AI Magical Mystery Tech?

[Cort Haus]

If anyone can explain to me how the Mali manage to research Polytheism when they do in the attached test savegame, I'd be very grateful.

It's vanilla 1.61, OCC, Prince, Small Pangea with 7 civs, normal speed, 3480BC.

Load the game, hit enter 3 times and a distant civ founds Hinduism in 3360BC. It's actually the Mali.

Then reload it, open Worldbuilder, and put a spy in Timbuktu to the northeast, and while you're there, count the number of huts left on the continent (there are zero).

Return to the game, press enter, and investigate the city. They are researching hunting, which completes when the human player founds Hinduism a few turns later.

So first, how do they manage to complete Polytheism in the non-worldbuilder test when there are no huts available, and they are researching something else, and second - why do they not do so when inspected using a spy from worldbuilder?

[Dominae]

It's Heisenberg's Uncertainty Principle at work, obviously.

[Cort Haus]

Yes, the act of observing seems to alter the behaviour of the subject.

It would be cool if someone like Trip or Alexman could have a look at this. It seems either

(a) I've overlooked something but I can't think what, or
(b) It's a bug, or
(c) It's a cheat (NOOOO, surely not!!).
{if SPY_PRESENT(human_player) then TECH_UP_SLEEVE(keep_hidden)};

I'd certainly call it an anomoly.

[Peter Triggs]

I loaded the game, pressed < ctrl + z > to enter debug mode, and saw that Mansa was due to complete researching Polytheism in three turns.

[Cort Haus]

Originally posted by Peter Triggs
I loaded the game, pressed < ctrl + z > to enter debug mode, and saw that Mansa was due to complete researching Polytheism in three turns.

Ah, debug mode. I haven't used that. So why is he researching Hunting when a spy is wordbuildered in?

[phlucas]

This is interesting. It is not important that the unit is a spy. I gave myself a wolf and set him into Timbuktu. He got kicked out on the next turn, but nevertheless MM switched away from Polytheism. Probably he sensed a hostile unit nearby and therefore went to Hunting. This makes sense, sort of (but he takes Animal Husbandry afterwards, and not Archery -- Hmm?). But this begs the question of why the presence of a spy can influence the decisions as well as the presence of a visible unit, of course.

[Prussia]

It might be a bug that most people wouldn't realize. Or maybe there are conflicting orders for AI and since the game isn't expecting spies that early, it might screw some things around.

[Dis]

it shouldn't know there is a spy there, right?

[alexman]

OK, here's a little-known secret. The AI behavior changes to become more defensive when they meet a human.

The AI does not know if you have a spy there, but even so, contact occurs.

[Cort Haus]

Thanks, alexman, that explains it.

[Nacht]

So basicly the AI panics when it meets a human player and screws up, like dropping research on a religion founding tech three turns from completion.

Wouldn't it be a good idea to let the AI never change research? The subtilities involved in deciding to change research halfway are really beyond the AI. It seems that just a rule of never changing would be more benificial for the AI.

[rah]

I don't know about that. If you trade a tech with the AI it may open up a research path to a coveted wonder that would now have priority. Why should the AI be locked in?

[Snotty]

I know, as Im sure the AI does, that the human player can be violent and unpredictable. Starting defence research asap after meeting a human seems a wise move to me

[Cort Haus]

Possible religious marathon gambit for lower levels here - with a starting scout on a small map, send it far and wide asap to meet the other civs to try and knock them off their Polytheism track, to allow monopolisation of early religion. There'd be a side effect of a more powerful AI military.

Opposite gambit - stay at home and try to avoid everyone for as long as possible to delay their archer build-up and let them all race for religions.

[Prussia]

Originally posted by Snotty
I know, as Im sure the AI does, that the human player can be violent and unpredictable. Starting defence research asap after meeting a human seems a wise move to me

The fact that you know this means the AI is TOO predictable.