AGEOD Forums

Help me understand a few things:

In Setup DB:

1. What does the number immediately after each Objective mean?
   
   
2. How many VP are awarded for each Objective per turn? {I assume each Stratigic city = 1 if controlled}
   
   
3. How can these be used/modified to affect AI behavior?

Thanks!

What does the number immediately after each Objective mean?

The objective value. According to this value an objective is considered minor (<= 2), medium or major (> 5).

How many VP are awarded for each Objective per turn? {I assume each Stratigic city = 1 if controlled}

Strategic City: 1 VP per turn
Objective City: Minor: 1 VP; medium: 2 VP, major: 3 VP

How can these be used/modified to affect AI behavior?

"HE" needs to answer this...

c'est l'homme!

I will be speaking for "him" then

VP and Obj value are very high (the best) incentives to raise the interest of the AI toward a region, but many parameters are taken into account, including all structures and their size, supply level, loyalty, control, number of land links.

That said, you won't convince the AI that she must attack a very very very high value objective if this is too risky. On the contrary if overwhelmed, she will garrison first the most valuable regions.

Last, I would like to propose a command to alter the perceived value of each region, to direct her better. This can be conditional. For example, drawing on a discussion we had with Hok, I would like to have the USA AI considers that Newport has only 10% of its value until Boston is under USA control. Why? Because without maritime superiority, Newport is a deathtrap for the USA, unless they are well established in the area.

That's a smart move Pocus, this lets us teach the AI a basic deep-thinking strategic capability based on realistic and historical strategies.

Pocus:

Does the AI have any "sense of urgency"?

By this I mean: is there logic that tells the AI that there are only x turns remaining? It seems that a 'weighting' variable might be useful that allows increased risk by AI if fewer turns remain. [bonus % reduction to 'risk aversion' logic as remaining turns dwindle]

Sorry if I just hijacked my own thread...

Pocus wrote:I will be speaking for "him" then

VP and Obj value are very high (the best) incentives to raise the interest of the AI toward a region, but many parameters are taken into account, including all structures and their size, supply level, loyalty, control, number of land links.

That said, you won't convince the AI that she must attack a very very very high value objective if this is too risky. On the contrary if overwhelmed, she will garrison first the most valuable regions.

Last, I would like to propose a command to alter the perceived value of each region, to direct her better. This can be conditional. For example, drawing on a discussion we had with Hok, I would like to have the USA AI considers that Newport has only 10% of its value until Boston is under USA control. Why? Because without maritime superiority, Newport is a deathtrap for the USA, unless they are well established in the area.

I have done a lot of experimenting regarding this topic over time. I have constructed scenarios where there was a chain of objective cities (one right after another). The AI would indeed "follow the chain" in an attempt to capture the cities along its path.

I was able to do this with variants of many of the popular shorter scenarios in BOA. It works pretty much every time. No matter what scenario as long as there are plenty of objectives the AI figures out where they are and brings enough force to capture them.

But, when doing these tests (in BOA), I was also able to figure out conclusively that the AI is also very interested in non-objective and non-strategic cities. (i.e.) it would divide a force into smaller components so they could (in some cases) travel fairly long distances to remote non-objective and non-strategic cities. (I DONT mean raiders)

The division of force in many cases would make the AI forces vulnerable to attack and deprive force for armies trying to caputre objectives.

Here is an example of what I mean. (see attachment). Note Henry Knox threatening Montreal. He is capturing non-objective, non-strategic cities that are meaningless in the Saratoga Campaign. Knox is urgently needed in Albany. He does not have any irregulars with him and is not conducting a raid.

Per discussions last year. I still feel that the AI should prioritize as follows.

1. Regions with Objective cities

2. Regions with Strategic cities

3. Regions (non strategic non objective) of highest value

It is the game itself that presents to the user the parameters for victory, which describes the Objective and Strategic cites that MUST be captured (the petit war notwithstanding).

It would be best IMHO, if the AI would also follow the same prioritization that is being described to the user by the game. (i.e. the caputre of Objective and Strategic cities).

Also if I am not mistaken,

1. The petit war is already in progress in any given scenario, never mind the strategic objectives of the regular army.

2. The AI will (and does now) direct its pool of irregular forces to the goals of the petit war.

3. I think we are discussing the intended activties of the regular forces and how best to prioritize there attacks and defenses.

