Iron War - Official Thread

[redrum]

@black_elk 20 VC might be too easy for the Allies. Though glancing at the game, it looks to actually be a somewhat interesting game into round 5. Are you planning to continue playing it?

[Black_Elk]

I went another round till they stomped Finland, but then the hardAI started to crawl, so I just gave up the ghost and launched a new game vs the Fast AI. Yeah the VC spread might be a little tight for the Allies early on, but I think a human could probably prevent the 20 VC win without breaking a sweat. I still evaluate the set up more in terms of the single player vs AI experience though, since I think that would be a popular way to play a map on this scale. As for Japan, its kind of the same deal, a human would probably be more aggressive vs Japan with a triple-team, or octouple-team or whatever, but when you're flying solo leaving it up to the AI, Japan brings the pain. Another fighter or a bomber for the far east might help stem the tide a bit. In the backfield vs Japan its usually a lot of smaller battles, rather than a large stack fest, so even a couple mobile attack points in the air might make a big difference.

Anyway, here's where I am in the new game at the moment vs the fast AI. This time I tried my hand at a Russian Baltic fleet and forward positioning as a way to trip up the Germans. So far so good along the front with G, they are surrounded by the red army. But Japan is starting to chip away things on the other side of the globe, and Italy is still hanging on in Europe.

0_1518484374337_Elk vs FastAI Axis Russia 9.tsvg

[Frostion]

@Black_Elk
Very good and constructive feedback, as always I really appreciate it. You have again pointed out some issues that have also crossed my mind, vaguely. But it’s first when other people point them out that they become clear.

• Yes, as you and @redrum point out, the 20 VC win is problematic and favours the Allied, maybe especially if the Allies focus on a very specific VC strategy. I have experienced several AI vs AI Allied fast-wins where the Allied AI, most likely coincidentally, wins fast by getting 20 VC . The map already starts out with a very uneven VC score: Axis 11 / Allies 19, and even though the Allies normally lose a couple of territories early on, it seems that the Allies can just focus their efforts on like Finland, Iraq and Iran and get the upper hand relatively fast.

I reckon that we could try to fix this issue by tipping the stating VC balance by 1 for starters, forcing the Allies to capture at least one more territory to win.

I have now tried to fix this by moving the Irkutsk VC to Oslo/Norway. This makes Norway a bit more important. It also sets the Allied start back a bit. And finally it makes “Plan R 4” a possible UK strategy and goal. Of course Norway has now become True-Neutral along with their Iron. I guess that all this also gives a human Germany more incentive to invade Norway.

• The Soviet far east does have a tendency to fall to Japan relatively fast. The Soviet AI does not seem to focus on building in that that area and protect it. But if the soviet forces become too strong from the beginning, then the Japanese territories in that area are in danger of falling to a Soviet invasion.

I have tried to give the soviets a bit more resistance to the Japanese, while still not making the Japanese overall weaker, there are now a bit more USSR infantry in the area and the Japanese infantry is withdrawn further away from the border of USSR. To not make the Japanese overall weaker (because they now have to build more units at the border) Japan also get a bit more Infantry on the map, but they are spread out on several territories, away from the Soviet far east (like Truk, Shanghai and Peking).

The Far East changes really alter the conditions in the east, but I difficult to see if it does what is needed. It seems that the AI can’t really handle protecting the USSR far east coastline, and that AI Japan sometimes goes all in and invades the USSR coastline, taking many territories simultaneously.

• Another thing, that has not changed but that is open for change/balancing is the territory of Japan PU income of 40 PUs. It could go up to 50 if needed. To me it sometimes seems that Japan is overall under a lot of pleasure from like ALL corners of the world, and that Japan is maybe slowly “worn down” as the game progresses. What do you think about that? Does Japan normally have problems in the long run? Could they need another 10 PUs (cost of 1 infantry) per turn from turn 2?

Here is a new XML test version 0.2.21:
0_1518553887906_iron_war.xml
0_1518553937673_vc.txt
(Right click, download, rename and override old files)

[Black_Elk]

So far I like it a lot. The front between the USSR and Japan feels a lot more stable. The Russians have just enough to hold the initial line around Vladivostok or Manchuria for at least a few rounds instead of just immediately folding back to the Irkutsk pocket. Basically you have a better shot at fighting along that whole Far East Pacific corridor, so long as you don't risk too much on the coastline.

