AI and History

[herwin]

I've been playing the Allies solitaire against the AI on normal. Nik Mod, with the usual results through May 1942. What is interesting is the failure of the AI to stick with Japanese strategic doctrine. I've been running 3 Allied TFs (two RN air and one US surface) in direct support of the Dutch forces in Java, operating from bases on Java, Timor, and at Darwin, and I've caught both the KB and some BB TFs with their pants down. The damage to the KB hasn't been that serious, but I've sunk two Kongo class BBs and the Maya and sent two older BBs home with torpedoes. I've spotted all the IJN BBs except the Yamato in the area.

My point is that the IJN GHQ would hardly have used their BBs to support landings in Indonesia. Right now, the IJN has 7 operable BBs, while the Allies have about twice as many, which was basically the IJN nightmare scenario. Sure the KB can probably fix those numbers easily, but Nagano and Yamamoto didn't know that at the time. Remember, Operation MI was planned to culminate with a surface engagement. The AI has a problem here.

[Capt. Harlock]

That is rather interesting. Two points to ponder: the AI is ideally supposed to do better that historical, since the IJN actually lost the war, and the program should be able to give an Allied player a good fight. Second, the presence of British CV's in the area was not historical; perhaps the AI was responding to that but for some reason could not commit the KB.

[jeffs]

I disagree on the quality of the AI....For a game as complicated as this, I do not think, unless one is spending millions of dollars on the best IA designers, that they AI can be expected to be competitive with good players (assuming it is not cheating on battle results).

Chess is much more limited in possible moves and yet it has taken many years and supercomputers to reach top class levels..

[herwin]

ORIGINAL: herwin

I've been playing the Allies solitaire against the AI on normal. Nik Mod, with the usual results through May 1942. What is interesting is the failure of the AI to stick with Japanese strategic doctrine. I've been running 3 Allied TFs (two RN air and one US surface) in direct support of the Dutch forces in Java, operating from bases on Java, Timor, and at Darwin, and I've caught both the KB and some BB TFs with their pants down. The damage to the KB hasn't been that serious, but I've sunk two Kongo class BBs and the Maya and sent two older BBs home with torpedoes. I've spotted all the IJN BBs except the Yamato in the area.

My point is that the IJN GHQ would hardly have used their BBs to support landings in Indonesia. Right now, the IJN has 7 operable BBs, while the Allies have about twice as many, which was basically the IJN nightmare scenario. Sure the KB can probably fix those numbers easily, but Nagano and Yamamoto didn't know that at the time. Remember, Operation MI was planned to culminate with a surface engagement. The AI has a problem here.

Now more of same. The AI sent the Fuso and the Yamato to bombard Soerabaya on June 4th. They were caught leaving by a couple of heavy bomber squadrons--the Fuso may *possibly* survive, and the Yamato took a torpedo and some bomb hits, so she will be out of action for a while. That leaves two Kongo class, three older BBs, and the KB to defend Japan. Looks like it's time to activate the US Battle Fleet (currently with 9 older BBs fully repaired), the Far Eastern Fleet (currently 1 BC, 5 BB, and 4 CV/CVLs), and the Fast Carrier Force (5 CVs) and try to box in the KB (currently off Amboina). Quantity has a special quality of its very own.

[herwin]

Six months into the campaign, I switched sides to see what the AI was doing. There were TFs spread *all* over the ocean, most containing damaged ships. TF 1 (the KB) was near Palau, completely out of fuel, and crawling towards Truk. The Japanese CVs averaged about 25% systems damage. The Japanese BBs averaged about 40% systems damage. The Kongo was sunk. Almost all the Japanese LCUs were in small pieces and out of supply. If I had known what shape the AI was in, I could have mopped up the campaign in the Summer of 42.

[Capt. Harlock]

Wow -- something is way wrong there. You'd think that returning a TF to port before it runs out of fuel would be one of the few things that AI would do better than a human player. Also, how many of the damaged BB's were repairing, and how many were left active?

[DD696]

Hi,

That has also been my experience playing as the Allies vs AI Japan. The Japanese forces run around out of fuel and with what I would consider far too much sys damage prior to getting some yard work done. I have also noticed the LCU's do become heavily fragmented. I have at times tried to send Japanese fleet units into repair yards in the hope that they would take care of damage, but they seem to reform and plod on. Another thing I have noticed is that the AI does not appear to move fuel to forward bases - even Truk has a minimal amount.

