Iron War - Official Thread



  • I still think it would be fun if there were a few different 'personalities' for the HardAI, named after different generals or whatever. Like maybe one guy has randomized purchasing for combat units, another has something similar for infrastructure type stuff, one makes more aggressive attacks prioritizing enemy TUV destroyed, one is ultra conservative prioritizing their own TUV on defense. Basically a dozen "spin off" versions of the same reasonably competant base HardAI, but just with one variable or two adjusted so they play differently. I think that would probably be more fun overall for players to have on hand than one AI that's meant to handle everything at the highest strategic level.

    Then the same AI game might play out very differently depending on which personality is assigned to the HardAI for a particular nation. I think some kind of compromise like that which uses a small degree of randomness might be more engaging in the near term for surprise outcomes, or humanlike digressions from the usual best practices, just to keep things interesting. By running a few different experimental AI drivers like that we might find some combos that outperform others, or at least mix up the bag a bit. I remember some older games like RiskII that had fairly weak baseline AIs, but which still managed to create a feel of dynamism and interest for the machine that way.

    If the behavior differences started out with something easier to control for, like purchasing preferences or attack thresholds, maybe it could be a thing that works for both Hard and Fast AI? Maybe it could eventually be randomized on a round by round basis, where the AI shifts back and forth over the course of the game, instead of just the initial assignment, but meantime it would at least give a place to start.

    In the same way that playing against a newb can sometimes be quite entertaining because they are more likely to go off script and force uncommon situations. Maybe something like that works for the AI too? I mean we don't have to set the bar all that high, since it's mainly for getting the rocks off while learning or in the downtime between PvP games. But it might also be useful from a game design standpoint, like a map being built around a certain standard behavior for a certain nation.

    Kind of off topic, but I still think the sweetest thing we could get going would be a way to flag a territory as higher priority based on some control maybe with an invisible marker. That way could be used for VCs or Capitals, high resources or Canals and such. Just like a generic tag that makes a territory higher up on the target to kill or defend to the death type tiles.



  • interesting idea actually but only if you found more people to work on the AI and really the Hard AI being the hardest AI is all I need ::)


  • Admin

    @Black_Elk @redrum @Captain-Crunch
    Maybe a quick and easy easy way to support maps with special units / strange purchases is a "chaotic/random AI" that worked like Hard AI, except that 50% to all purchases were random, and an xml setting that made the AI use this type of AI?



  • Yeah I mean that sounds pretty cool! I'm sure I'd have fun playing against it. The main point I'd drive home is that we are probably arriving at a point now where a single type of AI is insufficient for the number/variety of maps we have. The HardAI is pretty good at doing what is was designed for, namely general principles for the WWII maps that came out prior to V3, and it does a fairly competent job playing other similar maps. But I think we've probably hit a wall here, where it's easier to have an AI gang rather than one hard dude running the whole block. I think randomizing the behavior based on different variables would be really interesting, using different AIs with cool names.

    From a design standpoint, if we had a dozen different AI personalities, maybe people could build maps that play to certain strengths and weaknesses of that particular AI. Like just imagining extremes, maybe you have one AI who buys impulsively and spends half the pot on different units every time... well, at the other end of the spectrum you might have the dude who picks one or two favorite unit types and magnifies those buys every round. Or similarly maybe you have an AI who loves building ships and running the sea, another who's all for bombers, and just give them names or descriptions that fit.

    Those things are just on the purchasing side, but I can imagine a Dirty Dozen different AI's where each are basically a spin-off of the main HardAI. Then it would be a lot easier to design maps based on these different AIs instead of trying to have a single AI that can cover all situations across all maps. Otherwise you have that tension that Redrum was describing where a certain behavior makes a lot of sense for one map, or for one nation on a given map, but would nerf the AI in another possibly significant situation. If we had different AIs then we could play around with it a bit, under less pressure for ultimate perfection, while still operating within the same basic framework. At least that way we could experiment with some different ideas, without having to worry about unintentionally hosing the regular HardAI in the process of trying to refine it. For starters I would ditch the Easy AI altogether. It's name is deceptive anyway, since it should probably be called "old AI" at this point hehe. Or if its good to keep Easy AI for some kind of under-the-hood/future re-development reason, just change the name to something less enticing for the solo newb and stick it way at the bottom. Fast AI is good, because it's name tells you what it's really about.

    After that maybe just pick like half a dozen clever names and see if we can find a variable to tweak that would make their gameplay a little bit unique and let them loose on the solo playerbase to see whether it sticks...

    Perhaps you have one called "Ivan" who does what you'd want AI Russia to do, like just stack for defense and try to hold down a contiguous empire with very conservative purchases.

    Maybe "Tommy" is another AI personalty, except his deal is more about flying aid to his buddies, saving up to drop big fleets, and then establishing a beachhead near the front with transports.

    Then you got "Fritz" who likes to buy mobile ground units and hurl them headlong at the nearest enemy, sometimes makes desperate attacks when the odds are narrow, because he needs to maintain momentum and seize the initiative.