I also really like how this Norway thing is panning out. Definitely more interesting for the region with the 20th VC here, since a lot of powers have a shot at it. Nice call! I think it puts a new spin on that region for sure.

Started my first solo as Russia just to tease out the new opening options. Tried a similar play against Persia and Finland, but this time the Axis held off the early VC win. Pretty fun so far. Here we are in the 5th round, for comparison with that earlier game...

0_1518596497170_2.21 Elk vs FastAI Axis Russia 5.tsvg

I'd like to see how the Russian Ai does with the new stuff, so probably will try a Japanese solo next...

[Hepps]

@frostion This is a very innovative map. Even though I am not particularly keen on playing the AI, I am very happy to see this has begun a new phase of re-development.

Keep up the fantastic work.

[Frostion]

Thanks @Hepps I am totally aware of the fact that not that many people play vs AI that much, but at least I and a few others do, so I try to keep my maps very AI compatible. And lucky for us @redrum is very interested and diligent in regards to improving the AI.

[Hepps]

@frostion Yah. It's great that we have different people designing to suit different tastes in playing. It has really helped to develop Triple A as a well rounded game platform over the years.

[redrum]

@frostion If only @Hepps stopped designing such interesting maps that require new engine features, then I could spend more time improving the AI

[Black_Elk]

Yeah I think its kinda cool to have a map that works well for an AI match, since it services a gameplay niche that isn't always represented at this scale. Some of the smaller maps like v3 work alright, but there you really have to boost up the machine for it to be competitive and its rarely going to be as satisfying as a head to head game with another human player. I think Iron War has a few things going for it, beyond just the usual A&A fair, that keep it enjoyable for me in the stomp down "paint the map my color" type department haha. I like the way the production is handled.

I managed about 8 rounds as Russia last night before I had to go to sleep. US AI was starting to step up pretty major as I recall, taking Iceland, Norway and Morocco. Here is how it was winding down when I hit the sheets...
0_1518641257162_2.2 Elk vs FastAI Axis Russia 8.tsvg

Went another 5 rounds on a new game this morning, this time as Japan vs the Fast AI (definitely speedier going fast for the AI rather than hard.) I think the Russian balance has finally hit the sweet spot with those latest additions. They've got enough juice in the far east now to keep Japan from just steamrolling across the backfield, but not so much that Japan can't hold them in check with some concerted effort. Feels like its striking a pretty nice balance, at least from what I've seen so far. Also just noticed the British drop a factory in Norway, so seems like that is working out well too. Germany is still pretty deep into Russia, but its not the typical smash and grab with no mercy, since the Soviets are holding onto more income in the Far East. Anyhow, just wanted to post an update...

0_1518641179982_2.21 Elk vs FastAI Allies Japan 5.tsvg

ps. Took it 12 rounds to late 1945. While I was busy stomping Australia and trying to push 200 with Japan, the Allies sneaked in the 20 VC win over on the European side. Nice job from the Anglo-Americans! AI Germany still might have a little fight left. I imagine we could drag this one out for a while but it's getting late hehe
Fun stuff! Allies were definitely more formidable, keeping the pressure on.

0_1518659012392_2.21 Elk vs FastAI Allies Japan 12.tsvg

[prastle]

@black_elk I became a fan of the ai with @Frostion use for the ai in tribes. Every game was different. Not always the same even when human vs human. Since the ai wasn't always seeming to do the same thing twice. Just my two cents. Thanks for the great work @redrum

[Black_Elk]

Yeah I like the way it handles, especially when the AI gets a boost to be competitive. In a solo it's not as critical since the AI beats up on itself fairly well. I tried a couple 2.2 games where I picked one of the minor powers and let the AI loose on itself, just to see if it could hold its own. The balance in Asia felt pretty solid. Here's an example where I just took control of Brazil for 12 rounds. There's not much to do as Brazil, and it's not really a playable faction, so for all intents it was a placeholder to observe and AI vs AI game haha. Anyhow this is how it looked in 1945 with the AI beating up the AI.

0_1519683983400_2.2 Elk vs FastAI Axis Brazil 12.tsvg

Then I played a few full games against the fast AI just to see how things panned out. In the first I took control of all the Axis nations and gave the Fast AI Allies a 120% boost. Took that one almost to 1950 before I got sleepy.

