Page 3 of 3
RE: Axis AI vs Redmarkus
Posted: Fri Feb 17, 2012 5:07 am
by Redmarkus5
ORIGINAL: randallw
In my own games the computer threatens Minsk on it's first turn and takes it on the second ( I abandon any attempt to defend it ). It would be strange if the new AI takes it later.
Look at Larry's AAR if you don't trust mine. The AI doesn't take Minsk even after he vacates the city.
RE: Axis AI vs Redmarkus
Posted: Fri Feb 17, 2012 5:14 am
by Redmarkus5
Gents, the bottom line.
The
essence of the War in the East is not the challenges faced by the Soviet leadership at 5pm on 22 June 1941. It's the challenges they faced in front of Leningrad and Moscow
from late summer through to winter.
If the AI can't be programmed to do this (pic below) even to an opponent who doesn't attempt to resist, then the game does not model WW2 on the eastern front - it just models what happens when two armies of game counters meet on a map that resembles eastern Europe.
If the player opts to change history by employing different tactics, that's fine. But if the player decides to let history run its course, the game MUST be capable of inflicting some version of the historical carnage suffered by the Red Army in 1941.
RE: Axis AI vs Redmarkus
Posted: Fri Feb 17, 2012 5:16 am
by Redmarkus5
ORIGINAL: CarnageINC
Ahhh.....forget this AI stuff Red and just pick up in January of our last game [:'(]....Come back to the darkside!
Sadly Joel (I think it was him) is saying that the latest changes will be an issue for games in progress I believe.
I'll play you again from 22 June if you like?
RE: Axis AI vs Redmarkus
Posted: Fri Feb 17, 2012 5:53 am
by randallw
This Minsk thing sounds like another one of those city bypassing problems.
RE: Axis AI vs Redmarkus
Posted: Fri Feb 17, 2012 5:56 am
by Apollo11
Hi all,
Both Minsk and Riga fell to Axis in Larry's game... just as it should be... if "redmarkus4" continues his game against AI it will be the same... [;)]
Leo "Apollo11"
RE: Axis AI vs Redmarkus
Posted: Fri Feb 17, 2012 9:01 am
by Redmarkus5
ORIGINAL: Apollo11
Hi all,
Both Minsk and Riga fell to Axis in Larry's game... just as it should be... if "redmarkus4" continues his game against AI it will be the same... [;)]
Leo "Apollo11"
YES. They fell -
eventually. Which part of my posts above are you not understanding?
On Turn 5 for Riga. The AI ignored an empty Minsk for one full turn in Larry's game. It waited till T4 to attack it in my game.
This is exactly why I get so frustrated, annoyed and offensive in these forums. There is clearly an important bug here but all we can get out of the other side is this kind of fuzzy push back which runs on for weeks until finally we get agreement that, yes, there's an issue.
Go test it for yourself and please get it fixed.
End of thread.
RE: Axis AI vs Redmarkus
Posted: Fri Feb 17, 2012 9:01 am
by Redmarkus5
ORIGINAL: randallw
This Minsk thing sounds like another one of those city bypassing problems.
Jeez - you mean there's actually a known bug and it's got a name???
RE: Axis AI vs Redmarkus
Posted: Fri Feb 17, 2012 9:08 am
by Redmarkus5
Leo - go check out the "city bypassing problem" bug with your colleagues please.
RE: Axis AI vs Redmarkus
Posted: Fri Feb 17, 2012 9:10 am
by Redmarkus5
Actual end of Thread here
RE: Axis AI vs Redmarkus
Posted: Fri Feb 17, 2012 9:19 am
by Apollo11
Hi all,
So you are abandoning your own test before it even began because AI didn't take those cities as quickly as you want it?
You don't want to play it at least until end of 1941 to see how much AI can do in 1941 and end of 1942 to see how much it can do in 1942?
Well... OK...
