Log in

View Full Version : Strat Map AI Cheating



Inferno
01-08-2003, 04:04
One of the developers (I am 90% sure it was eat cold steel) came onto this forum and announced that the strat map AI does not get to see the moves you make before it makes it moves. In other words, it sees the map as it was before you moved any of your units when making decisions.

Well, I had an army of 1200 men in Aquitane and I decided to move them all into Navarre, relying on my alliance with the Argonese to protect that province while I moved down some reinforcements.

Next turn, the Argonese break the alliance by moving 101 men into Aquitane.

I have NEVER, EVER seen the computer launch an attack 101 vs 1200 in my favour. The AI was on Hard, by the way.

There are only two explanations:

1) The AI is crap (launched an attack that it was so obviously going to lose)

2) The AI cheated and knew I had moved those troops, even though a developer had come onto this forum and expressly denied the computer does this.

Sucks to be lied to.

LadyAnn
01-08-2003, 04:19
I have an easy explanation.

Your move was part of last year. Computer's move was part of this year. Computer see you left a province unguarded this year, it moves in.

So, nobody lies, although it is not quite realistic.

Annie

Inferno
01-08-2003, 04:23
Anne,

Unless I am very much mistaken, all moves SHOULD happen simultaneously.

I see what you are saying, but surely it all boils down to the same thing: I moved my units and then the computer made up its mind on what to do after it had seen me move my units. This is either grossly bad programming (unlikely) or cheating. There is no other way of describing it.

I moved my units, hit End Turn, then the Argonese IMMEDIATELY moved into Aquitane. My move and his move were made in the same year.

Perhaps I didn't make myself clear enough. http://www.totalwar.org/forum/non-cgi/emoticons/smile.gif

Lord Romulous
01-08-2003, 04:26
lol this is depressingly funny.

in other words lady ann.

we cheat, the player

we see the ai's move and then get to counter.


hmmmm it could be true i guess. i hope not.

so there are so far 3 explanations.

1. ai cheats
2. player cheats
3. ai is stupid or eratic.

hmmmm i dont like any of those explanations. anymore ???

Inferno
01-08-2003, 04:30
Quote[/b] (Lord Romulous @ Jan. 07 2003,21:26)]lol this is depressingly funny.

in other words lady ann.

we cheat, the player

we see the ai's move and then get to counter.


hmmmm it could be true i guess. i hope not.

so there are so far 3 explanations.

1. ai cheats
2. player cheats
3. ai is stupid or eratic.

hmmmm i dont like any of those explanations. anymore ???
Explanation 2 doesn't make sense...you make your moves before the AI does.

Lord Romulous
01-08-2003, 04:33
edit. disregard still thinking.

Kraxis
01-08-2003, 04:33
No, we are the first... We move our troops and the AI see them, but the AI does not take advantage of that like in STW.

But if the Aragonese took Aquitane from you, then it did not cheat as if you attack and the AI attacks the province they come from your army is halted and moved to defend the province.

Inferno
01-08-2003, 04:37
Grrrr.

You are missing the point.

I had 1200 men.

I moved them out of Aquitane into Navarre (held by me).

The AI attacked Aquitane on the same turn with 101 men.

The only way the AI could POSSIBLY have thought about winning that battle (101 V 1200) is if it KNEW I had moved my men out

Lord Romulous
01-08-2003, 04:40
what i was trying to say i think

is according to annes theroy.

IN year 1.
the ai moves.
we see the ai's move.
then the player moves.

then year 2 happens.
the ai sees your move in year 1.
and counters.
you then see ai move and
then counter.

this sounds really bad and i hope it is not true.

annes theory = player and ai moves are not simultanous

solypsist
01-08-2003, 05:05
you as the human player are always given the option to contue onwards with your attack on an enemy province, or to cancel the invasion. if you cancel your attack, your men are returned to their launching province, so the Arogonese are now faced with fighting your larger army.
yes, it seems the ai gets to "counter-attack" at best opportunity, but you can easily react by simpy changing your mind and calling your men back, since the human player always goes first on those decisions.

