Who knows the Answer?

kobayashi · September 2, 2003, 09:05

Seems the forum search function is disabled. Looks like I have to seek the aid of the collective wisdom here regarding bribing.

Besides distance from capital, do any other factors affect unit bribing? How about City bribing (Besides Courthouse and presence of spy/diplomat) ?

Veteran - seems no.
Defense Bonus
Fortified status
Type of government (other than democracy)

How about how bribing costs are calculated? WHat are the factors?

Gold
Cost of Improvements in a city
Cost of Improvements in all cities
Cost of units in a city
Cost of units in all cities

Would also be nice to know under what circumstances the AI likes to bribe?

p.s.
If the AI has the choice either to irrigate of build mines, does it alway irrigate. Do the goverment thresholds have any influence?

p.p.s
I sorta noticed that the AI doesn't attack units for which the odds of winning are bad. Does anyone know (roughly) the algorithm for determinng if the AI will attack.

Elok · September 2, 2003, 12:49

Well, cost of the unit, for one thing. In a lot of scenarios, unique units are given a cost of 99 so they can't be bribed without spending your entire treasury.

techumseh · September 2, 2003, 13:24

I'm pretty sure that city cost is affected by the technology gap between briber and bribee. If a civ is more advanced, it's cities are quite a bit more expensive.

William Keenan · September 2, 2003, 14:37

Unit Bribe
(37.25 / Distance + 2) * (1 + Treasury / 750) * Shields = Cost to Bribe

Distance refers to the distance from the capital. Normally barbarians have no capital so Distance is equal to 16. However, if the bribing civ operates under a communist government the Distance reduced to 10.
Treasury is the amount of gold in the Treasury.
Shields is the number of shields required to build the unit.

Double the Cost to Bribe if the unit is a settler type (role=5).

Immunity to Bribery is granted to any civilization, including the Barbarians, that operate under a Democratic government.

City Bribe

Cost to Incite Revolt = City Size * Wealth of Civ * Proximity to Capital

City Size = from 0 to 35.
Wealth of Civ = 1 + Treasury/1000. The revolting civ's treasury.
Proximity to Capital = 1000/(Distance+3). Where distance is the number of squares between dip/spy and the capital of the revolting civ. Distance is never greater then 16 and subject to several distance modifiers.
The Courthouse improvement halves the Distance (rounded down).
When the bribing civilization operates under Communism the maximum Distance is reduced to ten.

The Cost to Incite Revolt is adjusted by the following cumulative modifiers.

A Spy incites revolts at 5/6 the price of a diplomat.
A Veteran Spy incites revolts at 2/3 the price of a diplomat.
Disorder in the revolting city reduces the price by 1/2.
No Units in the revolting city also reduces the price by 1/2.
The Previous Owner of a city can incite a revolt at 1/2 price.
Operating under Democracy grants civilization immunity to bribery.
Palace improvement makes the city unbribable.

SlowThinker · September 2, 2003, 15:17

I suppose William answered comprehensively, but if you will miss something...
In Civ2-Strategy forum we have a thing named Great Library (link is in my signature). I am sure there are not only threads about combat system

but also about dips and spies...

Quote:

Would also be nice to know under what circumstances the AI likes to bribe?

In deity AI bribes cities quite often - if it has enough of gold.
In deity also non-dip AI units may bribe units.
But this stuff might be testable...

Quote:

I sorta noticed that the AI doesn't attack units for which the odds of winning are bad. Does anyone know (roughly) the algorithm for determinng if the AI will attack.

I think that Civ2 designers didn't care about combat system very much and so I suppose there will be some defects in evaluation of combat odds. And those defects might be used for some suicide attacks or too timorous behaviour of AI.
Try units with a non-standard ratio AF(or DF)/HP/FP.

Heresson · September 2, 2003, 18:53

You know... Once my sis was playing civ.And something strange was happening. She was inviding other civs with musketeers and catapults against tanks and howitzers...
And the AI simply didn't attack her! It was incredible. They were kind of running away from her, no matter what were her units. I never seen it again. Either my sis has paranormal abilities, or there was something wrong with my civ, which is pretty possible. There was something wrong with the computer at the time.

