bdtj1815 wrote:It might be more helpful if you stopped drooling and responded to my points! Am I wrong in my assessment of the AI's abilities, or lack of them?
Taillebois wrote:You are so wrong it is barely worth typing a reply. I suggest you try a simpler AGEOD game like AJE. Ranting angrily in your early posts is not going to endear you to many people and certainly hasn't to me.
Taillebois wrote:You are so wrong it is barely worth typing a reply. I suggest you try a simpler AGEOD game like AJE. Ranting angrily in your early posts is not going to endear you to many people and certainly hasn't to me.
vicberg wrote:Play France as SP
Pocus wrote:The AI is far from perfect, but is not utter crap. You can definitively play it SP. Some scripts might do weird things, like the Caucasus focus. Without scripts on the other hand, France will try to capture Gibraltar. Indeed, a human would probably not do that. As for the British, the AI will always keep a huge force in the homeland.
RebelYell wrote:I think with a combination of AI improvements and scripts the game can be fun to play solitaire.
veji1 wrote:Scripts are time consuming to do but would be the best way to make the AI functional. A game like WITP AE proved very fun to me with the good scripts written by the devs. The problem here is that force composition has always been the main weakness of the AGEOD AI : it can campaign ok, give you a fun game once the force composition is sorted out. But composing the forces or keeping them composed is the main difficulty. This was manageable in games like CWII or AACW (although btw I was mad as hell spring 62 wasn't a scenario in the original CWII release as it was the best to play in AACW).
What this game would need would be a set of scripts that are AI cheat cards, not to get massive ressources extra, but better composed :
- Basically the recruiting tools should be for players only and the AI countries should all be given their forces via faction specific events. Really in this game the AI doesn't need the money, the conscripts, the WS, etc.. this is just cumbersome and makes the recruiting process too difficult, lest the AI buys arty like crazy, so many events for factions giving them new troops regularly based on their historical capabilities.
- Specific force reconstruction events that looks at the situation of AI forces after a peace and gives it a complete new set of forces, organised in armies and corps and such (when the reforms have taken place), otherwise after a drubbing the Austrian or Prussian armies remain messes all game... One has to play them to rebuild their forces properly.
- Lock stacks : the AI shouldn't be able to dismember the Grande Armée corps or equivalent in other armies, lock the stacks to limit the mess the AI can make.
Etc.. All this is feasable but time consuming and would make the game much more challenging, right now the problem isn't that the AI is bad, it's operationally ok which what the customer can legitimately ask for. But it get's completely lost in the diplomatic / force building / force structuring (or here destructuring) operation.. And has a hard time coping with XX sides and logics.
With scripts and a more EAW style limited amount of factions, it would work a lot better.
RebelYell wrote:How much work would it be to make the AI count elements and their types it uses in the main army stacks?
Stack building starts by choosing the leaders, then sticking elements under them based on the rules that decide battles.
Make the AI ignore CP rules before military reforms happen and order as big brigades available, those are the core of future divisions anyway.
veji1 wrote:The problem is that I think the AI can't manage it, it's just to complicated. This is why events with triggers are the best method I would think, but it's very time consuming to write.
lodilefty wrote:There are a few AI Agents assigned, but the real issue is one of Branching Logic.....
the deeper into the game, the more difficult to script the alternative situations....
We did what time allowed by scripting early campaign "paths", where the AI is given "Capture A then B then C", but by the time 1807 arrives the situation has become way too unpredictable......
...maybe somebody better that I can program the numerous branches. Sorry, but it's as good as I can get it with anything resembling a reasonable return on invested time......
lodilefty wrote:There are a few AI Agents assigned, but the real issue is one of Branching Logic.....
the deeper into the game, the more difficult to script the alternative situations....
We did what time allowed by scripting early campaign "paths", where the AI is given "Capture A then B then C", but by the time 1807 arrives the situation has become way too unpredictable......
...maybe somebody better that I can program the numerous branches. Sorry, but it's as good as I can get it with anything resembling a reasonable return on invested time......
RebelYell wrote:There is no reason to, you can make profit by doing the revolutionary wars game and then do Rise of Prussia 2.
AI Leaders Affinity toward stacks
The concept is 'leader affinity toward a given stack'.
