The Last Conformist
Irresistibly Attractive
Possibly not the right thread, but how is the AI handling the new AA units?
The Big Question - How Does The AI Choose Which Units To Build?
- Thread starter Ozymandias
- Start date
I don't have any direct experience myself yet, but rumor has it that they're handled every bit as well as (per our earlier discussions) PTW handled artillery ...
The Last Conformist
Irresistibly Attractive
I can't understand it.
Just why couldn't the AI be taught to handle artillery (and apparently AA units) with the same modicum of competence it handles "normal" units?
What I find completely bizarre is that it will use the bombard capacity of its ships offensively (altho at least in PTW it tends to blast tile improvements rather than units or cities), but fail to do so with land-based artilley.
Just why couldn't the AI be taught to handle artillery (and apparently AA units) with the same modicum of competence it handles "normal" units?
What I find completely bizarre is that it will use the bombard capacity of its ships offensively (altho at least in PTW it tends to blast tile improvements rather than units or cities), but fail to do so with land-based artilley.
Dom Pedro II
Modder For Life
Well, how badly can they screw up the AA anyway? I suppose not using them in aggressive action? I dunno... I haven't played a normal game to that time period yet, but in the WW2 campaign, I didn't use AA aggressively at all. But then, of course, my AA couldn't shoot down a single enemy bomber in 200 attempts, so that doesn't say much for that....
Dom Pedro II,
I just did an experiment that confirms what you says:
I gave England 3 Aegis cruisers (unmodified stats).
I gave Germany 50 Jet Fighters (unmodifed stats).
The result: 2 Aegis cruisers sunk 1 Aegis Cruiser 2 damage points.
2 Jet Fighters shot down. They were shot down on strike 23
and strike 24! The remaining 48 Jet Fighters: No damage!!!
Best Regards
Rocoteh
ozymandias,
As you maybe know I have closed the book with
regard to ACW forever.
I have other projects going on however:
One of them is my own WWIII-scenario: "Russia Demands Respect". It works so well I probably will put it up for
download within 2-3 months.
It will be Conqests only.
To the point: Very good news!
It seems that AI now responds much better to build often
than earlier.
So far its a general impression, but I will start documentation
and will be back later if I can confirm it.
Best Regards
Rocoteh
As you maybe know I have closed the book with
regard to ACW forever.
I have other projects going on however:
One of them is my own WWIII-scenario: "Russia Demands Respect". It works so well I probably will put it up for
download within 2-3 months.
It will be Conqests only.
To the point: Very good news!
It seems that AI now responds much better to build often
than earlier.
So far its a general impression, but I will start documentation
and will be back later if I can confirm it.
Best Regards
Rocoteh
The Last Conformist
Irresistibly Attractive
Nice to see this thread revived. And that sounds like some very good news if "build often" now has visible results.
Hmmm ... Do you think it's worth running some tests to see how well the Build Often etc. flags actually work ... ?
-- Rocoteh, are there any particular experiments which might be of use in designing your current project?
All The Best -
-Oz
ozymandias,
At the moment no special experiments, but I will
keep track of how AI build compared to editor-flags.
As I said before, my impression is that an improvement
really have been made.
BTW, Do you remember the old S.P.I. days and the 3xN:
NATO, NAZIS and NUKES, that interested many people
then. I intended to do a NO NUKES WWIII-scenario.
But I have changed my mind and there will be Nukes
and Settlers.... I mean after a strategic nuclear exchange,
Social-Darwinism probably will rule again.
Best Regards
Rocoteh
At the moment no special experiments, but I will
keep track of how AI build compared to editor-flags.
As I said before, my impression is that an improvement
really have been made.
BTW, Do you remember the old S.P.I. days and the 3xN:
NATO, NAZIS and NUKES, that interested many people
then. I intended to do a NO NUKES WWIII-scenario.
But I have changed my mind and there will be Nukes
and Settlers.... I mean after a strategic nuclear exchange,
Social-Darwinism probably will rule again.
Best Regards
Rocoteh
Hi Rocoteh,
IIRC "Nato, Nukes and Nazis" was eventually made by the Command Magazine people -- If you'd like, I'll try to dig out an ad or a revue (I never had a copy), as they had a slate of what amounted to post-WW2 Nazi "Wonders".
Best,
Oz
IIRC "Nato, Nukes and Nazis" was eventually made by the Command Magazine people -- If you'd like, I'll try to dig out an ad or a revue (I never had a copy), as they had a slate of what amounted to post-WW2 Nazi "Wonders".
Best,
Oz
ozymandias,
Yes, that would be interesting.
The connection to Command Magazine was new to me.
