Iron War - Official Thread
-
@redrum What constitutes if a territory worth capturing for an AI? PU value, vc and capital?
-
@cernel Yeah FJ really could benefit from this. It a cool game just the AI's were notorious for messing the game up.
Although it's diplomacy is terrible too, but that can be player enforced a bit.
-
@cernel So "active neutrals" generally are like pirates/barbs/animals/etc and attacking them has very little value as every unit you lose is a unit you don't have to fight the real enemy. I'd rather start with an implementation where they are only attacked if its worth taking "X" territory to start and see how various maps that have active neutrals play (this is also mostly a minority of maps as most have passive neutrals). Adjusting the 'thresholds' tends to be must easier after we at least have players classified properly.
@General_Zod There are a bunch of things but the most important right now are PU value, has factory, and capitals (primarily because on simple maps like revised those are the main factors). There needs more consideration for things like VC, NOs, other non-PU resources, canals, etc. There is also a sense of strategic value so that AI will attack territories that are on the path to other objectives even if they have no value.
-
@redrum said in Iron War - Official Thread:
@cernel So "active neutrals" generally are like pirates/barbs/animals/etc and attacking them has very little value as every unit you lose is a unit you don't have to fight the real enemy.
Personally, if I would make such maps, I would not hide any "active neutral", but just the passive ones; so that, then, would be pointless. I'm thinking it is good to leave the player the possibility of selecting the AI they want, or whatever.
As I said, I'm positive on the "passive neutral" thing related simply to not having combat move, but, on the other hand, the "active neutral" package feels too complex to track and not very reliable, especially for mapmakers that don't know about it (hiding or not is really merely preferential).
This is probably something that should rather stay in an "AI.properties" configuration for the map, where a mapmaker could define what are the "non competitive" players, maybe defaulting to all players starting with "Neutral" or part of an alliance starting with "Neutral", if not set. -
Improving neutral categorization discussion moved here: https://forums.triplea-game.org/topic/776/ai-improve-neutral-categorization
-
@Black_Elk said:
enjoying the HardAI. The only downside compared with FastAI is that my laptop gets kinda fried if I let it go for too long. Like I'll get a map flickering and delay on actions if I go much longer than a full round.
Yes, this is my experience also. I hope that the hard AI can be developed in a direction that speeds up the calculations, so that also large maps can make use of the better Hard AI and not mostly rely on the Fast AI. And I am sure redrum is working hard on this
@Black_Elk said:
I think that this map has a lot of appeal for the single player vs the AI, its more interesting I think than Global or the smaller scale A&A maps, just because it has more going on to keep the single player engaged with it. I've probably played close to a hundred games against the machine, so clearly there is something here that keeps drawing me in.
Thx. It is also the idea that Iron War should be single player and AI friendly. I also like to play single player and coop TripleA, so this also has an effect on my map designs. But surely Iron War would not be developed to what it is now without your intense interest and testing!
@redrum said:
the AI attacks neutrals like crazy! This is because it sees them as another enemy player rather than neutrals.
The âneutralsâ in Iron War, as you see, are all passive and have no CM step. They are also not using their PUs. It is intentional that neutrals can be attacked, and the map would not really work if it was not possible for humans and AI to attack the neutral powers. Territories like Greenland, Iceland, Denmark, Congo, Tibet, Morocco etc. are practically meant to be invaded during play. Some of the territories are 5 PU territories, which means that a factory can be placed there.
I have also noticed that the AI does attack the neutrals even when there is no logical reason to do so, probably because of the TUV destruction that the AI prioritizes. It seems that the AI really likes to attack Ireland, Spain, Turkey etc., even though there are few PUs gained compared to the potential PU cost. The worst example in Iron War is Brazil who, instead of sending forces to Africa/Europe, uses most units in an effort to win all of south America.
In some places, like Ireland and Spain, I have placed un-proportionally large forces just to discourage the AI from attacking, but the AI seems to attacks anyway when it has enough forces. I would like to remove some of those units if the AI got smarter in this aspect.
It would be good for Iron War if the AI would only attack passive neutral territories very lightly defended, very PU productive/valuable, controlling a canal or stuff, even though the territories and units were not a threat. It should not attack just to kill if the enemy did not represent a threat, but it should also not NEVER attack as this would be a little dull. Redrum, maybe you could implement a very small probability of the AI doing unexpected stuff / going for low value territories. It would be nice if the AI could occasionally surprise players, of course only when the AI has the strength and does not compromise itself
PS: I think the AA gun cost will be increased from an 8 PU to a 9 PU cost. Also I will probably add a few barrels more to a few nations and territories.
