PvP multiplayer, but this apparently does have a different set of issues with netcode, architecture and the complexity/depth of how far we can take things like NPC Pawns.



One of the problems highlighted to me was synchronization and where the data is passed or processed - is it client/client, or client/server, how will it handle DC, how will it handle abuse etc.



long development time before game gets into the hands of players

This is something I've personally pre-empted in Design. We're launching Phase 01 out of the gate as early as possible to provide a sandbox look at a single village, so that the players or testers/demo people have something to look at as quick as possible. Phase 01 doesn't have an adversary God AI for this specific reason - we want to see what you, as a player, does first in the game with a village. We want to test, iterate, collect feedback and add/reinforce parts of the game that are fun whilst crushing/cutting out things that people dislike. 


many iterations between play testing and further AI development

This is true; we can't avoid this regardless. Many iterations for NPC village manager, many, many, many hours expected in the game also running automated without a God for QA to see how each game round plays. First we need to know that a village, without you or any God, can do the things it needs to. Then we add you, as a God, and they need to respond as you, the player, would expect. Then we add another village, and have them interact as you would expect. Then we add a second God, so on. Every single step of this includes testing, QA and feedback collection.



long time before the game may start generating income and publicity


Thankfully, the only danger of this is if someone else makes the same game before us. I can't control that, but I don't want to rush the game out either. Our minimum viable is Phase 01, which is where we want the public to have the change to play it. 


risk of ending up with a game that is not challenging enough

I agree with this, we can definitely end up with a weak AI or not engaging gameplay loops that result in the player feeling unchallenged. I'm anticipating this in Phase 01 and it's partially why we're doing so many paper prototypes; trying to identify how players can win or resolve problems, seeing what would motivate them, what would hinder them. For example: I asked another designer to start the Phase 01 prototype on a whiteboard. Place homes, place a forest, place a mountain, now you spawn, how will you take the village? He did the grab Idol, smash Idol on village. We then cycled through problems; you don't have enough food, what do you do? You now don't have a forest, what do you do? There are boulders blocking the mine, what do you do? Basically, without adding anything new, we use our pool of user stories to overcome these things. We're looking to prototype these concepts in the gameworld literally as early as humanly possible, so that before we actually have anything that looks 'good', we will aim to answer these questions in a prototype level.

risk of players losing interest in the game due to long development process

This is a personal fear of mine, and I think it is a risk we take to do what is happening here - which is to engage and build with the community. Our best option is to release Phase 01 as a demo/test for members of the community, to use it to generate interest and also to see if our game is even 'fun' for people to experience. 


risk of players losing interest in the game because they grow tired of play testing a game that is not fun or challenging enough


I agree with this, happened to me on most EA titles I own. They sit there, getting updates, but by the time of their release I simply don't care anymore. I'm open to discussing this one further.

risk of running out of development budget before the game reaches a state that is sufficiently playable


We hope that our skills in developing other projects will translate into this not happening. Hasn't happened to anything we've worked on yet and we've done a lot of internal calculations on what our milestones will take.

risk of multi player component never being completed


This is the primary risk. Seriously, this can happen to any game in development and we've seen sequels to AAA titles ditch multiplayer all together based on a range of problems.




Multi player first



shorter development time before game gets into the hands of players

Yes and no, for what we get back in no enemy God AI, we lose again in netcode and having to build the project in a specific way. But, we are looking to mitigate this as much as possible. The current gameplay loop and core mechanics can scale up for multiple Gods to exist in the same round. The design of the game is taking this into account, which is (for now) the best we can do until we know if the game is engaging or not.

less iterations between play testing and AI development

Less on AI testing, but more on human QA and testing sync/stability. AI development is sorted, sure, but we have to then worry about QA on 'are NPC villagers appearing on both players?' 'are values exchanged?' 'where is the data stored?' 'how many players/what ping/what conditions before drop?' etc. We can never escape QA/testing, we're going to eat that either way. 

game may start to generate income and publicity sooner


I agree with this, especially with multiplayer we could get exposure much sooner than expected. Things like Twitch etc.

game will be playable, even if the AI opponent is never completed


Yeah, but we'll get chewed out still on not providing a single-player experience, probably rightfully so. Some people don't want to have to play other humans, but I agree with this point - game would have a complete state per say.

potential for much better developed AI in shorter overall time frame
I still can't comment on this, I've been told that AI is a combination of the illusion of intelligence and the faking of player emulation. Basically, an AI can actually totally demolish a human player in most games, from RTS to FPS. Seriously, things like SorianAI in Supreme Commander or some of the twitch-shooter Quake bots are programmed to take full advantage of the ability for a computer to do everything it can to win. This includes cheating, being able to issue multiple commands on tick, awareness of player at all times etc. Some AI, even Chess, make mistakes on purpose to emulate human players. The quicker we can observe anyone playing any state of the game is quicker we can develop AI for it - including the villager AI. 

Remember, no matter what we do, we're going to have to tackle (and are now tackling) the villager AI. Towns will need to grow without you, that's a primary part of the game itself. That requires a lot of AI. 

Ba'al I'll try to poke them for more info and keep you up to date here! Thanks for that, a lot of food for thought for me to consider in the past couple of days!

What about programming it to not always do something beneficial at all? Humans have a tendency, whether intended or from naivety, to do things that have no noticeable benefit.

I honestly don't know much about programming.

Oh, and on topic; I see the logic in programming the more complex multiplayer functionality, prior to implementing single player. I would assume once that framework is laid, it would allow for greater flexibility in partitioning out the basic mechanics and features. The other way around, to my understanding, would require completely reworking the structure to support the new system, as well as having to layer in all the new requirements.

But I also have much to learn about game design.

I've always wondered can you base AI on environmental triggers, like time past + follower count + areas walked on=opponent behaviour trigger.  Surely you can make a better fake AI from the randomness of how you play combined with rules than to truly have an intelligent AI? I guess I should read up on AI!

You know, a heuristic AI (that's how environment-triggered AI is called) is actually kinda different to do because it is basically reflectory action simulated. And since reflexes can get complex and there are emergent actions formed from then the stuff requires good hardware and is fairly difficult to implement in the first place, not to mention the requirement for all the rules that it has to remember and resolve conflicts between.
In fact, creation of a strong AI is a creation of learning-capable heuristic AI and that's a Nobel Prize material that can result in world communism (as in, machines overtake all production and people no longer suffer from scarcity due to the economical efficiency skyrocketing and a huge part of populace dying off in the first decades).