With regard to AI-production I will allocate a large part of my
computer-time to my WWIII-scenario the next days so I
can confirm with stats that AI really have been improved.
Best Regards
Rocoteh
Yes, that would be interesting.
The connection to Command Magazine was new to me.
With regard to AI-production I will allocate a large part of my
computer-time to my WWIII-scenario the next days so I
can confirm with stats that AI really have been improved.
Best Regards
Rocoteh
Procifica
ACW Scenario Creator
3 months later, this STILL is the big question.
I want to point out that with regard to MF, it IS a factor to some extent. Extensive testing of ACW revealed that at one point, the AI would primarily focus on cavalry (despite it having inferior offense AND greatly inferior defense, though it was the only type of unit flagged just offense). Increasing cost and/or decreasing the effectiveness of cavalry reduced its tendency to build it.
I want to point out that with regard to MF, it IS a factor to some extent. Extensive testing of ACW revealed that at one point, the AI would primarily focus on cavalry (despite it having inferior offense AND greatly inferior defense, though it was the only type of unit flagged just offense). Increasing cost and/or decreasing the effectiveness of cavalry reduced its tendency to build it.
The Last Conformist
Irresistibly Attractive
Ozymandias: I'm all for the running of tests. However, I do not have extreme amounts of time available for doing it ATM, and I must track down that non-advancing MGL-creation bug I promised to do before I test anything else.
I'll be going into playtest / debug mode on mine own mod in ~2 weeks, and I'll definitely keep a detailed eye on all 31 Civs, more than a few of which have build often / never build flags checked.
If that looks susepct, then I'll structure some "simple" testing around the "Build Often" flag at the least.
@Procifica -- Yeah, it's still the question ...
.. was there anything like major combat going on during the time that the AI actually emphasized cavalry (major combat losses; major combat far from the production cities etc.).
@Rocoteh -- I'm digging through my entire "paper" game collection over the next few weeks, trying to get some additional closet space out of my apartment (I have a LOT of those games). I'll PM you when I (inevitably!) find the info.
Best To All,
Oz
If that looks susepct, then I'll structure some "simple" testing around the "Build Often" flag at the least.
@Procifica -- Yeah, it's still the question ...
.. was there anything like major combat going on during the time that the AI actually emphasized cavalry (major combat losses; major combat far from the production cities etc.).@Rocoteh -- I'm digging through my entire "paper" game collection over the next few weeks, trying to get some additional closet space out of my apartment (I have a LOT of those games). I'll PM you when I (inevitably!) find the info.
Best To All,
Oz
Procifica
ACW Scenario Creator
I'd have to dig through the 100+ pages of the thread to find the specific examples (which would take hours).
It was a frequent occurance from Rocoteh's playtests until I upped the cost a bit, and reduced its strength (did these bit by bit, a combo of the two worked best).
I do know though that the scenario starts with a small amount of cavalry, and this could have been a reason.
It was a frequent occurance from Rocoteh's playtests until I upped the cost a bit, and reduced its strength (did these bit by bit, a combo of the two worked best).
I do know though that the scenario starts with a small amount of cavalry, and this could have been a reason.
... um, don't do that ... 
*sigh* ... It's distressing to think how hard we worked to (unsuccessfully) discover any major, quantifiable correlations at all ...
So, you think having a unit in play at a mod's beginning acts as a "seed" of sorts? Interesting notion ...
Onward & Upward,
Oz
Procifica
ACW Scenario Creator
No, I'm thinking that the AI felt it needed more cavalry to meet its ratio requirement of O vs. D.
ozymandias,
If I remember right it was ACWRocoteh 3.8 that made
AI really decrease production of Cavalry.
In fact it ceased to produce Cavalry and started to
produce the new Divisions and Non-Division sized
Infantry units. Please check the ACW-thread Post Page 79
Post 1563.
Best Regards
Rocoteh
If I remember right it was ACWRocoteh 3.8 that made
AI really decrease production of Cavalry.
In fact it ceased to produce Cavalry and started to
produce the new Divisions and Non-Division sized
Infantry units. Please check the ACW-thread Post Page 79
Post 1563.
Best Regards
Rocoteh
@Procifica - (*sound of hand slapping forehead*) Ah! You mean the overall force ratio! Of course ... Which does indeed agree with our (1) general results of A vs. D for overall builds and (2) the tendency of the AI -- when presented with a range of units A=D ranging from high A / low D to high D / low A -- to prefer units at the extremes.
@Rocoteh - Man, you guys had some thread going there!
-- Thanks, I'll take a look.
-Oz
@Rocoteh - Man, you guys had some thread going there!
-- Thanks, I'll take a look.-Oz
