In response to the sabotage thing:

[La Fayette]

Here we are!
To make it clear I send 3 posts:
1) Latest results (attacking the ancient and glorious city of Thebes, capital of the Egyptians, with hordes of babylonian spies).
2) How it works (my conclusion about what is inside the box: namely how the programmers of civ2 decided that sabotaging would be).
3) Advice to players (what should one do when one wishes to attack a city with dips or spies).

------------------
aux bords mystérieux du monde occidental

[La Fayette]

Post #1: Latest results
a) Attack of Thebes (capital) by 7 babylonian non vet spies ("choose primary target": city walls)

Lost O = 1
Lost 1 = 7
Lost 2 = 3
Lost 4 = 5
Lost 5 = 0
Lost 6 = 1
Lost 7 = 5
Total = 25
This means 83 spies lost in 20 successful attacks + 5 unsuccessful attacks (due to the fact that I had only 7 spies ready to attack, since I always make tests from savegames and have never opened "cheat mode " up to now).

b) Same with 6 vet spies (capital, walls)

Lost 0 = 8
Lost 1 = 11
Lost 2 = 4
Lost 3 = 1
Lost 4 = 0
Lost 5 = 0
Lost 6 = 1
Total = 25

This means 28 vet spies lost in 24 successful attacks (+1 unsuccessful)).

[La Fayette]

Post #2: How it works

a)"choose primary target"
"Pile ou face", that is the name of the game in french (I think it is called "heads and tails" in english: you throw a coin and look at what happens, either or).
Look at the example above, if needed: my vet spy stands in front of the walls of Thèbes, then
"pile": the walls tumble down
"face": she's captured by wary Egyptians.

In more scientific terms, there is a probability of 0,5 that the walls of a capital tumble down when attacked by a vet spy. Any new attack is independant from the previous one (just like in the game of "pile ou face" any new coin is independant from the one before, which explains why it might happen to you that, once in a while,a large number of vet spies are captured before the walls tumble down, just like when one goes on playing "red" at the roulette).
Then the Egyptians get a second chance. They may capture the vet spy after the walls have been destroyed. This means a second game of heads and tails (with a probability close to 0,3 that the vet spy gets captured, or 0,7 that she escapes unharmed).

b) What happens if the spy is non vet, or the city is non capital or the improvement is non walls (temple or library,...)?

The program is the same:
test #1: do the walls go down? (yes or no, probability of yes=p1)
test #2 (conditional): if the walls are down, does the spy escape (yes or no, probability of yes=p2).
But the probabilities are different (most risky:capital, walls, non vet spy; least risky: non capital, non walls, vet spy).
From the tests made by SlowThinker and by myself, I derive that the probabilities are probably close to these:
1) Capital, walls, vet spy: p1=0,5 (p2=0,5)
2) Capital, walls, non vet: p1=0,25 (p2=0,25)
3) Capital, non walls, vet spy: p1=0,66 (p2=0,5)
4) Capital, non walls, non vet: p1=0,5 (p2=0,25)
5) Non capital, walls, vet spies: p1=0,6 (p2=0,5)
6) Non capital, walls, non vet: p1=0,4 (p2=0,25)
7) Non capital, non walls, vet spy: p1=0,85 (p2=0,5)
8) Non capital, non walls, non vet: p1=0,7 (p2=0,25)

c) "use her judgment" (or dip, since dips have no other choice)
The method for choosing is the same for dips and spies:
They choose their target almost in accordance with the list of improvements in the city window (ST has also tested some sort of priority for some modern improvements).
But there is a random factor which makes that, even if city walls are down the list, it may happen now and then that they are chosen as #1 or #2.
It seems to me that the improvement chosen is always destroyed and the commodity under building is chosen only if there is at least 1 shield in the box (which means only one such attack per turn).
The difference between dips and spies lies in the field of losses:
100% losses with dips (neither ST nor I are wishing to test vet dips now).
About 50% losses with non vet spies and 30% with vet spies (probably higher losses in capital cities; I test that to-night if I feel like it).

