Age of Wonders 3


We have a term for that.


Then there are those of us who stay on normal and still find most AI’s fairly challenging. HoMM3 is a good example, I have always gotten my ass kicked on that pretty regularly because I get distracted with exploring/city building while the AI is somehow putting together a huge army.

Looking forward to AoW3.


HoMM games aren’t a 4x game, at least for me they aren’t. You don’t develop an economy (just take sites), do diplomacy, trade, manage taxes, invest on research, etc. It’s more about recruiting armies and expanding. It isn’t the same developing a good AI for Civilization than for HoMM.

AoW is another case. While the game in general isn’t as complex as other 4x games, there are some complexity in race relations and group compositions, lots of special stats and abilities in the units, and lots of possible counters and special tactics that are hard for the AI to grasp.

In HoMM3 all the units are combat units, leaded by a hero. In AoW, you can use the units by themselves, and make raids with elves on forests, use invisible units to attack without retaliation, steal mines with flying units, use dwarfs to catpure areas and fall back to mountains where other units can’t go, etc.


Ah, but the drop in the Parabola as stated happens right after you’ve mastered the tools of the system and realize that you have nowhere to go, because the game doesn’t actually know how to use them. That would be a proper description of my experience with HOMM IV-V. In competently designed strategy games, mastering the system is where the fun starts. The drop only happens much later after you’ve proved your mettle through countless victories and defeats (and then, your enjoyment may not drop at all, and hit a comfortable plateau instead).


Fair enough on the difference between a true 4X such as AoW and a more simplified one such as HoMM3.

As with most things I guess for me AI talk is very much a YMMV situation. I am looking for a different experience in a 4X than I am in say Football Manager where I want that AI to be as sharp as possible in order to simulate the challenges faced in real life. When I play a 4X my goals are more to experience that imagined game world than to min/max my way to an optimum strategy, but I recognize I’m likely in a minority there.


I am trying to remember, but I think the AI issue in AoW SM was that it was a confused homebody. It could build units and spend resources, but it had a tendency to get stuck sitting at home. Every now and then there was the game where whatever paralyzingly variables didn’t occur and it would shockingly sally forth with tragic results to me, but most of the time it would do sort of an Xcom shuffle on home tiles or between only a few cities/ resource hubs. Such an “opponent” generally left the game unplayable outside of map design/ player funneling based solitaire (ie campaign) so the Chick Parabola was in full effect for even commanders in the extreme dilettante category.

I am pretty sure that was the hiccup on that one, but it has been many years. HoMM IV was similar but constantly suicidal to its 4 goblins just outside the castle and completely inept in tactical engagements.

Edit: As to the YMMV thing, what I am worried about and focusing on is not AI that is challenging to a player of a certain skill so much as an AI that can function and use systems within its own game to at least a rudimentary level. So low a bar seems increasingly entirely unreachable for some reason. If all an AI can do is run free units in a circle, then it doesn’t matter what skill level you brought to the game; there is no game.


I think my biggest issue is when the AI doesn’t play the same game that I do to the point that logical strategies (based on what would hamper/defeat me easily) are ineffectual against the AI. The biggest example is in terms of resources/supplies. Shutting down a player’s access to key resource nodes/sources to cripple their economy is an age-old strategy, and is a huge part of the old “Rush vs. Eco” interplay common in competitive RTS games like Starcraft.

If the AI’s got infinite funding, or can acquire resources in ways that the player cannot, then the “obvious” strategy of bleeding the AI dry in a war of attrition suddenly goes out the window.

Similarly, destroying troop-producing structures/cities in a game where the AI gets units for free.

I like to see the AI also being taught to “see” the game in realistically pragmatic turns. When it’s not, even trading can be bizarre sometimes. I remember a game (was either Civ V or GalCiv 2, the last two 4Xs I’ve sunk 100+ hours into) where enemy AI was unrealistically reluctant to trade tech, to the point where even ludicrously lopsided trades that a logical human player would be able to Okay in a heartbeat would get denied by the AI due to it vastly overvaluing tech.