PPS: I am currently mostly working on converting and balancing Iron War Europe to the new fuel system, so this takes some of my time. -
Neutral changes are merged as well as a few other tweaks and here is the latest pre-release: https://github.com/triplea-game/triplea/releases/tag/1.9.0.0.10182. The AI has some pretty drastic changes to Iron War with those changes. Tends to attack neutrals much less frequently and generally have better attack priorities.
@Frostion Yeah, the AI will still attack neutrals mostly just to grab their territory value but will generally prioritize them lower than capturing enemy territories. A good example is Baltic States which the German AI seemed to almost always attack round 1 where it should really prioritize Bylo or Western Ukraine and now pretty consistently does. You should be able to decrease mass neutral stacks now and the AI should only attack when it makes sense to gain territory production. I'd doubt you see the German AI attack Spain much even with half the units it has now. And yeah AA guns being 9 PU is at least getting close to them being balanced as at 7 PU they were crazy OP and 8 PU they are still fairly decent to mass.
-
Here is a 17 round all Hard AI game where the Axis appear to finally have it in the bag: 0_1524373939603_IW_HardAI_Axis.tsvg
-
Had a pretty fun couple games using that last pre-release. Whatever tweak was made I'm able to go like 3 rounds at a go pretty consistently now before the HardAI slows down. Its a definite improvement for playability.
In the first game I gave the Axis 110%, and then 130% in the second game.
0_1525057122685_Elk vs Hard AI Axis 110 bonus Russia round 13.tsvg
0_1525057139803_Elk vs Hard AI Axis 130 bonus USA round 13.tsvg
Both times my goal was to see how quickly the Allies could take the technical win at 20 by just throwing everything at the weakest VCs right away, then seeing how the games shaped up... like whether it was actually worth it to go all gangbusters right out the gate, or if bleeding the troops that early would come back to haunt me haha.
Of the minor Axis powers Iran is the most straight forward to pick off, since Russia can usually storm in out of Siberia and the Caucasus/Caspian region on their second turn. But after that its a bit of a trade off, which I like. Pushing further to take Iraq or Finland, almost guarantees that Germany is going to break through and shut down a couple Soviet factories with their mobile units. Still not sure whether its better to lose the north or the south, but either way things get hectic, which is fun to try and manage.
On the Pacific side, knocking off Thailand early on is possible, even to kill them off in the first round before they get a chance to place. But this is a little sketchy since you need to get the expected hits with British-India, French-Colonies and KNIL all right in a row. And set up blocking ships to prevent Japan from just putting everything into IndoChina on non com. Even then its kinda debatable whether killing Thailand immediately is worth it, since it takes like everything in the region to pull off. I tried it in the first game, even smoking the Japanese transports afterwards, but Japan just came hella heavy against China. Like it focused them in a more damaging direction than just stacking in south east asia would have produced lol. I think I may have really pissed off the AI too, since it was the first time I'd ever seen it run a pearl attack on J1!
The German HardAI seems much improved. Not sure if its the tweaked neutral thing or what's going on, but they seem to attack the British fleet now, which is much better for team Axis. In both games they came at me with all the ships and the air transport, something like 80% to the attacker with a couple German ships surviving on average. They seem to use the bomber and fighters for other battles, so not quite a full sweep like it might be, but certainly makes things more complicated for the Allies.
I was also pleased to note that Italy managed to snag Gibraltar in that first game, which is something I don't see happen very often. Could be the neutral thing again where the HardAI is now seeing those Brits as the priority threat over the neutrals in Turkey or Greece (which is where they used to go first.) Overall I'd say the play pattern for the AI Axis feels a bit more historical now.
Also saw the machine incorporating a lot more AAguns into their builds, which makes the air war a bit more challenging. I find myself buying more AAguns now to counter, since otherwise it can be tough to maintain hitpoint parity. All in all I think the AAgun unit feels a lot more relevant now. It might be a little overpowered for the cost, but I'm just happy to see its not a completely pointless purchase in this game, like it is in traditional A&A hehe. It actually works pretty well as a spam fodder unit, if only because its somewhat easier to clear out if you buy tanks than a similarly sized stack of infantry would be. On the whole they seem to encourage a few more tank builds over air to counter, which is sort of the purpose, so I like the balance.
I'd say most of the minor Allies feel pretty engaging now, although Brazil is still kind of one dimensional. Its basically a choice between feeding a hitpoint or two into Africa every couple rounds, or going after the neutral neighbors in South America. Of the two the latter option seems to have more of an impact since you can at least deny the Axis team a few resource points. But unlike the other minors its hard to achieve any kind of break out, and you don't really have to worry about managing a defense, so its sort of like just going through the motions. Most of the other minors can adopt a clear support role somewhere, or can become more relevant if one pf the big dogs throws some cash their direction, but I haven't really figured out what to do with Brazil hehe.