[La Fayette]

Post #3: Advice to players
This advice is about sabotaging (destroying improvements, especially city walls), not about bribing, stealing techs or poisoning water.

Spies (vet or non vet) have a choice between 2 methods: "use her judgment" or "choose primary target". Dips always use their judgment.

a) "use her judment"
If you want to make sure that you destroy the walls of the city you are greedily looking at, you should :
- send 1 dip and have a look at the number of improvements inside the city, let us say n;
- then come back next turn with n+1 dips and destroy everything until the walls tumble down (if you are very lucky, dip #1 might destroy the walls, but most commonly you will have to wait until dip #n or dip #n+1).
The method is the same with spies, but losses are lighter (100% with dips, 50% with non vet spies, about 28% with vet spies).

b)"choose primary target"
The 2 main advantages of this method are lighter losses (especially with vet spies) and non destruction of improvements inside the city (which allows you to sell them or use them when the city becomes yours).

NON VET
If the city has few improvements and you are using non vet spies, then the "choose primary target" is not really a good choice, because there is a risk of having many spies captured before the walls tumble down.
If you are in a hurry and don't care too much about the value of the improvements, then let your spy "use her judgment" (with n+1 spies ready, of course).

VET
It is highly advisable to use method #2 ("choose...") with vet spies, with lower losses and better chances of success.

[La Fayette]

La cerise sur le gâteau.
This is a gift to those of you who are fond of matrices.
Matrix #1 gives the average number of vet spies lost/walls destroyed, along with the tests made by ST and me.
Matrix #2 gives the same for non vet (column 1 relates to walls, column 2 relates to other improvements).

1) VET

Capital city: 1,38; 0,85
Non capital : 1,09; 0,70

2) NON VET

Capital city: 3,33; 1,95
Non capital : 1,92; 1,17

My conclusion would be (as usual): long live our beloved vet spies.

------------------
aux bords mystérieux du monde occidental

[La Fayette]

Dear friends
I know that some of you happen to write "sheilds".
I know that some of you happen to write "diety".
That is your choice.
I didn't choose to write "independAnt" (just as if I were writing french).
I blush and beg your pardon.

[SlowThinker]

I agree with "heads and tails".

I think we would have proceed 5 times more tests to have appropriate results. About p2, for example: you set the probability to 0.25 and 0.5 too prematurely. My results were 0.57, 0.31, 0.18, 0.43, 0.66 0.25, 0.23

About p1: I don't believe programmers set all 8 numbers separately. I think more probable is the possibility that they set basic probability and three modifiers for walls, capital and vet status. (Maybe they set that capital and walls won't be cumulative)

Look at this.
1) Capital, walls, vet spy: p1=0,5
5) Non capital, walls, vet spies: p1=0,6

4) Capital, non walls, non vet: p1=0,5
8) Non capital, non walls, non vet: p1=0,7

Non Capital induces increase to 0.6 for the first case and to to 0.7 for the second case.


[This message has been edited by SlowThinker (edited January 27, 2001).]

[SlowThinker]

I dislike that testing of probabilities. It is too annoying. I think authors of Civ should release all informations about probabilities.

[SlowThinker]

quote: Originally posted by La Fayette on 01-30-2001 12:12 PMWe know exactly how it works

I would disagree with "exactly"
But I would let the remaining work for other people

quote: Thank you for constructive dialog.

I liked it.

[La Fayette]

ST
I agree with you: it would be advisable to test much bigger samples to make sure of the exact values of those probabilities (and get bored to death doing that ).

We know exactly how it works and what to do when attacking any city with dips or spies (and I am quite satisfied with the summary of losses that I published above).
Hence I put the main results on your thread and consider that we have solved the problem of sabotaging (all important results have been given and remaining details have no practical importance).
Thank you for constructive dialog.

------------------
aux bords mystérieux du monde occidental