A major problem as has been noted elsewhere - using Buna as an example - is that the AI sends endless supply task forces there to be bombed into oblivion. It will also park surface combat TF's there to be used as bombing practice. I do try to limit my attacks on these "sitting ducks" by standing down the air units, but sometimes the temptation is too great to ignore. Many times I have decimated the Japanese Navy without the US Navy ever catching sight of a single Japanese ship.

If these problems could be solved, the game would be a lot better. And, for you PBEM'ers who are going to slam me for playing the AI - when you can provide me with 15-30 turns a day then I will be happy to play against you.

[herwin]

ORIGINAL: Capt. Harlock

Wow -- something is way wrong there. You'd think that returning a TF to port before it runs out of fuel would be one of the few things that AI would do better than a human player. Also, how many of the damaged BB's were repairing, and how many were left active?

All appear to be active; none repairing.

[herwin]

ORIGINAL: DD696

Hi,

That has also been my experience playing as the Allies vs AI Japan. The Japanese forces run around out of fuel and with what I would consider far too much sys damage prior to getting some yard work done. I have also noticed the LCU's do become heavily fragmented. I have at times tried to send Japanese fleet units into repair yards in the hope that they would take care of damage, but they seem to reform and plod on. Another thing I have noticed is that the AI does not appear to move fuel to forward bases - even Truk has a minimal amount.

Generally what I'm seeing.

A major problem as has been noted elsewhere - using Buna as an example - is that the AI sends endless supply task forces there to be bombed into oblivion. It will also park surface combat TF's there to be used as bombing practice. I do try to limit my attacks on these "sitting ducks" by standing down the air units, but sometimes the temptation is too great to ignore. Many times I have decimated the Japanese Navy without the US Navy ever catching sight of a single Japanese ship.

If these problems could be solved, the game would be a lot better. And, for you PBEM'ers who are going to slam me for playing the AI - when you can provide me with 15-30 turns a day then I will be happy to play against you.

Amen, brother!

[BLUESBOB]

When playing against the AI I always put it on Hard or Very Hard, and give it a variable set-up. It seems like the only way to give the AI a "human" like quality and takes away my ability to know every where and when of Japanese strategic plans.

I was playing a scenario where I had the Prince of Wales, the Repulse, and every CA, CL, DD, PG, PT, SS I could muster sitting in Soerbaja waiting to tangle with the Jap invasion fleets coming my way. I thought I would plaster them good. Then I spotted a Jap CV with other ships off Davao heading south. I thought it was the Ryujo doing what it always does...heading off escape from the Philippines. I decided to send everything out and take it on. A loss of a few ships would be nothing for the Jap loss of many.

As the Japs got closer to Soerbaja I got reports of a CV and BB's. Then I got a report of the Zuiho. I thought it was a mini-KB and stayed with my plan. I had about 50 ships to tangle with it...more than any mini-KB could handle.

Totally forgetting that my game was not a historical game, I went out to take on the "mini-KB" off Kendari. I got hit off Macassar. Turns out it was the Akagi, Kaga, Hiryu, Soryu, & Zuiho I decided to take on. The other fleet off Davao was the Ryujo, along with the Shokaku, Zuikaku and Shoho. I lost just about everything and learned a big lesson.

This wouldn't have happened had I kept to historical settings. I would have wiped out anything the Japs sent my way. Having knowledge of where everything is at any given moment and where it's going to go is not a very good way to test a strategy. The variables help keep you on your toes as a human player would.

[qgaliana]

Playing the AI now I get the feeling it doesn't react well to enemy forces, mostly it seems to attack the geography. Typical examples - it will ruthlessly pound large empty airbases, regardless of damage, and ignore others where TBs are ripping its transports. Invasion forces look sized for the base, not the troops defending it. Naval units concentrate in vulnerable areas without CAP...

It feels like playing a blind man. I don't get the impression that the recon information we see on the map is being handled by the AI. Anyone else get that feeling?

[Grotius]

Yeah, I know what you mean about recon. When I went from AI to PBEM, suddenly recon loomed much larger in importance! Still, I find the AI gives me a decent game if I play in a roughly historical fashion, on Very Hard, and ignore the occasional boneheaded move of my computer opponent. It also helps to play a scenario in which the AI has the advantage in numbers. I'm enjoying playing Japan in the 1943 campaign. I may take it to 1945 or start a new one in 1944/45.

I prefer PBEM, but I just don't have the time to commit to it right now. With the AI, I can play ten turns tonight, then skip it for a few days. In PBEM, that sort of erratic schedule isn't fair to my opponent.