Cheating AI

[Paul Bambury]

I've been playing EUIII on and off for a few months, trying different nations, strategies etc, and I'm increasingly frustrated by the apparent ease with which the AI nations discover and colonise the New World. After an attempt to make colonisation more expensive and difficult by modding certain values in the defines.txt file (amongst other things I made the base colonist cost value 50 ducats) seemed to have no effect on the AI, I decided to try and find out exactly what progress of the AI nations was.

Using the fact that one can load a save game and play as any country I created a series of monthly saves (playing as Munster) and then used these to see what the AI nations were doing. The results were quite surprising. NB, this was against a clean install of EUIII, patched to 1.2.1 with absolutely no modifications. Difficulty was set to normal.

I started by looking at Castille. On the 30th May 1453 The AI receives the same setup as a human player. Castille (screenshot) has 192 ducats and is running a deficit of -2.19 a month. Moving on to the 1st June 1453 (screenshot) and Castille has declared war on Granada. She now has a general and an admiral, and has started recruiting a regiment of cavalry. Yet miraculously, despite all this expenditure, Castille still has 170 ducats! With the admiral and general costing 55 each, Castille should have around 80 ducats. The only thing that seems to have been paid for is the cavalry. Additionally I couldn't help noticing that, given the 0% army and navy tradition, Castille has managed to recruit far better leaders than the human player would manage under the same conditions!

Unfortunately Castille went on to select Merchant Adventures over QftNW, so I continued this study with England, who gained level 1 government tech on 1st October 1454. The war with France has been eating into the ducats (screenshot). England has 4.5 currently, is gaining 0.34 monthly and 106.6 at the end of the year. By the 1st November 1454 England has recruited an explorer and sent him off at the head of the 1st Navy (screenshot). Again the AI has apparently not had to fork over a single ducat to recruit Horatio Saunders.

On 1st April 1455, Horatio has reached the Northern Guiana Basin, and the carrack Michael is in remarkably good shape following a 5 month Atlantic voyage (screenshot). In fact she's at 100%! Domestically England appears to have spent most of her annual income on the war effort, and now has on 5 ducats (screenshot). During April 1455 Horatio discovers the Coast of Hispaniola and a colonist is dispatched to Barahonas (screenshot). As with the recruiting of leaders, this action appears to have cost the AI player the princely sum of 0 ducats! Notice that the Michael is still holding up well. By June England, having got sucked into Portugal's war on Castille, has seized The Canarias. (screenshot) Another colonist is despatched there. Why not? It’s free!

1st September 1455. Since the colonist sent to The Canarias was successful, England sends another. And, as expected, it also cost nothing. (screenshot)

Throughout the rest of the year England keeps sending colonists as they become available. At the 1st of December we find 3 on the go, still without the necessary finances (necessary for the human anyway) to have sent them. Oh and the Michael has reached continental America after an 11 month voyage (screenshot), during which the weather has been the kindest in recorded history.

Conclusion? The AI is a dirty rotten cheater!

Question? How can we make the AI play by the rules?

 

[unmerged(5298)]

I like this post...

I think somehow connected gamebalance. I would add one thing too. When you are ahead of tech development, it suddenly takes more than 100 years to reach the next level. What's then the point of putting youre income into tech? Just a thought

 

[Darkrenown]

The AI gets spies for free too, not sure if the free stuff is intended to make the AI play better or is a bug.

The lack of naval attrition is a known AI cheat, apparently the AI can't cope with it. This has been true in all Paradox games I've played.

Castile's leader quality may be a cheat, but you can get good leaders with low trad or crappy leaders with high trad, there's some randomness there. Also luck may affect leader quality, was castile lucky?

 

[Darkrenown]

I like this post...

I think somehow connected gamebalance. I would add one thing too. When you are ahead of tech development, it suddenly takes more than 100 years to reach the next level. What's then the point of putting youre income into tech? Just a thought

That's the ahead of time tech penalty, techs too far ahead of their time cost a lot more.

 

[unmerged(67447)]

The AI not playing by the rules is my biggest problem with this game. They can also explore land Terra Incognito without conquistadors if they happen to be at war with the provinces owner. NOT COOL!

 

[Boblof]

I've suspected that the AI cheats with colonists, portugal alway seems to colonize far more than they should be able to do.

 

[unmerged(5298)]

That's the ahead of time tech penalty, techs too far ahead of their time cost a lot more.

Is there any point putting all the montly income into techs. Can I mint for some time, until the next nation reach my level?

 

[Fookison]

Take what the game gives you, work to its strengths. That is why we play the game. There are certain mechanisms in the Paradox games to ensure we cannot run and gun in the game and get bored 50 years into it by being so far ahead of the rest of the world. At times it is frustrating, but fun to solve the mystery. Tell me one strategy game where the AI is not "assisted"....

 

[unmerged(46403)]

Is there any point putting all the montly income into techs. Can I mint for some time, until the next nation reach my level?

You put money into tech...to avoid inflation.

 

[ppigor]

How to decide a problem

The AI is a dirty rotten cheater!

Question? How can we make the AI play by the rules?

AI does not use money, but duration of actions is the same that for a player.
Only the new update of game can correct a situation.

Reason is in complication of programmatic realization of AI.
I assert it as a programmer with experience more than 20 years.