Inferno
01-08-2003, 05:18
Soly,

I moved my troops into a region I controlled. Consequently, I do not get the option to move them back.

Sorry to be an #######, but did you actually read any of my posts?

Hakonarson
01-08-2003, 05:23
Sometimes a coincidence is just a coincidence - given the thousands and thousands of moves made in MTW by people on this forum do you think that the few occasions where this behaviour is reported really constitutes the AI cheating?

If it was cheating then I would expect to see it happening systematically - daily.

Not far away from this thread is another one about Stupid AI moves.

The answer is simply that the AI is stupid.

Yes it was invading your province expecting to fight 1200 men with 101.

It happens.....too often really.

you were just unlucky in that you chose to move out when you did.

andrewt
01-08-2003, 05:50
Agreed with Hakonarson. The AI sometimes attacks against impossible odds. Chances are, as I've read in another thread, royal knights were part of that 101 troops the AI send against you. Guessing from the number, the king is probably involved. I think it overvalues them in a fight. I've seen the AI attack another AI faction with the same impossible odds.

Jacque Schtrapp
01-08-2003, 06:45
Ok, first let me say that I have been attacked by the AI facing impossible odds, especially the Aragonese. Secondly let me state that I have spent a great deal of time reloading save games in order to determine whether or not the AI does in fact react to player moves that are executed during the same year.

If the human player makes a move on the campaign map and ends the turn, he/she can witness that AI moves for the same year before the florins/year are updated in the bottom right of the campaign map. These AI moves are supposed to be based on the AI's strategy regarding the humans movements from the previous year. Yet I have found save after reloaded save that (especially in the latter stages of the game) when I move a large army out of a province that I possess which shares a border with either an enemy/neutral/ally and leave that province virtually undefended I will be attacked. However, if I reload and do not remove the army from its province I will not be attacked. I have even gone so far as to reload the same turn numerous times adjusting the quantity and composition of my troops and have discovered that when the AI achieves a numerical or qualitative superiority it will attack.

Now those of you who claim that the AI "cannot see" the players movements that take place during the same calendar year, please explain how this is true when I have proved beyond the shadow of a doubt that the AI does in fact respond in 1276 to moves that I intentionally orchestrated in 1276.

Same holds true in naval situations. Share a sea zone with a foe, where the number of ships is in your favor and he will not attack. However, if you move enough ships out of the sea zone so that the numbers favor the AI, then he will attack the remaining ship(s) the same turn that you are moving your ships to a new sea zone. Reload and leave the ships and he does not attack.

The AI does see and react to player movements if it is in its best interest to do so. Still, all in all, its not nearly as prevalent or asinine as STW.

solypsist
01-08-2003, 08:13
Quote[/b] ]Well, I had an army of 1200 men in Aquitane and I decided to move them all into Navarre

Hmmmm...It doesn't say anywhere in there if Navarre was friendly or not; I just assumed you were attacking (why else move all your men into a province?)
So I did read the first post, but missed the second one. Sorry.

pdoan8
01-08-2003, 08:51
I don't think that AI cheat. AI is rather stupid in this case. Or as in a post I read somewhere, they are overconfidence with their royal units or they are too sneaky and doesn't think about the consequence of their action.

- I had seen an AI army consists of some royal knights and a couple of spearmen attack a well defended province with close to 2000 garrison troops. In my current campaign, the Danish break the peace treaty and attack Sweden which was mine for 4-5 years. Their force consists of 8 royal knights (including the King), 2 Spearmen and 2 Militia Sergeants. My force was about 2000 strong consisting of my main army (4 Byz Inf, 4 VG, 2 Kat, 2 Pronoiai, 3 Byz Cav and the commander) and the reserve (4 Treb Archers, 2 Passive Arb, 2 VG and 2 Byz Inf).
- AI (particularly the Italian and the Sicilian) frequently attacks a lone ship I just moved in a sea region which I also had another large fleet station there.

