Game is Busted (AI cheating)

[EroticOnion23]

Decided to pick the game up again, but noticed the balance have been gimped up the azz...Was playing greenskins, by the time my armies had mostly Big'uns, the TKs had fking necrosphinx, Skaven had full stormvermin armies, and dwarfs gyros + flamethrowers. wtf...not to mention the AI team up on you from all directions...

[Cacomistle]

Messaggio originale di Mactalon:

Messaggio originale di CaoLex:

This having to do more with the fact that AI have faster recrutment, faster and cheaper buildings and pays only half of an upkeep.

None of those are true. They get much more base-level income and possibly more recruitment slots but that's it. They have the same building times, costs, recruitment time and upkeep as the player.

Go play a game on normal, then play a game on very hard.

You can see how much money ai is making, and the 30% extra income doesn't explain how they get 3 armies on 1 city without upkeep reductions (also all ai have no supply lines, so even the easy ai "cheats" on this)?

You'll already mentioned recruitment slots so I'm not sure why you're including recruitment speed in the things he's wrong about. Recruiting more units at once lets you recruit faster.

I don't know about time for actual buildings, but the ai grows ridiculously faster on higher difficulties. Greenskins on normal won't be tier 4 till like turn 60, but on very hard you'll fight black orc stacks like turn 30.

The difficulty change honestly wouldn't even be noticeable if all it was is a 30% income increase (unless that number changed, but that's been it every time I looked).

Oh also you're statement that it's only base income doesn't even include their tax rate which you can see within 5 seconds of loading up the game. I'm going to do your post a favor and assume you were including this as part of base income.

[CaoLex]

Messaggio originale di Ser Pounce:

You're either playing VH or legendary or not playing efficiently if the AI is outpacing you.

Messaggio originale di Hrodh:

And there's the crux of the matter: what plays do you take? Multiplayer, campaign? Campaign on Legendary, normal, easy? Min-Maxing, cheesing, lore-friendly? Too many variables...
Would be interesting to program, but probably too much.

Check the game files throught mod editor. AI recieving cheats even on easy. Not all of the bonuses active on easy thought and some of them is weaker than on other difficulties.

By the way, there is exist a game "Polaris Sector" which consider player recieving 2 times resources the AI do as normal difficulty.

[Elitewrecker PT]

MMm no, the player has cheats on easy.

[Mactalon]

Messaggio originale di Cacomistle:

Go play a game on normal, then play a game on very hard.

You can see how much money ai is making, and the 30% extra income doesn't explain how they get 3 armies on 1 city without upkeep reductions (also all ai have no supply lines, so even the easy ai "cheats" on this)?

You'll already mentioned recruitment slots so I'm not sure why you're including recruitment speed in the things he's wrong about. Recruiting more units at once lets you recruit faster.

I don't know about time for actual buildings, but the ai grows ridiculously faster on higher difficulties. Greenskins on normal won't be tier 4 till like turn 60, but on very hard you'll fight black orc stacks like turn 30.

The difficulty change honestly wouldn't even be noticeable if all it was is a 30% income increase (unless that number changed, but that's been it every time I looked).

Oh also you're statement that it's only base income doesn't even include their tax rate which you can see within 5 seconds of loading up the game. I'm going to do your post a favor and assume you were including this as part of base income.

No need for any favours :P

I always play on very hard, there's really no need for that either.

I mean, yes, it pretty much does. While I've rarely in 2000+ hours seen the AI actually field 3 full 20 stack armies on only 1 city left (as they usually do a couple of stacks of 10-15), when I have seen those sorts of situations - what they recruit and the income they have left essentially adds up. They're often just stacks of chaff that gets bulldozed, and even then it still takes a number of turns for them to even get a full 3 armies going. They can't recruit units from buildings that don't exist either, which is why when they've been reduced to one city they often build the chaff armies because that's all that's available but they can buy in bulk.I did forget about the supply lines though, so in that regard the upkeep is untroubled by multiple lords being out, but the extra income is more than enough to hold together what they build. The player on one major city can field 2 armies, it's not ridiculous for the AI to do so. When they're down to one minor city they really don't magic out multiple doom stacks.

To make it clearer, recruitment slots and recruitment speed are actually two separate mechanics within the game, hence why some buildings in the game increase the speed by taking turns off, which is what I was talking about, and why I then went to the trouble of distinguishing it from recruitment slots later on. Yes, you will get more troops faster with more slots, but weirdly enough I was talking about the mechanics as they function in the game.

I'm honestly not sure about the growth, you might be right that they get a boost but in my own experience I've never seen the factions grow inhumanly quickly.
The AI has a different focus than what the players do, and with the aforementioned increase in funds they can generally afford to immediately purchase upgrades and consolidate their territory, while the player is often spread more thinly and has to make choices. As their main threat is also the player, they also often have more room to redo their armies if they're not being continually hammered, which I have seen them do a number of times once new unit types have become available. Again it always seems to stem more from the easier access to funds. When the AI is in more wars they do slow up, which is why the player will often overtake them. The AI also gets the benefits of research as well, which they likely also prioritise in a different order than the player. Left to their own devices they'll just steadily grow but it's all manageable.

