ROM bought and license-built about 100 SM.79B and SM.79JR versus several hundred fighters and light bombers each.
-
We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
AI does always upgrade production line when a unit is finished. If it uses less ic in the begnning for ships it can use more ic for infra and build more modern ships later.
Perhaps it just did not look that way to me, especially since my games have much longer production times (ships, most land, but not aircraft).
I have done a lot with this game in the last 6 months, but with a special desire in mind -- to create a better human USA vs. AI Japan game. It can lend itself to multiplayer as well with the right combination of countries played.
Some of what I have done I will share if desired for it is applicable to the normal vanilla game.
My starting point though is AOD 1.08 + SEMP 6 or whatever the latest is called. I rrealize that my discussion here that follows likely belongs in the mod section, but since there are so few of us remaining, I will give a brief overview here.
1936 Scenario only -- I have:
1. added infrastructure and/or factory lines to UK, AST, CAN, NZ, So.Afr. game start files
2. added locked 9 division garrisons in Berlin, London, Rome, Leningrad, Moscow, Stalingrad to stop cheap and easy capture.
3. added 3 div locked garrisons on key beaches in UK, Italy, Japan & key islands.
4. added 3 div locked garrisons in Tripoli and El Alamein
5. significantly increased build times for most ships and ground divisions
6.Added a lot of coastal forts for Japan in the Pacific.
7. Modified the AI files for UK, Jap, Ita navy to have them sail more by decreasing org/damage restrictions for sailing
8. Modified Jap AI files so that more ships/less infantry are built -- as a human player would do
9. Modified German victory events over USSR giving them a lot of free qualty divisions, oil, and a TC boost (sort of like what happened when Russia left WWi allowing the Ludendorff Offensive).
10. Wrote 40+ events for Japan so the AI will in a systemically but randomly capture unguarded Pacific islands/beaches if left unguarded -- to simulate what a human Japan would do. This keeps a human USA "honest".
11. Greatly increased the early Pacific war bonus for AI Japan
12. The 10 historic USA battleships completed 1940-1945 problem -- due to 20/20 hindsight a human USA player should not build these. Instead, I give them to the USA player for free in the historic years. Off-setting this freeby are a number of things -- including the 1938 USA recession caused by the start of the Social Security program which caused employers to stop hiring and lay off workers (sounds like Obamacare) and caused 15% unemployment. Japan gets the 2 super BB's free plus all starting DD's & CL's are raised to class IV simulating what a human player would be doing.
13. USA cannot research the 2 sub techs until 1943 simulating the magnetic torpedo problem that they had until mid 1943
14. Japan gets the free 1943 fighter in 1940 to simulate the Zero and what it meant in combat in the Pacific early in the war.
15. Japan starts the war with the long-lance torpedo -- 5 sea attack -- a new class III that only they can build.
16. The Kamakazes appear when certain Islands are captured and are represented by flying rocket units
17. The historic 8 USA BB's are damaged by the Pearl event recieving 50-95% damage.
18. A new land brigade is created uniquely for the USA simulating the superior artillery/tank attachments given most all USA divisions
That is most of what I have for now. It has been playtested a lot by me and 2 buddies who join in twice a week for networked play and testing. I currently update the game twice a week with new changes. Right now I am in the process of cleaning up the SEMP USSR surrender events -- which needs some work. My next priority is to legitimately influence the Jap AI to "learn" to recognise weak spots in island defenses and launch invasions with the proper force. I also feel that the German AI following a USSR surrender needs a LOT of work. It seems much easier to "teach" the AI to attack then it is to "teach" the AI to defend.
Go froglegs go! Keep up the good work! Hopefully some of this is included in 1.09, if you (and Pang) want it to be
A new set is uploaded. The only relevant change is the research ai of afghanistan. I have not tested it yet. I intend to use it later as a blue print for other countries aswell.
True, but perhaps the locked divisions are includable at least, considering the stupidity of the AI. Plus the infrastructure lines are certainly includable. And the Pearl Harbour event with the 8 ships being damaged is definitely historic and would help gameplay, Japan needs all the help it can get! And the Pearl Harbour event, at the moment, benefits the US way more then Japan, considering the slider moves the US gets.Most of that appears to be way overmodded to be included in any official patch.