I wouldn't be surprise if those 101 men were the King (or a Prince) and his 100 brave royal knights.

andrewt
01-08-2003, 09:03
I've had the AI attack a province with 100 units when I had 300. The enemy had Ghulam bodyguards (prince) and Armenian Heavy while I had spear and byz inf. I also had instances of the AI attacking a province during a turn I had just reinforced it and they were outnumbered by a really large ratio by superior troops.

HopAlongBunny
01-08-2003, 09:04
Does the AI cheat?

If not it certainly does take advantage of some weird things. Present game, I own Spain less Navarre Aragon Granada and Valencia. The French, w/o benefit of the fleets in the right places, moves troops from Navarre to Granada almost every turn. How does it do this?

Once I sacked and looted all of what would be modern-day Turkey; spent no longer than 2 turns in any province and burned everything to the ground; by the time I was back at my start point the first province was practically rebuilt. Odd? Yes http://www.totalwar.org/forum/non-cgi/emoticons/smile.gif

zipnar
01-08-2003, 11:10
when the ai attacks it tries to match what it attacks with to what you have to defend with, in order to give you a "fair" fight. that is why you often see 5000 peasants vs your 1000 good army.

i've played numerous games of MTW, and this is very very clear to me and i have no doubt whatsoever that this occurs.
it is also the reason i quit playing this game.

the ai does NOT try to win, it only tries to give you "good" fights. but since the tactical ai cannot beat a human with equal armies...it will always lose...and therefor the strategic ai will always lose.

you can test this...replay a turn where the ai is going to attack...and put varying amounts of units in the province to bee attacked...you will notice that the ai attacks with a force that coresponds to yours. lame.

the ai doesn't try to win.

Vrashk
01-08-2003, 15:15
I must agree with the ones claiming that the Ai cheats or at least makes it feel like it is cheating. I've seen it happen far too often (scenarios similar to the examples mentioned) and to put it simply:

The Ai DOES react to your moves the same year.

But it probably doesnt cheat, its just an unfair game system. It feels like it cheats since the player always go first.


I believe the campaign turns do not happen at the same time.

The human player moves first
then player 2 (And he sees what player 1 did)
Then player 3 ( And he sees what the first players did)

And when everyone has moved, the battles are resolved
(This is why the AI always decide to move in troops in his attacked provinces and attack undefended provinces etc etc)

It is the only possible explanation except cheating that I can see.

Strategery
01-08-2003, 19:14
I have sometimes felt that the AI does seem to react to your moves. OTOH I have also seen the AI frequently make hopeless attacks against well defended provinces, especially to initiate wars.

It hardly matters to me whether the AI is cheating or not, it is awfully beatable in any case. Maybe it should cheat.

barech
01-08-2003, 20:42
Hey I don't know what the explanation is but I had AI cross from Denmark to Sweden to attack me after I removed the land bridge and IT did not have any ships. Hmmmmmmm

Cadarn
01-08-2003, 21:25
Quote[/b] (zipnar @ Jan. 08 2003,04:10)]when the ai attacks it tries to match what it attacks with to what you have to defend with, in order to give you a "fair" fight. that is why you often see 5000 peasants vs your 1000 good army.

i've played numerous games of MTW, and this is very very clear to me and i have no doubt whatsoever that this occurs.
it is also the reason i quit playing this game.

the ai does NOT try to win, it only tries to give you "good" fights. but since the tactical ai cannot beat a human with equal armies...it will always lose...and therefor the strategic ai will always lose.

you can test this...replay a turn where the ai is going to attack...and put varying amounts of units in the province to bee attacked...you will notice that the ai attacks with a force that coresponds to yours. lame.

the ai doesn't try to win.
Whoa Try telling that to the AI Byz who have just launched a multi-province invasion of my Polish empire after decades of peace. Not play to win ?? My garrisons and reserve armies in the area were swept off the board.