I still think it would be nice if the minors could send Aid to the majors. Often the minors will have a remainder under 8 PUs that can't be spent (for the AAgun hitpoint) so it would be nice if they could send like 5 PUs to their associated major power. I think it would clean things up, allowing aid to move in both directions, with less cash leftover at the end of the round. That's my main suggestion for now, an option to send small amounts of cash to the big dogs, rather than turning the minors into mini-super powers and checker-boarding the map, which is sort of what I do right now.
Anyhow, fun stuff. I like the latest changes! Nice work guys!
-
Ps. I also tried a German solo vs the HardAI Allies at 150%. It was pretty brutal. The HardAI just stacked massive. Figured I'd hunker down and see if we could make it till 1945 hehe.
Here it is at the dawn of the Nuclear era, G is still kicking but my AI teammates are definitely up against the ropes. Italy has been doing reasonably well considering how the minor Allies have been banging away at Africa, but Japan got gutted pretty early on.
It seems the HardAI Japanese don't fully appreciate the centrality of their transports to victory, and will needlessly throw them away or leave them exposed, just to drop off a couple infantry units here and there. HardAI Japan also seem reluctant to replace their lost transports, which seems a bit curious since most of the other naval powers will buy transports pretty regularly. The Japan Sea Zone alone could cover territories worth 20+ production, but once the transports were gone they got stuck unable to defend the coast very effectively.
Anyhow, I think 150% is pretty much the upper limit before the continuing bonus makes the game so tough that you have to start switching the goalposts. But it was fun for test run.
0_1525073369930_Elk Germany vs Hard AI Allies 150 bonus round 12.tsvg
-
Thanks for the feedback @Black_Elk. I am taking it all into account. I am also myself doing a lot of testing, evaluation and start setup changes because of the new AI and fuel balancing. I hope to present a new version soon, but time is not something I have in abundance right now
But soon hopefully! -
Sounds good, will look forward to checking out the tweaks. Lately I've been thinking more about how to approach the AI difficulty levels idea floated earlier.
On the one hand I like the idea of something that is consistent/universal for all the AI nations on the enemy team, so that its easy to understand what's going on. But on the other hand, I think there may be some variation in what's needed depending on how hard we really want it to be. While still providing for some nuance, it'd be nice if we could describe what's happening on the machine's side in just a couple lines. The wording can be a little awkward, so it'd be cool if we could come up with clever names for the difficulty levels, but just for a starting point, we might have something like this..
Easy:
Standard game, no AI advantages.Normal:
The enemy AI team is not subject to fuel-based movement restrictions. +500 bonus to starting fuel for each enemy nation.(Lets just say whatever fuel number is required so the AI can go a dozen rounds or so without running dry. This alone should improve the challenge considerably, since it would allow the computer to move all its ships and aircraft each round without getting stuck.)Experienced:
+500 starting fuel, +25 PU's to starting income, +5 PUs per round recurring income bonus.Hard:
+500 starting fuel, +25 PU's to starting income, +5 starting steel, +10 PU's recurring income bonus.Very Hard:
+500 starting fuel, +50 PU's to starting income, +5 starting steel, and 115% recurring income/resource bonus.Extreme:
+500 starting fuel, +75 PUs to starting income, +10 steel, and 125% recurring income/resource bonus.Those are just ideas, but basically something to provide a bit more interest and variation between difficulty tiers than just a straight repeating income bonus would offer. I think an increase to starting income/resources, coupled with a recurring bonus per round would be ideal, since it would force a different play pattern out of the first round each time you go up a difficulty level.
Or another approach might be to only alter the starting resources at the lower levels, and then introduce the recurring bonus as you go up in difficulty. I think the recurring stuff provides a more consistent challenge, but there is also something to be said about front loading things, where it's like, if you can just get over the initial hump, then you can make a break away. With recurring bonuses, things get harder the longer the game goes, so there is an incentive to be ballsy about it early on.
-
I have released a new version of Iron War. The changes are as described here:
World v0.2.3 to v 0.2.4
⢠The Anti-Air unit now costs 9 PUs, not 8.
⢠Added some more Fuel barrels on the map.
⢠Romania is now a 20 PU territory, not 10 PU.
⢠Italy is now a 40 PU territory, not 50 PU.
⢠Siberia is now a 10 PU territory, not 5 PU.
⢠Brazil is now a 10 PU territory, not 5 PU.
