Iron War - Official Thread

[Black_Elk]

Another rematch, this time at 140%. I tried something different and went after Australia instead of India, to see how things would shake down. It was possible to pull off, but I had to send basically everything against Sydney ASAP, even building an extra transport at Truk to put as much heat on the Aussies as possible. It felt like kind of an inversion of the usual pattern. Instead of taking the center and then working outward, its was like a big push to the periphery then rushing back to the center before everything went to hell hehe. Was a little surprised to see that it actually worked pretty well. Thailand got dropped by the AI Dutch almost immediately, but somehow it felt like that worked to Japan's advantage since it gave a clean shot on those pacific KNIL transports, so even if Sumatra was stacked heavy, we knew those troops wouldn't be all hopping about in the backfield. India was pretty sav though. They're still alive even after a dozen rounds, pretty much the last man standing for the Allies at the center. Russia was pretty strong too. The AI Americans managed to take sneaky snake Norway for a time and sent some fighter support to Stalin in the process, which helped prop up the Soviets. They knocked off my black sea fleet from the air, which stalled things up and almost lost me the Middle East, before Italy came to the rescue. Played a stronger game in Africa than usual, and tried to come at British Colonies and France pretty hard, to make up for the fact that British India would be getting less pressure. Was a bit more effective than I thought it'd be. AI France seemed to freeze up and let Italy crack into Nigeria while the Germans came at Ivory Coast from the west, which was pretty sweet for Il Duce, allowing the Italians to go major.

One thing I've noticed is that a bunch of British Fighters seem to get stuck in Victoria. I'm guessing that the AI Brits move all their ships and then run out of fuel before they can get to the Canadian fighters. Not sure. I guess the same thing happens in E. USA, though they seem to do a better job of getting their fighters into the action. Most of the rest of the AI nations seemed to do pretty well the fuel bonus at 140%, Russia is especially formidable once the bonus starts getting higher. They're still pretty deep coming out of Siberia and Irkutsk despite finally folding along the front with Germany.

But we just showboated into Iceland with the Finns, so that's gotta give the AI Allies pause hehe. Anyhow, pretty fun stuff. I think I'll have to start playing as the other side next to see how they do.

0_1523859789898_Elk vs Hard AI Allies 140 bonus Finland round 12.tsvg

[Black_Elk]

First full game against the HardAI Axis. Started them at 110% with plans to work my way up. Before using a standard strategy I wanted to try the bomber thing again to make sure my initial assesment was accurate. I still think the bomber spam is the optimal strategy for the Allies and probably too overpowered. Its hard to say whether a more intelligent opponent might be able to overcome the bombers by purchasing fighters to scramble, but even then the minor Axis powers are particularly vulnerable owing to their limited production.

I see two possible solutions, either limit the unit roster of the minor Allies so that bombers don't become available until later in the game, or else give all the Minor Axis powers access to bombers as well, so they can at least compete with a similar strat bombing campaign in response. Of the two, I think the former option probably makes more sense.

France, South Africa, French Colonies, KNIL, and China could probably all do without bombers as a purchase option, or at least have their availability delayed until the endgame.

I think I'd keep the bombers for British Colonies, British India and Anzac, so their unit roster matches that of the British, but nixing bombers for all the other minors would probably go a long way towards preventing a merciless spam.

In this one by the 4th round the Allies are pretty much set... ready to light it up, and knock the Axis out of the production game entirely. Next time I'll give the Axis more to work with, and won't be quite so ruthless with the Strat Bs, but just wanted to show again how nuts it can get if all the Allies go hog wild with bomber purchases right out the gate hehe...

0_1523988044955_Elk vs Hard AI Axis 110 bonus Russia round 4.tsvg

[redrum]

@black_elk Yeah, only way to determine SBR balance is in a PvP game as the AI is helpless vs SBR on this map.

[Black_Elk]

@redrum

Yeah my gut tells me that in a PvP game, scrambling will not be enough to deter this sort of strat. The minor Axis powers don't really have much cash to put into fighter interceptors. I guess if the big 3 on the Axis team each sent like 2-3 fighters to prop up the minor powers it might be enough of a deterent, but then that's a lot of cash/fuel to keep the bombers away. Creating a halo of AAguns or cruisers might be another approach, but pretty expensive. The bomber only costs 30 per unit for the Allies, but probably costs almost twice as much for the Axis to set up proper defenses against it. They are also extremely effective as regular combat units to pick off transports or weak links on the ground, so for the Allies they are kind of steal, especially for the far flung minors that can really use the movement advantage.

