Is my AI out of adjustment?

[Professor Chaos]

Hi all -

I recently started my first campaign, playing Scenario 9 with the latest Beta against the Japanese AI. It is now mid-March, 1942. Much as I am impressed with the game, I am wondering if I should bother continuing with this game.

The problem is that the AI just keeps throwing piecemeal forces in all directions. It's working on finishing the job in PI and DEI, but keeps throwing other forces all over the map at other objectives. Unescorted invasion forces aimed at Port Moresby and Akyab have been massacred several times. An invasion force consisting entirely of one transport and a base unit showed up at Canton. Another base unit attempted to land solo in Java. A tanker and a destroyer showed up at Midway. Etc.

Meanwhile, the AI is mostly sailing the KB's carriers in circles around Truk and the DEI. Soryu was just sunk when it arrived at Canton (in the base hex) damaged, with an understrength air group, and a three-destroyer escort. Zuikaku was lost under similar circumstances. Shokaku sailed right into the minefield at Singapore. Two CVLs are gone when they sailed blithely past Soerbaja, where Force Z was waiting.


Now while it's fun to see all those carriers and transports go down, I'm wondering if there's going to be sufficient challenge in this game going forward. The AI has lost 405 ships so far, with relatively light Allied losses. I am not doing anything ahistorical in the game - conceded Wake, withdrew to Triangle in PI, defended Singapore and DEI with forces in place, usually kept my naval forces close to home bases, etc. It's not like the AI's losses are due to any strategic brilliance on my part.

To be clear, I'm not complaining about the AI per se, I'm just in doubt whether I should keep investing time in this game if the AI is not going to be capable of significant resistance given these losses.


(Meanwhile, despite this being "Quiet China," at least 3/4 of every resolution phase is occupied by reports of Japanese air and land attacks in China. I have not issued a single order to a Chinese unit since the game began, but after 3 months the attacks continue. I am going to try and withdraw every Chinese unit in contact with Japanese forces ans see if that helps)

[Amoral]

The AI does not adapt to serious losses or setbacks well. It is a script rather than an analytical engine. If it has got itself into a state where it is launching multiple unsuccessful operations it is unlikely to recover.

[rustysi]

There are several (don't recall how many) different scripts so starting a new game could get you a new one. From my experience (rather limited in AE) once it starts these type of attacks it doesn't stop. In a WitP game I had against the AI it started sending US battleships to attempt to bombard Milne Bay and didn't stop until I had sunk them all. New game.

[Lokasenna]

When these things happen, you should load up in head-to-head mode and change the things it's screwing up. After some time doing this, it'll go back to normal. You have to babysit the AI sometimes.

[PaxMondo]

ORIGINAL: Lokasenna

When these things happen, you should load up in head-to-head mode and change the things it's screwing up. After some time doing this, it'll go back to normal. You have to babysit the AI sometimes.

+1

Back up a few days before the disaster and change to head-to-head. Play 3 - 5 turns this way moving the units to a more effective locations and get some of them engaged in combat (this is key as the AI cannot summon units engaged). Then go back to normal play. 80% of the time this will work for me. When it fails, I need to go back a few more days earlier to initiate the fix and engage more units. It is a script, so all you have to do is to get the AI focused on other scripts or tie up units so that this script won't run.

[janh]

That's the best way to approach it for now. Just be careful to which "AI side" you are switching, i.e. if you are playing Allied and switching to Japanese AI you can accidentially activate the Russians since Japanese AI does not respect the garrison requirements in Manchuria and Korea.
Many times this approach works. You can even help AI recover a little by playing a turn or two at difficulty levels hard or very hary every now or then (giving it supply to recover squads and squadrons etc) in case you generally want to play at "historical level" to keep supply a meaningful quantity and be able to isolate islands or pocket land units.

You can certainly keep doing that until AI gets set on some scripts that are required for progress, e.g. recovery of Nomuea, Suva or such, in which case it will keep sending forces piecemeal for a long time (until the script terminates, which for defensive scripts usually is the max timeframe). When that happens and I begin to spend more time playing myself H2H, sort of, it is time to start over. AI will not recover from that, even if you gather all its most suitable vessels and LCU in the main port the script uses for counterinvasion and put all available combat ships there, it will not recognize previous defeats at that target and adjust the script accordingly, or group TF in larger, equally fast groups with escort TFs. The best thing in that case is to allow AI to grab these bases, especially if it is the Japanese AI in the expansion phase. Like most, on the defense this AI also acts better.

[Andy Mac]

I would love to clean slate the AI and do a rewrite even with just the tools we have but its just sooo much time

3rd time round I know I could do a better job but its time thats the issue

[janh]

I didn't mean to critize, Andy. The AI scripts are extremely complex and it is amazing how you kept an overview with all the units involved. It took me hours to read thru them and realize the structures and figure which units you associated with with task sets.