0_1519684121860_2.2 Elk Axis vs FastAI Allies Germany 18.tsvg

Afterwards I did the same thing but playing as Allies vs fast AI Axis with the 120% boost... Here it is at the dawn of the Nuclear era, with most of the world handled.

0_1519684209788_2.2 Elk Allies vs FastAI Axis China 11.tsvg

In general I'd say the game where I played as Axis was more engaging than the game where I played as Allies. I think it's mainly because the Axis get to really paint the map, and expand with a sense of momentum, whereas the Allies spend most of the game trying to liberate territory until they can establish a foothold somewhere. Also on the Axis side the minor powers (Balkans/Finland, Iraq/Iran, and Thailand) are definitely under pressure just to stay alive, whereas the minor Allies (France/British-Colonies, French-Colonies/ANZAC/KNIL, and China/Brazil) are a bit more secure, at least when you're taking control of the whole team. I found that the best use of the minor Allies was to just spam aircraft, especially bombers if you can get them up past the 30 ipc threshold on income, after which point you can pretty well murder the Axis production. Still it was pretty entertaining. I think this might have been the first time I took control of all the minor Allies, since usually it feels like a lot of spinning plates and I just give them over to the AI. In this game Japan got hammered after leaving their transports exposed to an air raid, so they were kind of nerfed out the gate. Made it a lot easier for Russia to put the hurt on Germany. I had fun but got a little tired as 1945 set it, so had to call it a night.

FastAI was smooth compared to HardAI, for the lag factor, but I still find myself reloading after a round or two or when things hang, or to scrub a turn and see how the AI responds to different stuff. Sometimes seems faster to relaunch than to wait for the fastAI to figure itself out if its struggling haha, but they usually hum along pretty quickly compared to the hardAI grind.

[Black_Elk]

I wanted to show a more extreme demonstration of that Allied bomber spam I mentioned earlier. In save below you can see that by the end of the third round the Allies have completely shut down Germany's ability to produce units just with bombers alone. In all likelihood they're about to do the same thing with Italy next round, at which point the Allied bomber force will probably be large enough to pick off factories from the minor Axis nations at will....

In this example game I set up two main hubs from which to rain down death and destruction on the European Axis, with bomber armadas stationed in England and Egypt. From those spots I can reach pretty much anywhere that the European Axis blocks might want to build a factory. So basically, as long as I can put like 4 bombers on each of the possible factory locations per round, I can totally lock the European Axis out of the production game, just by using bombing raids.

I was pretty sloppy here too, not really doing anything with my combat moves other than bombing or bomber positioning, so I imagine this kind of play could be even more effective with a little more consideration about ground and naval forces. But yeah, just figured it might be worth looking into. For the most part I really enjoy how factories are easy to destroy and relatively cheap to replace, since I think it makes the territory trading game more interesting, but this lock out pay with the all bomber exploit is pretty powerful once you get things going.

0_1519763750916_2.2 Elk Allies vs FastAI Axis all Bombers round 3.tsvg

[Frostion]

@Black_Elk
I think the problem with bombing is that the AI can’t seem to protect itself against it. In Iron War a human has always been able to bomb the life out of the AI. No matter how the AA gun stats have been, the AI has never really been using AA effectively to defend against bombing.

Right now the AA Gun cost only 7 PUs (Infantry cost 10), making the AI actually purchases a few AA guns, but probably not with air defense in mind, rather just as a ground fighting unit in mind. The guns always march away from the factories.
I have always just kept myself from bombing the AI as I see it as unfair.
And I have always hoped the AI would get smarter at this one day.

How do you think humans would use the AA guns? Would they not be purchased?
Maybe the AA gun should just have other stats? Maybe be like a vanilla type A&A Anti-Air gun. If the current Bomber stats got nerfed to like 1-2 bombing damage instead of 1-3, I am pretty sure the AI would never purchase them.

[Black_Elk]

Not sure, but another aspect of the problem (at least from the AI) is that the machine also doesn't seem to realize that it needs to purchase factories to stay in the fight. So for example, here is that same game a round later...

0_1519766254818_2.2 Elk Allies vs FastAI Axis all Bombers round 5.tsvg

You can see that even though I've destroyed every Axis factory on the European side of the board, Germany and Italy fail to purchase new factories when their turns come up, despite having a ton of cash to do so.