I think the most sensible approach would be to delay the arrival of the bombers by making them a tech advance for most of those minor Allies. Even 1-2 rounds to get prepared would make a big difference. My feeling is that 1941-42 would be ideal. Giving a chance for the Axis to get their act together before they start getting squeezed from all directions hehe. At least you could build up AAguns over a couple rounds.

Right now on the Allied team everyone but Brazil has access to bombers from the outset. On the Axis side Germany, Italy, Japan and Thailand can buy them (Finland too if G sends cash), but Iraq and Iran don't get them at all. I wouldn't anticipate that everyone would go for the bomber game. Some just don't like the unpredictability of SBR, but right now it seems kind of one sided, with the Allies gaining a huge turn order advantage in the sequence if they start early and stick with it.

Cutting out bombers for France, South Africa, French Colonies, KNIL and China would at least even things out, so a dedicated bombing campaign by the Allies would be somewhat more manageable. Again the most vulnerable targets are the smaller nations with just one production spot and limited income. For the Axis that's half the team, Balkans, Finland, Iraq, Iran, and Thailand. On the Allied side the most vulnerable to bombing is British India, followed by South Africa. France and China are probably hard pressed too once they are knocked down to just one factory. Although Urimichi and Gabon are easier to defend by spreading out with aa guns along the likely flight paths. The other Allied nations have safe spots over in the Americas so probably can come up with alternative builds. I haven't attempted a carrier based bombing campaign, but that might be an interesting option for Japan vs Australia or the US.

On the Allied side defending against Axis raids against thr subcontinent, the cruiser in the India coast sz is probably key to detering bombers against the India/French India factories. With a cruiser and a couple AA guns fanning out, you can at least force the Axis to run a gauntlet.

But yeah, that's all PvP. Against the Machine the bombers rule, so probably just have to go easy on the AI and use them only for combat hehe.

One thing I do really enjoy is how the strat bombing game feels more nuanced and consequential than it does in traditional A&A. Something highly satisfying about knocking a factory off the board from the sky, even if they are relatively inexpensive to replace, the shock of a well timed placement phase vs the enemy stall can be decisive.

[Black_Elk]

This one was pretty fun. After what felt like a solid performance by my Allies (especially in Africa and India), AI Japan made a sudden and dramatic comeback at the end there. In rapid succession they smoked the American fleet, knocked off China, and held the Russian's feet the fire in the Far East, setting up a recovery for their team. I gave the AI a 120% boost this time and made no bombing runs. As if to punk me for my charity, the Machine started doing bombing runs of their own! Out for blood tonight.

The AI Japanese really tried to ice it too, spanking me in Philippines after MacArthur thought he had it in the bag, then stealing Truk and Hawaii before our yankee jaws had time to hit the floor heheh.

It's definitely entertaining when the AI catches me off guard and doesn't pull any punches. Germany/Italy also did some nice coordination with a little Denmark trap to isolate the British fleet once I moved it through the strait to help stack Baltic states, alas those Brits are cut off. Now it looks like the European Axis are trying to blow the lid off Moscow pretty soon here. Throwing in the towel and calling this one for the computer haha. Nice job HardAI!

0_1524113178159_Elk vs Hard AI Axis 120 bonus USA round 7.tsvg

Will be back for the rematch lol

[redrum]

@black_elk Out of curiosity, what engine version have you been using? The latest stable (9687)? Or a pre-release? As there have been a few fixes to the AI in the pre-releases.

[Black_Elk]

I'll have to check for the latest when I get home. I was using 1.9.0.0.9853, but that was from a few weeks back now I think.

I've been 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. If I see the AI stall on noncom or something, that is usually my cue to quit tripleA and reload the game, which generally speeds things up. Saving at each block is helpful to keep the pace up. Not really sure if there's more I could do to help the machine crunch its numbers. I figure it has a shit ton to think about each turn, given the scale of the map, but I've definitely noticed stronger play from the Hard than the Fast, so seems worth the extra effort hehe.

Anyhow, will give the AI Axis another try later tonight. Thinking to go 125% this time, just to see what kind of damage the Japanese can do.

[redrum]

@black_elk Alright, pretty recent then. There has been a few small adjustments since that version and the default memory has been increased from 1 GB to 2 GB which could have an impact on AI speed as well. Be interested to see if that memory increase makes any difference for you.