⢠Start setup changes.Many of the changes have been done to adjust the fuel availability and balance the fuel needs of nations. Also, as the map tries to implement a bit of historical realism (alongside the playersâ freedom to act crazy) some start setup adjustments have been made as a result of the newly updated AIâs behavior.
Brazil and Balkan have been boosted a bit in PUs, so that they are maybe more interesting to play if players take control of these nations. Hard AI Brazil still goes like 75% of the time on a mission to conquer all of South America, but at least some times they do help out in Africa and Europe. Since the neutral South American countries are mostly 1 PU territories, I think most human players would not waste 10-20 PU units on conquering this area, even though it is a strategic option for human players.
As part of the starting setup rearrangement, the UK fleet now starts 1 sea territory further away from the German, so that the Germans do not always wipe out the UK fleet in round 1, but instead focus on conquering some land territories. With the new setup, the German AI seems have a variety of move options to chose from in round 1. And many of the Allied countries use round 1 to move units around and positioning them according to Germanyâs actions and battle luck.
@redrum As @Black_Elk I have also seen the AI, especially Japan and Germany, move transport ships into positions where they are sure to be killed. Many times the AI moves a transport, disembark the units as part of an attack, and then the transport is left undefended at the coast and killed soon after. I think the AI needs to protect sea transports a bit better, maybe by moving escorting defensive ships with them?
@Black_Elk I would also like to set some levels of difficulty some day, and your play experience is the best measurements we have to date. You have played a lot with increased % income, and this is what I would like to use as ârecommendationsâ. Maybe at some point the engine will support more varied options like AI personalities and stuff, and this would be great to implement as part of any recommended or pre-fixed setup options. Right now I have not implemented recommendations in the notes or coded some sort of map/xml trigger system for difficulty settings.
@All I really hope to use some time updating Iron War Europe now, so that the Europe version does not fall behind
I hope to do some intensive testing now on Europe. But please keep playing the World version as that version is probably best balanced and most playable atm. -
@frostion I like the changes. If you have some saves where the AI consistently leaves transports vulnerable please upload a save game so I can take a look. Generally, transports are one of the most difficult parts of the AI to do well but always looking for ways to improve it.
-
@Frostion Here is a 27 round AI vs AI game after land transports were just added: 0_1527297105272_IronWarAI.tsvg
-
@redrum Thx. I will take a look at it when I have the time. I have not had that much time lately to look at anything TripleA
So I have not been able to look at and make save games of the AI sacrifice transports thing.The bad management of transports, where the AI combat moves transports alone and do not follow up with escorting ships, is a problem not to be confused with situations where the AI actually wants to move combat ships with the transports but then does not have the fuel. This also happens in Iron War. Maybe you have noticed.
-
PR with AI land transport implementation is merged and can be tested here: https://github.com/triplea-game/triplea/releases/tag/1.9.0.0.10276
-
@Schulz bumped
-
Been a little bit, busy at work, so haven't had as much time to game lately, but here are a few quick saves using 1.9.0.0.10355.
I started out with some Solo games to see how the AI held up. I didn't give the AI any kind of bonus, because I wanted to see how the baseline challenge felt after that land transport update, since it seemed like a major game changer.
For the first few rounds I'd say the AI does a pretty reasonable job of moving their mobile ground forces, though they do burn through their starting reserves of fuel pretty quickly.
Not sure if the AI has any way to prioritize which fuel units move first, like if it calculates that it should move mobile ground or ships before the more fuel intensive air units, things like that. But I think for the AI to provide any kind of serious challenge into the endgame, it needs a major fuel bonus.
Having the mech units finally functioning properly definitely helps the AI balance with Russia and Germany, since mech is such a large part of their starting forces.
I will probably play some full games next for each side to see how it feels with no bonus, then gradually increase to see how it compares to earlier versions. My guess is that it will probably still hover around 20%, though I think that could probably be reduced if it was possible to add just fuel without increasing the PUs or other resources.
0_1530315932628_elk vs hardAI Allies Germany Solo round 8.tsvg
0_1530315942437_elk vs hardAI Axis Britain Solo round 13.tsvg
-
@black_elk The AI doesn't take fuel into account when planning purchases or moves. Only thing it does is make sure it has enough fuel before making a move. But yeah having a separate fuel bonus setting would especially be useful for the AI.
Glad to see the land transports are a significant improvement.
Hello! It looks like you're interested in this conversation, but you don't have an account yet.
Getting fed up of having to scroll through the same posts each visit? When you register for an account, you'll always come back to exactly where you were before, and choose to be notified of new replies (either via email, or push notification). You'll also be able to save bookmarks and upvote posts to show your appreciation to other community members.
With your input, this post could be even better đ
Register Login