Let's say a leader is an Infantry commander. You'll want to seek mostly infantry stacks, not cavalry stacks. You can be more precise even. A leader is a siege expert. He will want to go to stacks with some infantry (+), artillery (++) and siege artillery (++++ !!). A militiaman will want to be among militias...
In the end, it boils down to two things: creating lists and giving a list reference to the leaders.
// #Infantry commander
AIAffinity0 = $famElite|150|$famLine|250|$famCavalry|25|$famMedArty|125|$$famLightInf|150|$famSkirmisher|150
// #Militia commander
AIAffinity1 = $famElite|25|$famLine|50|$famMilitia|400|$famCavalry|25|$famHvyArty|0|$famMedArty|0
// #Siege expert
AIAffinity2 = $famElite|125|$famLine|125|$famMilitia|50|$famCavalry|0|$famSupply|150|$famHvyArty|300
// #Cavalry commander
AIAffinity3 = $famElite|0|$famLine|25|$famMilitia|0|$famCavalry|300|$famSupply|0|$famHvyArty|0|$famMedArty|0
// #Bad commander
AIAffinity4 = $famElite|0|$famLine|25|$famMilitia|50|$famCavalry|0|$famSupply|0|$famHvyArty|0|$famMedArty|0|$famLightInf|0|$famSkirmisher|0
// #Raider
AIAffinity5 = $famRaider|300|$famLightInf|125|$famSkirmisher|125|$famLine|25|$famMilitia|20|$famCavalry|50|$famSupply|0|$famHvyArty|0|$famMedArty|0|$famElite|0
// #Artillerist
AIAffinity6 = $famElite|25|$famLine|50|$famMilitia|25|$famCavalry|0|$famHvyArty|300|$famMedArty|300
0 means not interested. 100 is the norm. Above 100: interested
If a family is not cited, the value is -1. -1 means auto-calculate: if the general is a land commander it means 100 for all land families and 0 for all ships, and the reverse if he is an admiral.
Thus, you don't have to specify for a General that he doesn't want to see ships in his stack.
How do the parameters work?
With these values, when the code checks for the affinity of a leader toward a stack, it will get a consolidated number between 0 and xxx (can be above 100).
100 would means 'average', i.e if you don't add a specific affinity list to a general, as the defaults will be 100, then you are back to the previous code.
Under 100, the leader will have less interest than the norm for the stack. Above 100 ... you guess.
So the feature allows to have generals with several varying interests for different types, and can calculate what will be the average, weighted affinity, of this general toward a stack, even if the stack is a complete mix of units...
Affinity will alter the probabilities, but the algorithm by itself can't match a human mind. Still, just expect in some cases that the choices of how to repartitate leaders will be better with than without.
Don't hesitate to go a bit above the roof. Example, siege artilleries are rare, so even if you have an affinity at 300% for it, as there can be perhaps only 2 such elements in a stack of 100+ elements, this won't change much the affinity! So here, don't hesitate to go to a value of 1000 or even 2000!
For affinities toward more common elements, like cavalry, artillery, you can still target the 300 mark...
The second part of adding affinities is to give to the leaders, in the models DB, what is their affinity list. With the lists created, you will do these aliases (in WIA, found in AI_Interest.ini in the /Aliases folder):
$AIAff0_inf = 0
$AIAff1_mil = 1
$AIAff2_siege = 2
$AIAff3_cav = 3
$AIAff4_bad = 4
$AIAff5_Raid = 5
$AIAff6_Arty = 6
In the models.xls file, add (or edit) the column AIAffinity, leave the value as NULL if there is no special ability, or add one (and only one) of the aliases if the leader has a particular affinity for a role. Then extract your new models (using the csv -> csv Splitter method)
Kev_uk wrote:Well one thing that concerns me is the Steam thing, whereby you are going to get folk who will quite easily tear it to pieces, none of whom understand what it is about, the history, and Ageods past titles and what they are trying to achieve. I believe WoN has potential to be better than its current state, instead of us all playing PBEM (which the Steam crowd will not, really will they?). I will probably buy it again on Steam when it released anyway.
Pocus wrote:The AI is far from perfect, but is not utter crap. You can definitively play it SP. Some scripts might do weird things, like the Caucasus focus. Without scripts on the other hand, France will try to capture Gibraltar. Indeed, a human would probably not do that. As for the British, the AI will always keep a huge force in the homeland.
Users browsing this forum: No registered users and 5 guests