A superb play by the AI in this case, just the sort of co-ordinated empire-breaking attack us players love cooking up.

I'm fighting a rearguard action with what's left but they've hit most of my main troop factories in Hungary, Carpathia, Kiev and Bulgaria. Great work.......

jrexchandler
01-08-2003, 22:05
Sometimes the AI makes bad moves. A few days ago, the entire Danish royal family (king and 5 heirs) attacked (120 royal knights) Sweden, which I had defended with 1800+ troops, royal knights, feudal men-at-arms and seargents, and spearmen.

Sometimes, just like humans, the AI makes mistakes.

Also, sometimes the AI makes good moves. For the first time ever, I was completely outflanked by the AI last night. They were able to beat my force with quite a few less troops.

And, so what if they did cheat? We get to see their moves before we move. Also, since they broke the alliance, their influence will go down. And, since you have such a large force, just move them back and destroy their faction.

Problem solved, and offenders punished.

Rex

Hakonarson
01-08-2003, 22:22
Quote[/b] (Jacque Schtrapp @ Jan. 07 2003,23:45)]If the human player makes a move on the campaign map and ends the turn, he/she can witness that AI moves for the same year before the florins/year are updated in the bottom right of the campaign map. These AI moves are supposed to be based on the AI's strategy regarding the humans movements from the previous year. Yet I have found save after reloaded save that (especially in the latter stages of the game) when I move a large army out of a province that I possess which shares a border with either an enemy/neutral/ally and leave that province virtually undefended I will be attacked. However, if I reload and do not remove the army from its province I will not be attacked. I have even gone so far as to reload the same turn numerous times adjusting the quantity and composition of my troops and have discovered that when the AI achieves a numerical or qualitative superiority it will attack.

Now those of you who claim that the AI "cannot see" the players movements that take place during the same calendar year, please explain how this is true when I have proved beyond the shadow of a doubt that the AI does in fact respond in 1276 to moves that I intentionally orchestrated in 1276.

Same holds true in naval situations. Share a sea zone with a foe, where the number of ships is in your favor and he will not attack. However, if you move enough ships out of the sea zone so that the numbers favor the AI, then he will attack the remaining ship(s) the same turn that you are moving your ships to a new sea zone. Reload and leave the ships and he does not attack.

The AI does see and react to player movements if it is in its best interest to do so. Still, all in all, its not nearly as prevalent or asinine as STW.
You haven't "proved" anything of the sort - indeed you seem to have an odd understanding of the word "prove".

Moreover no-one has to provide any explaination as to why the AI makes teh moves it does - it is you making the accusation, therefore it is YOU who has to provide all teh information to justify your own position, not everyone else who has to show that your accusation is wrong.

Unless you go to the code and show how it's happening all you have showen is that in some individual cases the AI has made those moves - yuo have shown nothing at all about why it has made those moves.

Unfortunately I don't expect you to accept this - howls of "justified" outrage rarely answer to the requirements of logic or objective evidence because they are so deeply based in emotional terms.

You are not going to get any joy from your demands that the programmers admit that the AI cheats - indeed as yuo said at the start they have quite explicitly said that the AI does not cheat.

Exile
01-08-2003, 23:13
Quote[/b] (Strategery @ Jan. 08 2003,12:14)]I have sometimes felt that the AI does seem to react to your moves. OTOH I have also seen the AI frequently make hopeless attacks against well defended provinces, especially to initiate wars.

It hardly matters to me whether the AI is cheating or not, it is awfully beatable in any case. Maybe it should cheat.
This echoes my thoughts on the matter. I've seen both things happen, At times the AI seems psychic at other times it acts like it is psychotic/suicidal.

In the end its tipped in favor of the player. The failings of the AI more than make up for any advantages it has. Play the game and enjoy, just remember to always leave your borders garrisoned with a force at least equal to the force the AI could attack you with.