Peaster · September 2, 2003, 19:07

Quote:

Originally posted by kobayashi

p.p.s
I sorta noticed that the AI doesn't attack units for which the odds of winning are bad. Does anyone know (roughly) the algorithm for determinng if the AI will attack.

Don't know, but I think the algorithm must be different in MGE and FW. There's at least one FW scenario that doesn't work well with MGE because of this.

Boco · September 2, 2003, 22:27

I noticed that role 0 units are used much more aggressively when they are plentiful. There's a situation in El Aurens where changing the cost of a unit from 80 to 100 dramatically reduces the aggressiveness of the AI for this unit. I don't think the cost directly affects how aggressively the AI uses the unit, but rather its abundance. Later in the game, when it's in short supply, the AI is a little more reluctant to throw it away.

In short, I suspect that the AI is much more willing to attack with a unit when there are a few more in reserve. If it has a dozen units available, it's more willing to try a Banzai than when there's only 1-2.

Role 3 units have clear modifiers that affect their targeting. Favored defenders include air units and sea units (at least in port). It will also consider the cost of the defender. I know of an example where a unit costing 10 is consistently overlooked by role 3's. When I raised its cost to 80, the AI immediately strafed it. Odds for victory were somewhere around 40%.

techumseh · September 2, 2003, 23:58

Welcome back, William!

Elok · September 3, 2003, 09:54

Whoa...the William Keenan of "barbarian paper" fame? WB indeed...

William Keenan · September 3, 2003, 10:38

Quote:

Originally posted by techumseh
Welcome back, William!

Thanks, tech. Life is so busy these days, leisure hours are hard to come by.

Quote:

[SIZE=1] I sorta noticed that the AI doesn't attack units for which the odds of winning are bad. Does anyone know (roughly) the algorithm for determining if the AI will attack.

Capt. Nemo did some substantive research on this topic, I recall. He determined that the AI will only fire missiles at units with an acceptable target cost to missile cost ratio. It’s logical to expect that a similar ratio based formula is used to determine the aggressiveness of other unit types as well.

Boco makes some good points. Surplus of AI units, Role, and odds of winning are undoubtedly factors also.

Valuk · September 5, 2003, 10:43

The answer is 42. Haven't you read Doouglas Adams?

William Keenan · September 5, 2003, 11:22

Quote:

Originally posted by Valuk
The answer is 42. Haven't you read Doouglas Adams?

Yes, but, what is the question to the answer of the meaning of life?

SlowThinker · September 15, 2003, 14:20

Quote:

Originally posted by kobayashi
I sorta noticed that the AI doesn't attack units for which the odds of winning are bad. Does anyone know (roughly) the algorithm for determinng if the AI will attack.

I ran some tests that indicate how Civ2 engine evaluates defense strength...
http://apolyton.net/forums/showthrea...16#post2337865
...but I also found out the willingness to attack don't follow this algorithm, so another tests would be needed.

kobayashi · October 10, 2003, 10:22

repeat request

If the AI has the choice either to irrigate or build mines (on the same spot), does it alway irrigate? Do the goverment thresholds have any influence?

Panda · October 10, 2003, 19:06

*usually* irrigate, but I suspect you could modify the AI behaviour. For example, make mining more/less rewarding, or set mining to "never" in the govt prereq if you wanted to disable mning altogether.

This is just my opinion so far, and not backed by empirical data *apart* from the observation from a recent scenario of mine that the AI loves irrigating.

What are you getting at here?

kobayashi · October 11, 2003, 10:32

I am toying with the idea of having every large patches of the same terrain. This means they must all be irrigatble as well as minable. But if the AI doesn't mine then the idea doesn't work.

SlowThinker · October 11, 2003, 12:04

Quote:

Originally posted by kobayashi
I am toying with the idea of having every large patches of the same terrain.

What do you mean? Same amount of bonusor same amount of time to Irrigate/Mine?

Mercator · October 11, 2003, 13:25

Well, the AI can certainly mine... Hills can both be mined and irrigated, but the AI mostly (only?) mines them.

The only problem could be that the AI always picks the same option for a terrain type, rather than what would suit its needs most.

That is, it might mine one terrain and irrigate another, but it will always choose the same action for a given terrain type.

I'm not entirely sure about that, though. Maybe it actually makes a somewhat intelligent choice, mining some squares and irrigating others.