AI Bonuses

[Marshall Ellis]

Hey guys:

I am playing around with a new method of AI bonus prevention. I am canceling all bonuses IF the MP has been defeated by a human for the period of the enforced peace. My hope is to reel in the AI bonuses enough to make it possible to achieve a victory here even with the AI bonuses in play. Give me some thoughts on pluses and minuses that you can see here (If you would, please)...

[easterner]

In my last game Turkey won. The Pr, Rs, Au tag team tied Fr up so that while it defeated all 3 piecemeal the bonuses gave Turks all the time they needed to win. The minors are winning an 11 year Campaign in 4 years. The bonuses need be watered down. Napoleon said, "Ask me for anything but time." The bonuses accumulate so fast that by 1806 they are 2x the human player. The bonuses did not accumulate this fast prior to v1.4.

Even better how about the A.I. just perform better. There is no reason for the Brit or Sp armies to sit in Portugal or other out of the way spots doing nothing. There is no reason Bagration crosses the Swede frontier to prevent lapse, then sits out game there after Swedes are conquered. There is no reason Turk & Rus fight phony wars. No reason why Rus doesn't finish Caucasus conquest. No reason why Rus army sits in Grodno or Brest Litovsk when at war with Turks. Or why they ignore a Corps quietly occupying St. Pete's & Moscow. Or fail to support Pr or Au when at war with Fr mutually. (Rus should lend Corps to Allies). Sp needs to figure out how to build a depot line in Morocco as game after game the cross Algerian frontier (range 3) and stop dead whole game. Ultra rarely do they amphib. Algeria. Br need stop sending 2-3 SP cross channel endlessly at Lille, violates principles of war (concentration of force).

[obsidiandrag]

I don't have any problems with actually keeping the Mp/$ and even a bigger pp+ ALL GAME as long as maybe we get rid of the VP bonus... this way they are not getting an unresonable ammount... more like free econ manipulation all game...
 
and I am all for more agression especially in certain situations for the AI..  Even to the extent of say Austria kicking down Paris door, Russia maybe declaring on Austria for the free land and a possible Poland while they are busy and cant take pressure off of France, but were winning.  Nothing like kicking while already down to France..
 
 

[NeverMan]

ORIGINAL: obsidiandragon

I don't have any problems with actually keeping the Mp/$ and even a bigger pp+ ALL GAME as long as maybe we get rid of the VP bonus... this way they are not getting an unresonable ammount... more like free econ manipulation all game...

and I am all for more agression especially in certain situations for the AI..  Even to the extent of say Austria kicking down Paris door, Russia maybe declaring on Austria for the free land and a possible Poland while they are busy and cant take pressure off of France, but were winning.  Nothing like kicking while already down to France..

I actually would prefer the other way around. I don't mind the VP bonuses, so long as 1) It's based on their PP standing 2) No PP bonuses.

PP bonuses prohibit instability and fiasco zones, which is absurd, IMO.

FURTHERMORE, I agree with easterner that if you implement the way you suggest Marshall, that you can have a country win simply because the Human player didn't go to war and beat them. This makes it so that the Human player has to beat every country at least once in order to win, which again, make NO SENSE!!!

[Marshall Ellis]

ORIGINAL: NeverMan

ORIGINAL: obsidiandragon

I don't have any problems with actually keeping the Mp/$ and even a bigger pp+ ALL GAME as long as maybe we get rid of the VP bonus... this way they are not getting an unresonable ammount... more like free econ manipulation all game...

and I am all for more agression especially in certain situations for the AI..  Even to the extent of say Austria kicking down Paris door, Russia maybe declaring on Austria for the free land and a possible Poland while they are busy and cant take pressure off of France, but were winning.  Nothing like kicking while already down to France..

I actually would prefer the other way around. I don't mind the VP bonuses, so long as 1) It's based on their PP standing 2) No PP bonuses.

PP bonuses prohibit instability and fiasco zones, which is absurd, IMO.

FURTHERMORE, I agree with easterner that if you implement the way you suggest Marshall, that you can have a country win simply because the Human player didn't go to war and beat them. This makes it so that the Human player has to beat every country at least once in order to win, which again, make NO SENSE!!!

Then should I make it to where they get no bonuses if ANY MP has defeated them?

[obsidiandrag]

Actually, if the interest is to make the AI harder ie harder to beat without it cheating, you would WANT it to be harder for them to go into instability and fiasco so it is harder to actually beat them right? Not give them 17VP per econ phase so you physically can not beat them.

[easterner]

No because AI rarely surrender to each other. They are so courteous that if they declare war on a minor and an ally gains control they rarely even cross the border.

1) Minor Allies must produce in Economics. That is priority one as not producing hurts the AI owner, badly.