BTW - Strategery - ROFL, you have a hilarious user name. Maybe I'll change mine to "Lock Box" http://www.totalwar.org/forum/non-cgi/emoticons/wink.gif

Gregoshi
01-11-2003, 07:28
The following is a reply from Hermann the Lombard in the Entrance Hall:

-----
This is a reply to Hakonarson in the Main Hall. For context, I provide his quote as well:

*****

Quote[/b] ](Hakonarson)

Quote[/b] ](Jacque Schtrapp @ Jan. 07 2003,23:45)
If the human player makes a move on the campaign map and ends the turn, he/she can witness that AI moves for the same year before the florins/year are updated in the bottom right of the campaign map. These AI moves are supposed to be based on the AI's strategy regarding the humans movements from the previous year. Yet I have found save after reloaded save that (especially in the latter stages of the game) when I move a large army out of a province that I possess which shares a border with either an enemy/neutral/ally and leave that province virtually undefended I will be attacked. However, if I reload and do not remove the army from its province I will not be attacked. I have even gone so far as to reload the same turn numerous times adjusting the quantity and composition of my troops and have discovered that when the AI achieves a numerical or qualitative superiority it will attack.

Now those of you who claim that the AI "cannot see" the players movements that take place during the same calendar year, please explain how this is true when I have proved beyond the shadow of a doubt that the AI does in fact respond in 1276 to moves that I intentionally orchestrated in 1276.

Same holds true in naval situations. Share a sea zone with a foe, where the number of ships is in your favor and he will not attack. However, if you move enough ships out of the sea zone so that the numbers favor the AI, then he will attack the remaining ship(s) the same turn that you are moving your ships to a new sea zone. Reload and leave the ships and he does not attack.

The AI does see and react to player movements if it is in its best interest to do so. Still, all in all, its not nearly as prevalent or asinine as STW.

You haven't "proved" anything of the sort - indeed you seem to have an odd understanding of the word "prove".

Moreover no-one has to provide any explaination as to why the AI makes teh moves it does - it is you making the accusation, therefore it is YOU who has to provide all teh information to justify your own position, not everyone else who has to show that your accusation is wrong.

Unless you go to the code and show how it's happening all you have showen is that in some individual cases the AI has made those moves - yuo have shown nothing at all about why it has made those moves.

Unfortunately I don't expect you to accept this - howls of "justified" outrage rarely answer to the requirements of logic or objective evidence because they are so deeply based in emotional terms.

You are not going to get any joy from your demands that the programmers admit that the AI cheats - indeed as yuo said at the start they have quite explicitly said that the AI does not cheat.
***

Hakonarson,

Regarding "proof," Jacque Schtrapp stated a hypothesis, that the AI "cheats" defined as choosing a move based on the player's actions or inactions within the same turn. He then conducted an experiment involving multiple repetitions where he attempted to control all variables except the player's choice of moves. The results of the experiment support the hypothesis.

Now, does that constitut "proof?" Perhaps not in and of itself, but the hypothesis meets the criteria of falsifiability (it can be DISproved). If other "researchers" are able to reproduce the experiment in multiple game situations, then collectively they will have produced results that strongly support the hypothesis.

It's true that we don't have the precise data on the number of repetitions so we can't even calculate the size of the error bars...but you could conduct such an experiment yourself and see if your results support the hypothesis or not. You could even ask J.S. for a copy of HIS saved game, allowing you to reproduce his experiment as closely as you wish.

Obviously there is a lot of anecdotal evidence that the AI sometimes responds to human moves within the same year...but we know that anecdotal evidence isn't worth much. We remember the incidents that support an idea and fail to notice the cases which don't.

So did he "prove" his point? No, but he obviously made a good-faith and relatively scientific effort to do so...and if you wish, you can try to disprove his point by experiment (as opposed to using Aristotelian reasoning).