All I can say is that AI works - it is proven and it has no relevance for grand campaign if AI takes Minsk and/or Riga few days later than human player would do...
BTW, AI is AI (and we should never forget that) - in WitE it really does splendid job and it offers human players good adversary regardless of the side they choose to play!
Leo "Apollo11"
RE: Axis AI vs Redmarkus
Posted: Fri Feb 17, 2012 11:03 am
by Redmarkus5
"it is proven and it has no relevance for grand campaign if AI takes Minsk and/or Riga few days later than human player would do... "
A few weeks later, not a few days later.
There is no point continuing the Soviet test beyond T4 because I didn't take any defensive actions during the first 4 turns, so the result now will not tell us anything about how the AI deals with a real opponent. Plus, I find it boring to test an AI that's not going to really start attacking me in full force until Turn 8, or whenever.
I am testing from the Axis side now to see how the AI handles the Soviets in defence.
RE: Axis AI vs Redmarkus
Posted: Sat Feb 18, 2012 7:25 am
by Apollo11
Hi all,
ORIGINAL: redmarkus4
"it is proven and it has no relevance for grand campaign if AI takes Minsk and/or Riga few days later than human player would do... "
A few weeks later, not a few days later.
There is no point continuing the Soviet test beyond T4 because I didn't take any defensive actions during the first 4 turns, so the result now will not tell us anything about how the AI deals with a real opponent. Plus, I find it boring to test an AI that's not going to really start attacking me in full force until Turn 8, or whenever.
In the grand scheme of long Grand Campaign for AI it is of no special importance that Minsk and/or Riga are taken immediately. For human it will be different (especial those humans who micromanage rail repair and concentrate on long single tracks for support of advancing forces) but for AI it is not...
Also - the AI will always advance slower than human player - the initial AI turn is scripted and it is the deepest thrust the AI will ever do - all later AI moves are much much slower and paced!
IMHO this is good and it ensures the AI "consistency" and paves way for methodical advance (and not "rush" like experienced humans would do - that is outside of reach for AI)...
Therefore, based on all this I wrote that AI performance can only be judged in long run!
Axis AI will do the following in the grand campaign:
#1
Methodically advance till Blizzard in winter of 1941 (usually it will advance east less than historic Germans)
#2
Defend in Blizzard of winter 1941/1942
#3
Consolidate in spring 1942 and advance in the summer / autumn 1942 (this is the time I saw the best AI advance and this is where AI usually took Leningrad, Moscow and Crimea in test games)
Leo "Apollo11"
RE: Axis AI vs Redmarkus
Posted: Sat Feb 18, 2012 11:21 am
by AFV
I am impressed the AI is as good as it is, truly. For a game that cost $70-80, its impressive. However, I think we all can agree it could use improvement. Perhaps any improvement is not feasible or not possible in the short term. Or perhaps it is, hell I don't know, I don't run the company.
So, with that said. the first few turns in WITE are truly unique for the Axis in WITE. I certainly understand redmarkus concerns.
So:
Is it impossible or unreasonable to script in some early turn behavior beyond turn 1 for the AI?
For example: City x undefended, take it.
City x weakly defended, attack it.
If turn = 1, 2 or 3, take more risks than usual?
Leo, you seem defensive regarding the AI. The issue is not what the AI will do during the first year of the war. I have confidence it will put up a good fight.
The issue here is during the first few turns of the game, where the AI as a German is faced with a unique situation it will not see, ever again, -a completely surprised, weak, disorganized Russian army incapable of defending itself. It seems to me, it might be possible to script in some early turn behavior beyond turn 1 for the Axis AI. Then again, maybe that is just not possible (for a game that costs $80).
RE: Axis AI vs Redmarkus
Posted: Sat Feb 18, 2012 11:33 am
by Apollo11
Hi all,
ORIGINAL: AFV
I am impressed the AI is as good as it is, truly. For a game that cost $70-80, its impressive. However, I think we all can agree it could use improvement. Perhaps any improvement is not feasible or not possible in the short term. Or perhaps it is, hell I don't know, I don't run the company.