2) If you want to give them production bonuses fine. But those do them little good as too often they place troops in useless places like Portugal.

3) Give extra VP based on standing, Dominant 0, Stability 1, Instability 3, Fiasco 5. Numbers are guidelines probably need adjusting based on play results. Another possibility is take opposite tack. AI get normal VP, Player gets penalized if dominant and no other power is. That could encourage player to slow down. One of great ironies is AI plays historical (lots of actionless turns) It also reminds me of many clueless humans I've played against, they have brilliant opening move, then no idea what to do after 1st 6 turns.

4) Surrender: Au & Pr quick (say 55% SP loss from Dow turn), Tu and Fr average (75%), Br, Rs & Sp slow to surrender (90%). With a random number to surrender after those losses. Loss of capital should trigger a random # to surrender, full occupy unchanged, over 60% capitals occupied trigger a random # to surrender. Fighting to destruction, particularly Au, Pr & Tu is a game loser, a lesson I've seen many humans not learn.

England wins if everyone loses, yet I've never seen a game hit 1814, so speeding victory to 1809, 1810 does GB a disservice.

[obsidiandrag]

I am not sure we should base bonuses on if they have already been beaten or not, IMO I would think that after they have been beaten is when the bonuses would make the most sense for $/MP to help rebuild and prepare for the next war.

I have not had the AI yet able to wage another war after the enforced peace from a defear (dont get me wrong, they WILL declare again). I usually just leave a corps or 2 there for the time required and then take them over again as soon as it is over. If the AI had a way to offer a conditional early (tactically to limit losses from a larger or more ready foe) then still be able to wage a good war later that would be better. As it is now, the AI waits until you have desimated them and taken 3 home provinces since you have already taken or they have lost to instability all minors origionally in their posession before surrendering. So after the war they can not rebuild very well and you have more than enough resources to stand by and then hit them again...

[Marshall Ellis]

So maybe leave the money/mp bonus but elimiate the VP bonus?

[easterner]

Proportionality is the final word. If maxing the Dominant zone and getting 12-15 VP each turn is good for a humiliating last place i.e. 57% (Player) to 102% (Tu or Sp etc) What's the point?????

[obsidiandrag]

Actually if you think about it, giving the AI pp not VP...

The Political Point chart WILL bring them back down each econ phase so they will not be maxing out, only if the AI is actually winning more than just sitting there. All the PP bonus would become would be like an economic manipulation for them only free. This way the country will have bonuses for play but not as much for the VP win in 1809 anymore. The chart will start kicking in the -1 -2 -3 as the AI starts getting higher up and not "Needing" the bonus anymore.

[yammahoper]

PP instead of VP is what I think too.

[NeverMan]

ORIGINAL: yammahoper@yahoo.com

PP instead of VP is what I think too.

But there are things that are involved with getting low PP that the AI would never then have to experience, taking a good portion of the game away.

[StCyr]

maybe recall what AI means at all ? just a hint: no, it is not about cheating...

[iamspamus]

Well, but most AI programs do cheat. That's the point. They are not and never will be as good as some [;)] players. So, yes, we want them to perform better, but we also want them to cheat in a way that does screw the game or make it impossible to win even if you've done "everything" correctly.

I was just summing up. I'm not sure what is the best method of fixing the problem.

ORIGINAL: StCyr

maybe recall what AI means at all ? just a hint: no, it is not about cheating...

[StCyr]

dear iamspamus,
I would get mad if the AI-programmes of my chess software would see it the same way [;)]. But you got a good point there, no question- cheating may "improve" a weak AI in a way that the game becomes more challenging.
But making the AI cheat should be the last option. The aim should be to have an AI for EiA where you simply pick no major power but watch all other AI controlled nations face each other. And have some kind of simulation by that, perhapse on a low level. But a working diplomatic and military AI on a low level would be the basic to improve the AI system. I mentioned it several times before - for example, read the EiA manual how to become a new major power, use these options within the individuell political concept for each major power (if there is any so far...).

[Jimmer]

Actually, I can't think of a computer game where the AI doesn't cheat, at least in some areas. Civilization (1 through 4, but this was removed in Warlords), for example, was famous for having the AI "know" where the good terrains and tribal villages were. At the higher levels, their starting exploratory troops would make a run straight towards the closest tribal village. And, they would plant new cities in what looked like awkward places, but later turned out to be perfect placement (once the then-hidden resources became visible).
 
So, given that, the real question is how MUCH "cheating" they are allowed. Personally, I like the idea of having the AI's abilities be tailorable by the player. I hate it when the computer "knows" the map before they should legitimately be able to see it. But, allowing them the equivalent of higher combat dice rolls seems fine to me. Other players might desire the opposite.
 