4. If Objective cities cannot be of highest prioritization in a scenario, then WHY describe them as OBJECTIVES.

I'll discuss of that with pocus

lodilefty wrote:Pocus:

Does the AI have any "sense of urgency"?

By this I mean: is there logic that tells the AI that there are only x turns remaining? It seems that a 'weighting' variable might be useful that allows increased risk by AI if fewer turns remain. [bonus % reduction to 'risk aversion' logic as remaining turns dwindle]

Sorry if I just hijacked my own thread...

Some operations are conducted more aggressively when the end turn approach, but this is done on a per algorithm basis. There is no overall increase in aggressiveness, this can be the AI's doom too. You'll be able to test that very soon with the new commands.

Stwa wrote:Per discussions last year. I still feel that the AI should prioritize as follows.

1. Regions with Objective cities

2. Regions with Strategic cities

3. Regions (non strategic non objective) of highest value

It is the game itself that presents to the user the parameters for victory, which describes the Objective and Strategic cites that MUST be captured (the petit war notwithstanding).

It would be best IMHO, if the AI would also follow the same prioritization that is being described to the user by the game. (i.e. the caputre of Objective and Strategic cities).

It is prioritizing, but it splits also the forces when it seems possible to do so. Its hard to find the middle ground between an AI which does something, boldly, but without too much aggressiveness, and a too passive AI just maintaining 90% of its forces in objectives before doing anything else (as in many games, like Civilization etc.).

I can do that easily. Do you want an AI which garrison much all that it has before doing anything else with the spare forces? That is easy, but you'll feel like playing against a passive AI. Because we have historical setups, with forces roughly equivalent (+- 25% I mean) and not games like in Civ where the AI has 3 to 4 time as much forces as you!

Pocus wrote:It is prioritizing, but it splits also the forces when it seems possible to do so. Its hard to find the middle ground between an AI which does something, boldly, but without too much aggressiveness, and a too passive AI just maintaining 90% of its forces in objectives before doing anything else (as in many games, like Civilization etc.).

I can do that easily. Do you want an AI which garrison much all that it has before doing anything else with the spare forces? That is easy, but you'll feel like playing against a passive AI. Because we have historical setups, with forces roughly equivalent (+- 25% I mean) and not games like in Civ where the AI has 3 to 4 time as much forces as you!

YES!, please hurry...

Well, if you can really do it easily then make it an option, becuase now, sometimes I think the AI is dispatching too many "SPARE" forces, and forgetting the objectives of the scenario.

In the example I gave above with Henry Knox. Those regular forces were taken from Ticonderoga and in my view they were not "SPARE" forces. Another person might see it differently.

And don't we already have all the IRREGULARS for the wandering around and all the "fun" stuff. Must we wander around with the Regular Army?

Pocus wrote:It is prioritizing, but it splits also the forces when it seems possible to do so. Its hard to find the middle ground between an AI which does something, boldly, but without too much aggressiveness, and a too passive AI just maintaining 90% of its forces in objectives before doing anything else (as in many games, like Civilization etc.).

I can do that easily. Do you want an AI which garrison much all that it has before doing anything else with the spare forces? That is easy, but you'll feel like playing against a passive AI. Because we have historical setups, with forces roughly equivalent (+- 25% I mean) and not games like in Civ where the AI has 3 to 4 time as much forces as you!

If we had a 'modified' AutoGarrison, it might help.

As I understand it, the current version only garrisons Forts, Depots and larger cities. Something [a new or modified script command] to put a small garrison in level 1 towns might help.

lodilefty wrote:If we had a 'modified' AutoGarrison, it might help.

As I understand it, the current version only garrisons Forts, Depots and larger cities. Something [a new or modified script command] to put a small garrison in level 1 towns might help.

Remember, this game tries to use historical OOB's so creating fantasy garrison units might not be a good idea.

Guys, with all do respect here, I am not really taking about garrisons.

I am talking about what the AI does with its regular forces. I think these should be only concerned with the Objective and Strategic cities.

Of course while the regulars are in-transit to these places they will "capture" non-strategic cities on the way.

The irregulars can be used by the AI for its hair brained missions to St. Louis, Texas, or where ever.

lodilefty wrote:http://www.ageod-forum.com/showthread.php?t=9752

Yes, thank you, things are definately looking up. It will be fun to mess around with this and see what it does....