-----

MizuKokami
01-11-2003, 08:31
the only possible explanations i can think of that might explain away a small army moveing into a province the same year that(as far as the ai should know, it was way defended) you decided to move your forces out of the territory, is if they have a spy in that territory sending back reports of troop mobilization. either that, or the ai is just plain hopeing that you will move those forces.
in responce to hak, i too have seen the ai move forces into a territory that they should have thought was impossible to attack, and am quite certain that the ai is responding to moves they shouldn't have seen. i agree entirely with those that say the ai is doing something that it shouldn't be able to do, but would rather call it haywire programming rather then outright cheating. the ai isn't quite smart enough to cheat. http://www.totalwar.org/forum/non-cgi/emoticons/rolleyes.gif granted, we have no absolute proof as to this attribute of the ai, but obviously we have an abilty to determine logic thru experimentation. where as others have the ability to ignore logic simply because they don't want to agree.
however, i do get quite crazed when i see it happen, and have a long list of adjectives for the ai when it does happen. here's to hopeing it will be fixed some day.

p.s. here's to hopeing that the sarcasm isn't lost on the blind http://www.totalwar.org/forum/non-cgi/emoticons/tongue.gif

Vrashk
01-19-2003, 03:02
Or my reply *Coughs*


Anyone know if it is like I say that the player ALWAYS move first in strategy map ?

If so, I believe:

Strategy map moves, does contrary to belief NOT happen at the same time. Everyone moves in their pace and at the end, all battles are resolved.

Efrem Da King
01-19-2003, 07:55
We all know the ai cheats, thats why I play only normal and easy difficulty levels http://www.totalwar.org/forum/non-cgi/emoticons/wink.gif .

Cooperman
01-19-2003, 13:38
I think the AI was modelled on the grand old duke of york as it quite oftens marches troops into an enemy province against impossible odds then marches them back again usually getting itself involved in a war it has no hope of winning.
Dunno why it does that but if you moved your troops out when it did that then it got lucky.
You didnt say what your military was either, the Aragonese tend to play quite an agressive game attacking larger neighbours with their Royal Knights although they usually come unstuck rather quickly.
In general the AI sux so bad that it could use all the help it can get but it's not cheating.

Bob the Insane
01-20-2003, 19:07
From what I have seen playing the game, I am not decided one way or the other at the moment. By that I mean I have seen the AI make both really stupid moves and moves so good they require psychic powers...

I will point out on issue though. Spies, if you have spies in the right place, they will pass on info like "Faction A will invaid Province B in X turns". This gives you future info to work with. There is no way for the AI to do this against your faction (with the above psychic powers..). Or possibly by 'peaking' at the moves you are making during the AI part of the turn.

Could this be modeling the effect of AI spies actually spying (not producing revolts..)??????

Nick Marcus
01-20-2003, 19:38
I've got a far better example where the AI cheats EVERY TIME

We've all been there: Nice juicy province under our control. But a far to weak garrison, or none at all. When the attack comes, your screwed. Defend it with 50 peasents or get the hell out of there.

Now reverse the situation.
Nice juicy enemy province. Not really enemy, ally more like, for almost 20 years. The leave the place completely undefended, but you don't make a move.
Because you KNOW that the moment you set food on that land, the frikking huge enemy army on the next province comes down to defend it.
It's even apparant when huge amounts of reenforments come pouring down to defend an attacked, but well defended, province.

=> they get time to prepare the defences, we don't.
=>An attack of us is seen by the ai in the same way we see rebellions

barocca
01-20-2003, 22:02
Quote[/b] (Bob the Insane @ Jan. 20 2003,12:07)]From what I have seen playing the game, I am not decided one way or the other at the moment. By that I mean I have seen the AI make both really stupid moves and moves so good they require psychic powers...

I will point out on issue though. Spies, if you have spies in the right place, they will pass on info like "Faction A will invaid Province B in X turns". This gives you future info to work with. There is no way for the AI to do this against your faction (with the above psychic powers..). Or possibly by 'peaking' at the moves you are making during the AI part of the turn.