EDIT: I loaded up your last save game you posted and the memory should make a big difference actually. Trying to play Iron War with only 1 GB even without the AI is probably slow. As the map itself ideally needs around 1.5 GB to fully load everything up in memory so you probably had hard disk swapping the entire time.

[redrum]

Decided to run an all Hard AI game and let it run til round 7 when Germany begins to overrun Russia and appears the Axis should have this one in the bag: 0_1524161355825_Iron_War_Hard_AI.tsvg

[redrum]

@Frostion @Black_Elk So the first thing I noticed is the AI doesn't properly consider AA when purchasing as they are a valid land unit for attacking/defending. So I went ahead and fixed it: https://github.com/triplea-game/triplea/pull/3384

You'll most likely see the AI buy more AA guns now primarily for fodder. It seems to go for a blend of AA guns and infantry now given the costs which is fairly reasonable.

[redrum]

@Frostion Next thing I notice is the AI attacks neutrals like crazy! This is because it sees them as another enemy player rather than neutrals. Do you think there is an easy way to determine which are 'neutral' players by looking at certain parameters we already have? Or does it make sense to add an attribute to player?

[Captain Crunch]

@redrum hey did you just adjust the AI purchases (AA and Infantry) for the overall AI or just for the Iron War map??

I havent tested the latest AI in awhile since you have not been doing tactical improvements but your last post maybe you altered the AI than when I last played it ... was that a tactical AI update??

[Hepps]

@redrum Isn't the neutral parameters already hard coded? I would have thought there are already things in the code that would allow the AI to omit them as a threat.

[redrum]

@hepps I mean when you have an actual player that acts as neutral. Open up Iron War and you'll see what I mean

[wc_sumpton]

@redrum
I think in Iron War there really is no 'Neutral'. 'True Neutral' is assigned the relationship 'Closed-Borders' which has the 'archType' of 'War'.
I believe this is done so that the AI can attack 'Neutral' territories.

Cheers...

[Hepps]

@redrum You are nothing if not always correct. I should know better than to stick my thick neck out... lest it be hacked off with the sword of enlightenment.

[redrum]

@wc_sumpton Correct. But what I'm getting at is that the AI should treat them essentially as 'neutral' when determining attack/defense. As those neutrals can't attack and its generally better to only attack neutrals if the territory value is high or its a strategic position. The AI shouldn't attack 'neutrals' for positive TUV trades. This is what it is currently doing in Iron War.

[redrum]

So my thought is one or both of these player parameters determine if a nation should be treated like 'neutral' and not have the AI TUV trade:

defaultType: "AI", "DoesNothing"
isHidden: "true"

Thoughts?

[Cernel]

@redrum As I reported in the past, this has always been a problem with Feudal Japan FFA. I remember to have reported the AI destroying itself to kill huge minor clan stacks.
For some reasons, I believe none of those parameters are really very good, and there is the item that is not helping existing maps (for example, that is doing nothing for Feudal Japan, unless someone updates it).
As you have said, the matter is that they cannot attack; so I would just have your AI consider "Neutral" whatever player that lacks a "Combat Movement" phase.
A better, albeit likely less feasible to implement, way would be the AI being able to see who is lacking a win condition, and this would be better on a FFA perspective, as the AI should also take into account who is closer to its win condition (so, those having no win condition are always at 0%, and the others the AI would have to evaluate (so, practically, in a FFA a player in a very bad shape, that has almost no way to win, would be seen just like a "Neutral" one)). Obviously, this would cover the above point too, as not having a "Combat Movement" phase would almost assure you cannot win, aside from very strange victory conditions.
But I'm guessing seeing who can and is closest to win the game is very hard to implement, so just testing for the absence of a "Combat Movement" phase should cover the matter at hand.
Side note "Neutral" is really not a good definition, and I suggest to rather reference it as "Null", since "Neutral" (the player) is not "Neutral" (the relationship), but always at "War" with everyone.

[theredbaron]

@redrum I just wanted to poke my head in to say that this sounds like a smart idea. In the end, DoesNothing AI is functionally the same (from a player's perspective) to static neutral units. I would actually suggest dropping the isHidden and sticking with only checking for DoesNothing AI, as whether it is shown should probably not affect whether the AI should perceive it as a threat. This will be a great improvement to AI behavior on this map. Frostion may be able to expound on this, but I imagine this problem also exists in Age of Tribes, so we'll have improvements on a few fronts.