I am sure that the developers of game will create adequate AI and that it will not sprawl on long years.

 

[Paul Bambury]

The lack of naval attrition is a known AI cheat, apparently the AI can't cope with it. This has been true in all Paradox games I've played.

Didn't know that. But EUIII is the first Paradox game I've played. Still puzzled by that though. I mean presumably the AI can cope with the concept that naval units damaged in combat need to be brought back to a friendly port for repair? How hard can it be for Paradox to get the AI to check the status of ships that are exploring, and bring them to nearest friendly port should the state of repair drop below a set level, say 55%.

Castile's leader quality may be a cheat, but you can get good leaders with low trad or crappy leaders with high trad, there's some randomness there. Also luck may affect leader quality, was castile lucky?

Castille is indeed a lucky nation in this game. Portugal isn't, so I had a quick look to see what the quality of the leaders Portgual recruited was, and they were much more in line with what one might expect (single point of fire, two on shock, that sort of thing) for 0% tradition. Still got 'em for free though!

 

[Veldmaarschalk]

The AI doesn't cheat, the AI plays by the rules that are set for him.

Only a human player can (and a lot them do) cheat.

The AI just has different rules then us humans, and that is because we as the human player are much smarter and better at playing this, or any other game, then the AI.

 

[ppigor]

It seems to be half-correct, sorry...

The AI doesn't cheat, the AI plays by the rules that are set for him.

It's right certainly, but you will not find players, which will agree with a fact that AI can do more than they.

 

[Paul Bambury]

The AI just has different rules then us humans, and that is because we as the human player are much smarter and better at playing this, or any other game, then the AI.

I do appreciate that the AI (in most strategy games) is given certain advantages of the human player, particularly so at harder difficulties. It does however seem that the AI in EUIII is getting quite significant advantages over the human player, even on normal and easy settings, that I consider to be unfair.

For instance, one of most obvious consequences of getting free colonists is the AI can go to war with impunity. If the human is at war (particularly in the early game) pretty much all of their resources have to go into fighting (and paying for) it. The AI meanwhile can spam colonists/spies regardless. This isn't giving the AI an advantage to compensate for us smarter humans; its stacking the deck.

AI does not use money, but duration of actions is the same that for a player.

But the AI does use money. It uses it to recruit regiments, naval units, build province improvements. If its bright enough to do it for these things, why not colonisation?

 

[Darkrenown]

Didn't know that. But EUIII is the first Paradox game I've played. Still puzzled by that though. I mean presumably the AI can cope with the concept that naval units damaged in combat need to be brought back to a friendly port for repair? How hard can it be for Paradox to get the AI to check the status of ships that are exploring, and bring them to nearest friendly port should the state of repair drop below a set level, say 55%.

There's actually quite a lot to factor in for the AI when thinking about naval attrition so making the AI deal well with it would probably be quite complex. I guess it doesn't work out well from a cost/benefit point of view. I'd like to see the AI handle this without cheating of course, but this is a cheat I can fairly easily forgive.

Getting leaders etc. for free, on the other hand, seems pretty weak to me. That's why I'm hoping it turns out to be a bug that will be fixed rather than a deliberate AI helper cheat. IIRC the AI managed to pay for stuff just fine in Vic, and seems to balance its budget ok in EUIII with the stuff it does pay for.

 

[unmerged(12131)]

Strategy Game - NO cheats

Take what the game gives you, work to its strengths. That is why we play the game. There are certain mechanisms in the Paradox games to ensure we cannot run and gun in the game and get bored 50 years into it by being so far ahead of the rest of the world. At times it is frustrating, but fun to solve the mystery. Tell me one strategy game where the AI is not "assisted"....

Buy Fritz10 Chess Program. It does NOT cheat and even beats Grand Masters like Kasparov at full strength.

 

[Darkrenown]

Galciv2's another game with good, non-cheating AI.

 

[Chaingun]

Buy Fritz10 Chess Program. It does NOT cheat and even beats Grand Masters like Kasparov at full strength.

Chess is an entirely different game than EU3. EU3 has a much larger complexity. Chess AI is smart simply by looking lots and lots of moves ahead. The state of the game after each is easy to determine. EU3 has this thing called "imperfect information", nor are there any clear-cut moves.

That said, I'm not saying it's impossible to make a smarter, non-cheating EU3 AI (which should be the goal). Though, it seems in other threads, people already complain about it being too hard.

 

[Mr Me]

If I was the Ai I would feel that it was the player who was cheating. The player can adjust game speed and pause at any moment he wants. He can think ahead and already knows of many of the events that may hit him. He can go back to a previous save if he makes a big mistake. He can choose the best choice in an event instead of almost always chosing choice A.

Do you see what I'm getting at here?

 

[ppigor]

The amendment

But the AI does use money. It uses it to recruit regiments, naval units, build province improvements. If its bright enough to do it for these things, why not colonisation?

AI does not spend money neither for colonization, nor for hiring of leaders, it is checked up. AI receives them absolutely free of charge.
At declaration of war AI employs so much mercenaries, how many considers it necessary. This quantity cannot be employed on an available money.
It seems to you, that AI uses money, but it is error. AI does not use money even on payment of advisors and the maintenance of army and fleet, all of money go on researches.