Could this be modeling the effect of AI spies actually spying (not producing revolts..)??????
a possibility,
perhaps if the AI has spies in your provinces it gets to "peek" at your moves....

cugel
01-21-2003, 03:06
Jacque Schtrapp is right about the AI adjusting its strategy (how many men to invade with and whether to attack at all) depending on what the player does. I've reloaded a saved game numerous times to see this, sometimes to avoid having a game spoiled by the AI's stupid move in attacking my shipping (say) when it may have a local superiority, but a crushing overall inferiority compared to my fleet.

I gues my response to claims of "AI cheating" (if you want to call it that)would be:
SO What?

The AI in general plays rather badly and doesn't put up a reasonable contest most of the time (on expert). It needs every advantage it can get. In programming this game the developers did the best they could to create reasonable gameplay in a number of different areas: 1. the tactical battle AI, 2. multi-player, 3. the strategic AI. There are many shortcomings in all of these areas, but considering the difficulty as a programming exercise, I'm surprised they did as well as they did. Perhaps there are better pure strategy games out there, certainly there are games with a more robust diplomacy model (Europa Universalis and Civilization come to mind), but the combination of tactical and strategy components is unmatched IMHO. That's why I still play despite all the problems.

Elwe
01-21-2003, 03:55
The AI doesn't cheat... it works exactly as it is programmed to work.

Cheating, by definition, means breaking the rules; something that a program is incapable of doing. The AI acts the way it acts because it is programmed to act that way. It doesn't 'cheat' at all. It merely follows a different set of rules to those imposed on the player.

Personal opinion: I prefer the AI acting the way it does. Adds to the challenge of the game http://www.totalwar.org/forum/non-cgi/emoticons/wink.gif

Cheers.

Bob the Insane
01-21-2003, 11:32
Quote[/b] (Elwe @ Jan. 20 2003,20:55)]The AI acts the way it acts because it is programmed to act that way. It doesn't 'cheat' at all. It merely follows a different set of rules to those imposed on the player.
I would have to disagree with you there. On a personal level I would say, what you said is the very definition of AI cheating... That it plays to it's own rules to give it an advantage over the player.

But if, as I suggested, this behaviour is merely the computers version to using spies, then all is fine and happiness again...

Asmodeus
01-21-2003, 13:04
Quote[/b] (barocca @ Jan. 20 2003,15:02)]
Quote[/b] (Bob the Insane @ Jan. 20 2003,12:07)]From what I have seen playing the game, I am not decided one way or the other at the moment. By that I mean I have seen the AI make both really stupid moves and moves so good they require psychic powers...

I will point out on issue though. Spies, if you have spies in the right place, they will pass on info like "Faction A will invaid Province B in X turns". This gives you future info to work with. There is no way for the AI to do this against your faction (with the above psychic powers..). Or possibly by 'peaking' at the moves you are making during the AI part of the turn.

Could this be modeling the effect of AI spies actually spying (not producing revolts..)??????
a possibility,
perhaps if the AI has spies in your provinces it gets to "peek" at your moves....
This makes a lot of sense. It would be a serious disadvantage if the humans could be warned before hand of an invasion but the AI spys were incapable of it. The AI also uses a hell of a lot of spys judging by the number that get caught by border forts, a lot more must make it through so you never know if they are there or not. Enough to make anyone paranoid.

After all, this is essentially the primary purpose of spy, to know your rivals intentions in advance. It seems to make sense to me anyway.

I reckon it gets a peek at your move if it has a spy there.

I've also seen the exact opposite occur; I invade an enemy province and on that same turn it moved some armies to another province, unintentionally weakening itself. No spys there on that occasion

I guess only the developers know this for sure.