I'm guessing this is probably related to that thing where the AI doesn't really defend its core factory territories (like in that earlier game where the Axis leave West Germany exposed to amphibious). Maybe because the unit type is different from the vanilla A&A factory, on account of being destroy-able?

Perhaps a workaround for the bomber spam would be allowing VCs to spawn infantry units or something, so that even if all factories are knocked off the board, the AI will still have a way to produce hitpoints, provided it has the cash and control of a VC?
A human would surely be more sensible with their factory purchases, and with interception vs bombers, but the machine doesn't seem to have a knack for this sort of thing haha.

[Hepps]

@black_elk What I heard is.... the best opponent is ALWAYS a human opponent.

[Black_Elk]

Doubtless, though I still think its worth exploring ways to tweak things so its harder pull a fast one on the machine haha. If only because the map is still pretty engaging for a solo, and my impression was that it was designed to make at least some use of the AI. A head to head match with all nations under human control would almost certainly look totally different than any of the saves posted above, but even then I could picture scenerios where a concerted bomber spam vs destroyable factories might undermine the production scheme. That's why I was thinking maybe a class of factory that can't be bombed or destroyed, which spawns infantry at the capital/VC might be interesting. I guess then you'd have 2 ways that units enter play, heavy equipment (the stuff that uses resources) requires factories to spawn, whereas regular infantry would just require control of one of the large population centers represented by the VCs. Feels kind of realistic and might make sense for a minor nation like China or Balkans where you could limit the roster based on the kinds of factory territories they have access too.

[General_Zod]

@Frostion

A nice way to fix weaknesses in AI play is to create a feature that knows when a nonhuman is playing a given nation.

Then the very well known areas where AI is weak, can be patched. ie. Stronger AAguns which are built into AI owned factories. Stronger or cheaper units for AI. Change the values of key territories on AI turns via triggers, so AI sees them as more important even can go so far as change the territory into a capital if you want to ensure AI values it.

And many other ways to give the AI help.

[General_Zod]

@Frostion Btw, I'm wondering if your gonna switch fuel models once the things are sorted in that department?

[Black_Elk]

I could see that, although I think part of the appeal of these A&A style games is how the mechanics are all basically transparent. So that even if you give the computer a buff (like with an income bonus) at least the player still knows what they're up against. Long as the player has a way to pop the hood and find out what's going on underneath it. That said, some way to prioritize target territories for the AI would be cool. Or to somehow set sz or land tiles with a marker so that the AI treats it as a seriously do or die spot to hold/take, the way British/American players might treat Berlin (or the batlic sea zone that borders it) in traditional A&A maps.

[redrum]

Yeah, improvements for the AI on AA gun management and per map settings are on the list: https://forums.triplea-game.org/topic/105/ai-development-discussion-and-feedback

In terms of whether AA vs strat bombing is balanced, its hard to say without playing a few full games. Looking at the numbers it appears the following is the summary:

• AA gun (7 PUs) - defends 1/10 @ 1 air unit for strat bombing
• Bomber (30 PUs) - SBR @ 1-3 damage
• Factory (25 PUs) - 5 bombing HP, destroyed on max damage

So unless the enemy stacks at least a few AA guns to guard a factory, usually 3 bombers will kill a factory and 4 would all but guarantee it. Probably in most cases you can reach at least 2 factories with bombers given the 6 movement so for 30 PUs, a nation can either buy 1 bomber or 4 AA guns (say 2 AA to defend 2 territories). Scaling that out for 4 bombers (120 PU), then enemy could buy 16 AA guns (8 AA per 2 territories). With 8 AA guns, that pretty much is gonna kill 1 bomber but the other 3 would probably kill a factory (30 PU vs 25 PU).

So from a PU perspective its probably reasonable but the issue is once you get to a certain bomber threshold, you can probably replace 1-2 bombers per turn and keep destroying the factory every turn. Its probably a bit overpowered given that countries like Germany/Italy only have 2 factories to start. I'd probably say bombing AA could use a buff or the bombing damage or HP for bombers/factories be adjusted. And yeah against the AI it would be devastating especially since it doesn't understand that AA can stack.

My gut says that air units in general are overpowered for their cost. Being able to buy a fighter or dive-bomber for the price of 2 infantry seems too cheap.