    Somewhere in there maybe you have an AI that focuses on developing production infrastructure or a wide variety of unit types. Probably analogous to what we'd want from the USA in a standard WWII type game. Once the basic template is established and people see how different AI types might fit together to make a more interesting game, then you could add to the list different behaviors that might be fun. I honestly have no idea, but it seems like that might be somewhat simpler than going at it Highlander style, where there can be only one hehe.

    In the Iron War map, I can imagine at least half a dozen different AI behaviors that might be advantageous to have kicking around, depending on which Nation is up in the rotation. In most TripleA games, there are at least a couple basic types of player nations, where a human is expected to do different things based on who they play, so probably the AI should mirror that. In that vein I'd say start with AA50 as a model. We do Six different AI's with cool names and behavior themed on the 6 player nations on that map. Use them as the experimental guinea pigs, to see if we can get them to model a more human like approach on that map (since its kind of like legacy edition A&A at this point.) If successful we adapt them or add new ones to handle other different maps.

    At least I think it could be a fun experiment to run. Maybe a little more entertaining for our AI guru, having something to mess around with that isn't quite as monolithic as the standard HardAI?

    ps. Still going vs the HardAI Axis with the boost. We took the TKO in 43, but Axis are still showing teeth and its pretty fun to try and manage the Allied fuel reserve while preparing for the final strikes. AI Japan has done a pretty reasonable job of reorienting onto the mainland and not letting their main fleets get cracked. Allies are trading W. Germany on income for the edge, but the Germans still have their main fleet and an expanding tank force holding onto Berlin. Balkans just made a sneaky Blitz on Moscow after AI Germany canopened the way. Guess they made Stalin pay for his overconfidence hehe. Looks like they got a few rounds left. Probably will play on just to see how long Japan can play off the fuel reserve they built up. Catch you next round!
    Axis 20 synth fuel per starting factory flat rate bonus Russia round 8.tsvg

    AI Axis made a pretty glorious final counter offensive to push back the Allied occupation of W. Germany. Balkans then blitzed to punish Paris hehe. They took a lot of TUV down, but the big German ground stack also went by the wayside, so basically a done deal in Western Europe. On the Pacific side Allies are steady rolling up Japan after a big build up over several rounds... Italy is stacked heavy in Egypt and Rome, but also getting boxed in now. Probably time to throw in the towel but I'll give it to the AI for making a clean last stand! hehe

    Axis 20 synth fuel per starting factory flat rate bonus Russia round 11.tsvg

    I think they could have gone on longer if the AI Japanese rebuilt their Tokyo factory, but they seemed to expend all their energy around China instead. Curious if the AI values territory income or just production value? Both AI Germany and Japan seem to underplay their highest value territories on defense/recovery, compared with the way they will target other lower value enemy territories in the area. Other than that I suppose the main thing that would probably make the AI more competitive is to prioritize VC control at the end of the round, for whatever threshold the Win is set. Like in this case, the AI just going all gangbusters trying to hold as many VCs as they can if they have the chance at 20. Right now the AI is still playing more total domination style ala Classic with solid general principles on the TUV trade, but not really set up for the VC aspect, even though that is how victory is determined for many games these days. Probably something like that would be at the head of the wishlist for me, making VC control supersede other considerations when Victory is reach. But anyhow, I like how the AI plays on this map when they have the synth fuel and a flat boost, fun stuff!



  • Here is a fun one where the AI Allies have been keeping us on our toes... Axis went with the slow push to expand carving up the conquests between the various minors along the way, but got stalled up in the Urals and the Russians had us fighting tooth and nail to take out Siberia. AI Soviets meanwhile have been launching some effective counter attacks using their bomber stack, with few furious bombing runs to sweep Japanese production on the mainland. Yakutsk is still kicking and British fighter support has kept the far east fairly active. US pacific fleet just did an end around towards Tokyo when the Japanese fleet finally launch from Truk , and AI Stalin has used the opportunity to re-established a final stronghold in Vladivostock! The Americans are stacking fighters to the ceiling! Germans just bought their first nuke to deal with the situation... hehe good times.

    Elk vs Allies 20 synth fuel per starting factory flat bonus Japan round 12.tsvg

    AI Australia felt better after the increase on starting fuel, they've been keeping Japan on edge bouncing defending fighters around. South Africa was knocked off and British-India is near collapse, but the rest of the gang is still up to no good in West Africa. Think it could go another half dozen rounds

    A pretty Epic battle vs AI Britain for control of Mexico City! hehe been taking this one full invasion USA, but the North American air fortress is pretty impressive. Japan didn't have enough strength at the ready after the long campaigns vs India and Russia to turn around and directly take on Australia or USA. So instead they blew past Hawaii and the West Coast and headed to Panama. Axis are sweeping across the South America now, with German tanks leading the way and a massive Italian army finally ready follow along for the Atlantic crossing. Even with tactical nuke strikes by both Germany and Japan over several rounds, the British and Americans are determined to bloody us still. US AI has fifty some odd fighters defending the homeland. Good times! This one might go into the 1960s lol

    Elk vs Allies 20 synth fuel per starting factory flat bonus Germany round 20.tsvg


Log in to reply
 

27338
1429
1658
Who's Online
Visitors Today