Anyway, if the AI isn’t playing the same game that I am with a reasonably similar set of goals to me (I don’t mind the idea of variable AIs, where some opponents are overly aggressive, others are researchy turtlers, etc.), then–as noted above by others in the thread–once I start to really understand the game’s systems and how they interplay, the fun factor immediately drops off enormously.


The AoW AI problems ever since the first installment have always been related to leaders. In AoW1, if you tried to play the same map with leaders on and leaders off, the experience was vastly different.

Leaders on, the AI was far more timid and always kept huge hordes of units around its leader in the capital, but did not know how to expand properly unless it had massive advantages.

Leaders off and the AI was far more aggressive and you couldn’t kill it by cutting off the head alone (which allowed for sneak victories while losing the war when leaders were on). So it came at you much more fiercely. Especially on some maps which were constructed to have good AI opponents, it was brutal.

In AoW2 and Shadow Magic, it was by definition impossible to play with leaders off, which meant the AI fell into the same trap as with AoW1, but also had to deal with far more complex empire management. The economic system of the sequels also rendered mana nodes irrelevant because you got far more mana from your city buildings and by founding new cities and building them up. So no wonder the AI was confused, especially after you mauled a few of its armies and took some of its cities.


I loved AoW games, but honestly, the AI didn’t understand half of the abilities coded into the game.


I do not think I agree that there is no game at all if there is no super AI because all the beloved 4X games of yesteryear almost all had bad AI.

I think that was the point of the post above where several posters stated the games attraction was not only the challenge of beating the game but also discovering things in the game. My point being that if bad AI means no game than a lot of the craving for old time games would not exist. How does one explain the popularity of MOM except that people were drawn to other aspects if the game that made up for the lack of strong AI. Maybe there is an age gap here with older folks just happy being to have something, but as strategy gaming evolves the acceptance of weak AI diminishes. I also suspect that multiplayer growing has a lot to do with wanting a stronger AI SP experience in a great portion of the gaming population.

I agree that strong AI should be a part of any strategy/simulation but not going to happen unless there is strong demand. Civ 5 and the New Sim City come to mind.

I think this is the biggest problem in our support for strategy 4x games on crowd funding. Here is a chance to influence the outcome more but that does not appear to be happening.


On the topic of AI, how far are we from the development of an AI that “learns” a game via repetition and derives a statistics based gameplay from it? Or if not via this method then another, and so we have a botAI that can play different 4X’s?

It will happen in most of our lifetimes, the question is just when.


Stardock is going to try something along those lines in GC3, where the AI learns from games from the Steam Cloud, but not fully in that direction like I had hoped.


Supreme Commander 2 had some pretty good AI that did something along those lines, although not in real time. The AI would run thousands of games against itself, discovering what strategies and tactics did and didn’t work. This would then be touched up by the AI developer before a patch and the update rolled out.


For those like myself who enjoy the endless replayability of a well crafted random map generated (ala Civ, for instance) this blogs for you.


Awesome read!


Yes, the RMG in AoW3 is very nice indeed. Haven’t had a chance to run it with the last build to see what has changed. I do remember that one of the first things I did upon entering the beta test was find two or three ways to break the RMG in the then current build. Note that almost everything in that dev journal entry is also new info for those of us in the beta.


Could someone explain “game flow”? I don’t really get it from the journal entirely, and it’s piqued my curiosity.


Surely, it will be something like slow/normal/fast, affecting the magic research speed and maybe the strength of indie units on the map.


That was my first thought as well, upon seeing that screen shot, but this line sort of made me wonder:

Could still mean game pace (research speed, starting resources, who knows) or it could just be referring to the ascetic flow of the map, too (grassland into winter tilesets into lava tiles, etc.). I assume you are correct though, the more I think about it.


The game flow option is not explained at all in the entry. I asked for a clarification on that because the NDA ties my hands. It is, when you do get your hands on the game, rather self-evident, so from a mechanics standpoint the rest of the post is far more interesting. Especially since it is previouslu unavailable in-house developer info.