Elwe
01-21-2003, 22:51
Quote[/b] (Bob the Insane @ Jan. 21 2003,04:32)]
Quote[/b] (Elwe @ Jan. 20 2003,20:55)]The AI acts the way it acts because it is programmed to act that way. It doesn't 'cheat' at all. It merely follows a different set of rules to those imposed on the player.
I would have to disagree with you there. On a personal level I would say, what you said is the very definition of AI cheating... That it plays to it's own rules to give it an advantage over the player.
But just because an opponent is following a different set of rules, doesn't mean it's cheating. 'Cheating' means that you have certain rules imposed upon you that you are deliberately breaking. It isn't playing to it's own rules, it's playing to the rules given to it, that differ to those given to the human player. A fine distinction to be sure, but a valid distinction non-the-less.

An advantage over the player implies that by acting in this way, the AI should win (almost) every game. And yet, there are countless posts on this forum that state the exact opposite, that the game is too easy. So where is this advantage that the AI has gained? http://www.totalwar.org/forum/non-cgi/emoticons/wink.gif

But I like yous suggestion that this is the way the AI represents the use of spies. It helps explain the reasoning behind the way the AI behaves. Good thinking http://www.totalwar.org/forum/non-cgi/emoticons/pat.gif

Cheers.

econ21
01-22-2003, 11:30
Anyone got a link to what Cold Steel said on this? I know that in Shogun, you moved first and the AI responded. It made the strategy map rather tough and the gameplay a little frustrating and unforgiving. I think in Medieval they toned this down and consequently the Single Player game is much more enjoyable to me. Exactly how they toned it down, I'd like to know - but what I have observed is still consistent with a sequential, as opposed to simultaneous, AI move happening to a moderate degree. I certainly play as if the AI can react immediately and try to leave all my border provinces with at least a deterrent force.

Personally, I am happy with this as it helps the computer AI out, means it is very tough to wage an offensive war without sufficient material and doesn't seem unrealistic. Would you not expect some opportunism if you leave a province undefended in the way described? Anyone who has played Imperialism II - a wonderful game, BTW - will know that it is VERY easy to exploit an AI that reacts only to your move on the previous turn. (Basically, you can swop one army continually between two provinces, inducing the AI to make weak attacks against what it believes is an undefended province and so win lots of stars for your general. Yuk, give me a "cheating" AI any day.)

Could it be due to spies? Maybe, I don't tend to use spies and am not sure how they work. But I doubt it. Anyway, it should be easy enough to test - run the experiments right at the beginning of the early period, when no one has the tech to build spies.

Hakonarsen - I'm a little surprised at your response, as I always respect your insights on history and other matters. Surely we don't need to break into the source code to resolve this? A replicable experiment would be sufficient and is the "scientific" method. Maybe Byzantines and Turks would be the obvious case to play with, as those two seem to be at each others throats very quick on early.

I don't think it is "stupid AI behaviour" - I think the AI attacks when it "should" win (eg on autoresolve) and does not factor in the fact that the human can outfight it. I don't believe 101 men should be able to beat 1200 (I am sure the original poster is not leading 1200 peasants).

Lord Aeon
01-22-2003, 14:10
Well, was going to suggest something along those lines (re: spies).

Scenario 1: Let's say youre going to move that large army out of Aquitaine to invade Normandy, and are going to bring down soem reinforcements from Brittany. The Aragonese have spies in Aquitaine, whoch inform it that you're planning an invasion of Normandy for year after next (i.e. the CPU gets to react to that move as if it knew what you were going to do before you did it).

However, there's no way for the CPU to get wind of any planned fortification of Aquitaine, so it would have no way of knowing you'd be bringing al large force down from Brittany --> produces the problem you've described.

Scenario 2: Maybe the AI was simply taking a chance. Maybe you have a tendency to leave provinces completely undefended whil you move troops around, but decided on a few occasions (such as this one) to fortify. The AI makes a bad move. Perhaps the AI saw that you needed, for whatever reason, to bring a large amount of troops into some province adjacent to Aquitaine, and the army in Aquitaine was the only likely candidate for the move. It makes a well-timed attack (albeit with a poor selection of troops).

Finally, if the AI simply has no other faction to attack, it will attack yours, even with no hope of victory.

Just a few suggestions.