Here are modified AI files for Pang Bingxun, but everyone can check them of course. They mostly affect building schemes for some countries.
With the possible exception of ITA those building schemes are for countries where i intend to leave things (mostly) as they are, because they have already been given thorough testing in 1.08.
A new set is online. The most important change is deleting everything in \db\events\ai_reactive.txt. Those events kept on messing up the UK AI.
What's the problem there? These events have a purpose and if they are firing too early an adjustment of the triggers is better than anything else. Hint: dormant units counts as active units for the checks made by many events.
They have a purpose, or at least they had. But adding anything into the building queue via event is messing up the building scheme, especially due to very short production lines size = 1. Using different events chain to load different building schemes might be an option. But since relative building schemes are in use those events appear to be simply outdated.
You should at least keep the reaction ship builds (size 1 should not be a problem there). Is there an option to add production sized higher than 1 via event?
It can still mess up the building scheme. Before deleting the events i had 240 ic need in production with 60 ic allocated. This is inacceptable. It is more elegant to adapt relative building schemes than to skript the building queue directly.
No.
My bad, i didn't tested 1.08 AI as much as 1.07, but taking Germany as an example, i can still see inaccurate results. They build too many motorized infantry and medium tanks instead of light in pre-war/early war time. Also, they build too many transport ships, which is a waste of IC IMO. They do well at war, but not exactly all the time.
I understand, that things are made for balance, but in most of my test games, using my schemes Germany managed to do "Fall Weiss" with quite historical forces, in almost historical time (about 1 month) and with quite historical manpower losses (around 20k). Other campaigns were also more historical, than with vanilla, so the game itself is rather realism-friendly. Isn't historical accuracy preferred for AI vs AI ?
This is WAD. I decided so because of given limitations in AI.
Unless they prepare Sea Lion they are likely to not build more than 10 transports. So if there exists a case of too many transports they only build slightly too much. Given that AI is likely to lose some transports a reserve of those cheap units makes sense.
To some degree, yes. But more import is that AI performs as good as circumstances will allow. AI should be more than a sparring partner to reproduce the historical outcome of WW2. Making Germany able to nuke USA after Bitter Peace seems more important.
I have no idea what limitations You mean. I tested german schemes many times and composition of land forces with (about) 90 INF, 5-7 MOT, 6-7 LARM, 6-8 ARM plus several MNT (which is still not historical BTW) worked just well.
Yes, this is not so important, but still ~10 transport in almost every game is a bit of valuable IC. Can't they just start with 2 or 3 transports already built ?
To some degree, yes. But more import is that AI performs as good as circumstances will allow. AI should be more than a sparring partner to reproduce the historical outcome of WW2. Making Germany able to nuke USA after Bitter Peace seems more important.
I don't expect the game to be nearly 100% realistic and historical accurate. I expect it, to be as much of it, as possible, not at all cost though. I suppose it's only an example with that nuking possibility, yet good example. I never saw any events for underground factories built by Germans IRL or testing facilities reflected by bonuses for researching techs. Things, that can be used, not only in mods.
Limitations are that AI cannot calculate ahead that l.Arm will be upgraded to Arm. It cannot utilize "small" spearheads like human player can. Therefore many Mots are needed to make the german AI at least some sort of a challange.
The AI should be able to change to composition of forces in the Königsberg area fast. And imo it should also be able to surprise with an early Sea Lion.
A new set is online. Germany starts now the Bismarck build at the start. There are a lots of new trades. The UK received a relocation of stockpiles to accomodate the needs of the AI.
This seems reasonable, although i'm still wondering why the AI was doing good enough with more historical schemes.
Well, there are in fact some tricks, that could make Germany (and german AI) more challenging and surprisingly more historical accurate, but i can accept this to be left for now if not permanently.
Now i hope for some updates of forts in 1936 scenario (not only though) like german "Ostwall" or soviet "Stalin's Line". Good for alternative, but realistic history outcomes.
Well, there are in fact some tricks, that could make Germany (and german AI) more challenging and surprisingly more historical accurate, but i can accept this to be left for now if not permanently.
Now i hope for some updates of forts in 1936 scenario (not only though) like german "Ostwall" or soviet "Stalin's Line". Good for alternative, but realistic history outcomes.