So, with that said. the first few turns in WITE are truly unique for the Axis in WITE. I certainly understand redmarkus concerns.
So:
Is it impossible or unreasonable to script in some early turn behavior beyond turn 1 for the AI?
For example: City x undefended, take it.
City x weakly defended, attack it.
If turn = 1, 2 or 3, take more risks than usual?
Leo, you seem defensive regarding the AI. The issue is not what the AI will do during the first year of the war. I have confidence it will put up a good fight.
The issue here is during the first few turns of the game, where the AI as a German is faced with a unique situation it will not see, ever again, -a completely surprised, weak, disorganized Russian army incapable of defending itself. It seems to me, it might be possible to script in some early turn behavior beyond turn 1 for the Axis AI. Then again, maybe that is just not possible (for a game that costs $80).
Only the 1st turn of grand campaign is scripted.
BTW, the opening move was made by one of our "Grandmasters" - Andy "Sabre21" (one of the best Axis / Allied players we have)! [:)][&o]
It took weeks of hard work to do it - it is no trivial task (because the script must work in various user selectable settings)...
After that 1st turn the AI is playing normally and it is not possible to script further moves.
BTW, in grand scheme of things the AI is not depending on taking those towns you mention - all bypassed (and encircled) enemy units will eventually be destroyed - so it is no big problem for AI - the only difference is that human player will do it much faster - the AI will do it eventually! [;)]
In all other turns after 1st turn the AI will fight as usual (i.e. "grinding") but it will most certainly try to encircle the enemy if AI is "Attack Mode" and appropriate calculations show that certain criteria is met. This was doe in most recent patches...
Leo "Apollo11"
RE: Axis AI vs Redmarkus
Posted: Sat Feb 18, 2012 11:42 am
by Commanderski
I just started a new campaign against the Axis AI with 110% across the board for the AI and 100% for me. I also used the Better Position Grand Campaign. On Turn 1 it took Riga, I have 1 unit next to so I can retake it and it's already within 1 hex of Minsk. In my previous campaign it took Minsk in tuirn 3 but still didn't Riga.
The AI will not take a city with it's armor or even most of the motorized divisions in the opoening drives It tries to take as much territory as possible and will bypass everything it can until it runs out of gas or hits the front lines. The infantry that comes up behind will take the cities.
At 110% there were only 3 unrouted and 2 routed units left behind the lines in AGN's area. That's very impressive. At this level I think I have quite a fight on my hands.
I'll start an AAR later today.
RE: Axis AI vs Redmarkus
Posted: Sat Feb 18, 2012 6:48 pm
by randallw
Ah hahhhhh....so that's how the AI is programmed, to avoid moving against a city with mech units, because of the attack CV penalty, and making advancement a priority for the mech?
RE: Axis AI vs Redmarkus
Posted: Sun Feb 19, 2012 2:14 am
by Joel Billings
Using finess attacks with few units is always hard for an AI. The AI will do better when it has it's infantry and isn't just working with the limited armored units. It will also do better with 110% logistics help because it's not as good as a human player at keeping his spearheads in supply via maxing rail repair and use of air transports. Even so, I agree with Leo that you have to let the AI roll through at least to the end of 1941 and then compare how it did. If you are asking the AI with limited help to repeat the impressive blitzkrieg of the first month (or two) of the war, you are asking for a lot. Even experienced wargamers have to get some experience with WitE to do this. Taking cities early on is not a true measure of the AI as it really does not care that much about them. If the AI is too aggressive with it's mobile forces without infantry support, it can find itself in trouble, so it's a balancing act. If it has strength, it will look to encircle where it can. My suggestion is boost the help level higher if it's not giving you the challeng you are looking for. If it then suddenly surpirses you and it becomes too tough, you can always tone it down mid game.