They work for a good many cases, but sometimes it were better AI could cheat a bit and check the player's dispositions at a location or in a (sea) area before committing TFs or air raids. I've looked at them a couple of times, but I saw no way to trick AI into adapting to situations -- since you have to set the script's unit sets and requirements beforehand, before any game and not during when the true situation develops. If that were possible to do with the script mechanisms, I'd be playing with that right away, but for what the script engine can do, I think you have already optimized everything. I don't think anybody could do it any better than you did.

If you have any ideas to share on how to improve things though...

[Professor Chaos]

Thanks for the responses folks! As a newbie I'm sure I'm not uncovering issues that haven't been recognized before. And Andy, I definitely didn't want to start an AI-rag fest.


It's not so much the AI pounding against unachievable objectives, but the fact that the carrier forces are either circling around Java, or showing up solo at Canton while major invasion fleets are getting massacred by relatively small forces at Akyab or Moresby.

It seems to me the issue is timing - I assume that once DEI operations are concluded, the AI would release major assets for use in the Indian Ocean or New Guinea. So I'm surprised there's no "wait until DEI and PI are conquered" before pushing further. Particularly by March, when the Allies can concentrate significant naval power in those areas.


But perhaps I have just drawn an aggressive AI. The AI had Rangoon and all the Solomons by the beginning of March. And the attacks on Port Moresby might have worked if I hadn't immediately concentrated forces there.

[Banzan]

When i started WitPAE, i often used small raider TFs (CLs + DDs) and planes to interrupt naval operations.
Nowdays i let the AI take what it wants for the first 6-8 months and only fight back on land for this time. This, together with the updated scenarios from Andy Mac were major advantages for my play experience. So much more fun now

[Ron Saueracker]

This AI issue is why I'd appreciate a fully dedicated Babes post Japanese expansion scenario that starts after the DEI and PI are Japanese. Must be easier to sort out an AI which basically defends (as Japan) or attacks (as Allies) I'd assume.

[Professor Chaos]

I give up.

It is now mid-May, 1942.

• I just sank a force of BBs and CAs that decided to bombard Port Moresby while my carriers were nearby escorting a supply TF. I didn't even know they were there.
• While the invasion forces are being pulped, the IJN's carriers are still sailing in endless circles around Kwajalein, or once again headed solo to Canton.
• Despite this being a "Quiet China" scenario, there are still dozens of air raids and several land battles every turn in China, including most recently an amphibious invasion of Pakhoi.

The AI also tried to invade Diego Garcia, and is planning an invasion of Auckland. At this point I have sunk between a quarter and a third of the IJN's carriers, battleships, and cruisers.


I would like to believe something has gone awry with the AI in this game (scenario 9), and this is not a typical AE experience. But having sunk perhaps a hundred hours into this playthrough, I find it very difficult to contemplate starting again unless the experience is likely to be different.

[&:]

[Lokasenna]

You could always try Scenario 10, Ironman! You'll suffer no lack of targets in that one.

Are you using the database updates and unofficial scenario patches? I don't really remember, but Andy may have tweaked the AI scripts with the scenario patches.

It sounds like your AI is in a loop and you need to load up in head-to-head for a while. Maybe go back a couple of days (or weeks) if you have auto saves. Change the orders of things like the CV TFs, the BB force you just sunk, etc. The game against the AI really does need babying every now and then, as the game is just too complex for the AI to work optimally all of the time.

[Professor Chaos]

I wasn't using the unofficial data updates, because there are none for the Quiet China scenarios. I was using the latest public beta of the program files.


I've given up on this play; if I decide to start a new game I might try one of the updated scenarios. China wasn't very quiet in this play anways.

[rustysi]

Myself, I'm just using the AI and this forum as a means to progress to PBEM. Hopefully I'll get there in about a year. Then hopefully I'll have the time to commit also.

[Lokasenna]

Actually, things like the invasion of Pakhoi are well within the Quiet China scenario. For Quiet China, the AI will (try to) achieve certain goals before it settles down. Pakhoi should be on that list, as should Hong Kong and the area between Canton and Kweilin. Unsure about the rest.

I heartily recommend that you give it another go, either against Ironman AI, as Japan, or even against a human! Or perhaps try the Guadalcanal scenario or something. It's pretty fun.

[CaptBeefheart]

Let me guess, the 14th Chinese Expeditionary Army is planning for Auckland? Let's just say that might be misdirection on Andy Mac's part. [;)]

The last couple of times I played the game I did Ironman and I let the two sides duke it out until at least May 1942. You can set both sides to "computer" and "continuous" and let them play. Go to bed, wake up and see where things are by hitting the nine key if memory serves (or is it F10?). Anyway, one of the keys stops the game and then you can load up as Allies and see how things have gone to that point. That'll give you more of a challenge. Also, you should play on hard difficulty.

Cheers,
CC