Now, an idea: In EiA and EiANW, the combat tables are built around six-sided dice which are rolled at a table. Discreet mathematics can be used, making the "odds" be an exercise in calculating step-function probabilities. In other words, there are only 36 possibilities, each chosen from a continuum of values (i.e. the set of real numbers between X and Y).
 
BUT, in a computer game, there's no reason they have to be laid out this way. The combat tables could be generated such that there is a continuum of results, rather than a small set of potential outcomes. For example, let's say that the combat table ranges from 0.2 morale loss through 2.6 on rolls of 0 through 7. In that range, there are 8 specific numbers (each with 1 digit to the right of the decimal) that can be rolled. What happens to the other 17 possibilities? They simply don't exist.
 
Why can't the computer (AI) be allowed to roll a result that is in-between the values the tables would allow? In the example above, let's say the possible results normally would be 0.2, 0.5, 0.8, 1.2, 1.5, 1.8, 2.2, and 2.6. Allow the computer to "roll" results that included the missing entries (0.3, 0.4, etc.). In fact, allow the computer to roll 0.55 and 2.16 and other strange rolls. Just keep the displayed values rounded for display purposes, but keep the "extra" points with the numbers.
 
If this were implemented, then it would be a very simple matter to give the computer an "AI fudge factor": Simple add 0.1 (or, 0.7 or 1.3) to the end result.
 
A similar concept could be applied to casuaty percentages: Remove the limitation on having to have percentages that are multiples of 5, and let the computer have in-between values. Again, the fudge factor would be to simply add a factor to the end result garnered by the "die roll".
 
The bottom line is that it would be possible for the AI to be "6% better" in combat than an equivalent force belonging to a human. Or, 7%, of 5%, or any other number one might choose.
 
With combat being the at the core of the game, though, this has the potential of being hugely unbalancing. So, it would be imperative to allow players to tailor these values (at least while the idea is being beta-tested, and probably forever). In other words, "easy", "normal", and "hard" AI would have selectable value for the parameters. "Parameters" would include not just combat dice rolling, but also the VP and/or PP bonuses currently applied. It could also include the odds of being friendly with another AI player or human player. Or, forage values. The possibilities are endless.
 
Comments?

[easterner]

Combat though isn't the A.I.'s problem, its strategy. The chit picking of the computer is adequate, could be better (it picks Withdraw from capital city!!! too often, should pick Cordon a little more against 4-5 opponents) but Morale is still key and the computer has it or doesn't.

But Strategy wise it's clueless attacking in penny packets and building massive armies in areas that will never see combat. Doesn't build depot chains well either.

A.I. Failures
Turks often end up with massive army in Egypt after Turn 1.

Rus build massive armies in Grodno/Abo then never move/use them. Rarelr sends troops near let alone in Tukry which it is always at war.

Portugal, there's nothing there, but every A.I. that owns it gets an urge to send massive troop quantities there.

GARR: A.I not keen on them though pretty good at garrisoning captured minors and home capital it leaves coastal cities and provincial capitals empty.

Brits too aggressive with 3SP, too passive with 40SP.

Au & Pr need to stop fighting to last factor.

Fr A.I. should declare war more frequently.

A.I. needs to get a clue with Amphib.

A.I. needs to respond to Corps on or near national capital not leave stacks snoozing as they fall.

A.I. need take a lesson from CIV A.I. Be relentless have a goal, mass forces and go for it. EiA A.I. has 8 Corps, commits 2. CIV has 8 units, uses 6.

Is Spain sacrosanct? Never seen Spain attack or get attacked by another A.I.

A.I. allies seem to think they are just bystanders and allow their allies to get crushed. A.I. needs to 'gang up' on winning power. A.I. needs loan troops (listening Russia) to allies if that can get them to front.

[Marshall Ellis]

Easterner:
 
I think you will see some strategy improvement in 1.06. I have trained it to build longer supply chains deeper into their enemies. Wars with Spain do not seem to happen too much because of a few items (I even remember this in older board games). First, she rarely has enough money to take her fleets out of port for very long thus preventing numerous amphib ops. Second, nobody else is setup good enough to invade Spain. Granted France could BUT a prudent French player (someone who learned from the actual history) would never try this while at war with Pr, Au or Ru. And lastly, those pesky guerillas can really drag out an invasion attempt! I guess my point is that Spain is simply in a boring position (Even the board game was like this). The only thing of value that she has is her fleet. I only say this becuase you're right! Spain plays the sitzkreig quite well.
Thanks for the feedback!
 
 
 
 

[easterner]

Marshall, can't you find something better to do than cloud the issue with FACTS! [:)]

Yes, Sp indeed daunts invasion. But she needs do better scooping up Africa and snagging up loose Italians. In one game I played I nearly fell over dead when a Sp amphib took Genoa, though it would have been cheaper walking in from Pama. Need more of that.