Black Orc stacks by turn 30 is something of a stretch though, at least not with all the GS armies fielding them in any noticeably serious manner, if that was happening more consistently the Orcs probably wouldn't be getting constantly smushed in every campaign XD The AI is given leeway to be able to keep pace with the player, but it's rarely outrageously fast in that manner. I mostly see Big Uns stacks with Goblin support about that point pretty much every time.

[Cacomistle]

Messaggio originale di Mactalon:

Messaggio originale di Cacomistle:

Go play a game on normal, then play a game on very hard.

You can see how much money ai is making, and the 30% extra income doesn't explain how they get 3 armies on 1 city without upkeep reductions (also all ai have no supply lines, so even the easy ai "cheats" on this)?

You'll already mentioned recruitment slots so I'm not sure why you're including recruitment speed in the things he's wrong about. Recruiting more units at once lets you recruit faster.

I don't know about time for actual buildings, but the ai grows ridiculously faster on higher difficulties. Greenskins on normal won't be tier 4 till like turn 60, but on very hard you'll fight black orc stacks like turn 30.

The difficulty change honestly wouldn't even be noticeable if all it was is a 30% income increase (unless that number changed, but that's been it every time I looked).

Oh also you're statement that it's only base income doesn't even include their tax rate which you can see within 5 seconds of loading up the game. I'm going to do your post a favor and assume you were including this as part of base income.

No need for any favours :P

I always play on very hard, there's really no need for that either.

I mean, yes, it pretty much does. While I've rarely in 2000+ hours seen the AI actually field 3 full 20 stack armies on only 1 city left (as they usually do a couple of stacks of 10-15), when I have seen those sorts of situations - what they recruit and the income they have left essentially adds up. They're often just stacks of chaff that gets bulldozed, and even then it still takes a number of turns for them to even get a full 3 armies going. They can't recruit units from buildings that don't exist either, which is why when they've been reduced to one city they often build the chaff armies because that's all that's available but they can buy in bulk.I did forget about the supply lines though, so in that regard the upkeep is untroubled by multiple lords being out, but the extra income is more than enough to hold together what they build. The player on one major city can field 2 armies, it's not ridiculous for the AI to do so. When they're down to one minor city they really don't magic out multiple doom stacks.

To make it clearer, recruitment slots and recruitment speed are actually two separate mechanics within the game, hence why some buildings in the game increase the speed by taking turns off, which is what I was talking about, and why I then went to the trouble of distinguishing it from recruitment slots later on. Yes, you will get more troops faster with more slots, but weirdly enough I was talking about the mechanics as they function in the game.

I'm honestly not sure about the growth, you might be right that they get a boost but in my own experience I've never seen the factions grow inhumanly quickly.
The AI has a different focus than what the players do, and with the aforementioned increase in funds they can generally afford to immediately purchase upgrades and consolidate their territory, while the player is often spread more thinly and has to make choices. As their main threat is also the player, they also often have more room to redo their armies if they're not being continually hammered, which I have seen them do a number of times once new unit types have become available. Again it always seems to stem more from the easier access to funds. When the AI is in more wars they do slow up, which is why the player will often overtake them. The AI also gets the benefits of research as well, which they likely also prioritise in a different order than the player. Left to their own devices they'll just steadily grow but it's all manageable.

Black Orc stacks by turn 30 is something of a stretch though, at least not with all the GS armies fielding them in any noticeably serious manner, if that was happening more consistently the Orcs probably wouldn't be getting constantly smushed in every campaign XD The AI is given leeway to be able to keep pace with the player, but it's rarely outrageously fast in that manner. I mostly see Big Uns stacks with Goblin support about that point pretty much every time.

Alright the black orcs is more like a turn 40 thing. They weren't getting smashed in campaign for quite a while though. And the biggest reason they do now is because they go against dwarves, who are possibly the strongest ai faction.

On recruitment, I don't think speed is an actual stat. I think how many turns units take to recruit is called duration. So I think most people are thinking of speed as total output. But if we want to be as clear as possible, as far as I know yes they only get slots.

As for the income adding up, I'm talking about factions like Tilea. First of all, the ai doesn't get the upkeep increases player does on higher difficulties (I think units on easy cost 10% less upkeep and 10% more on legendary than normal). But something like a swordsman costs over 100 upkeep. A province like Tilea might make about 2500 income total (2 income buildings for 1000, level 3 dock for 400, level 5 dock for 800, and a bit extra from settlement income). Add 30% and they're making 3250. Add passive income and its about 6000. If they fielded just armies of trash tier units, then maybe. But I've seen them field 3 stacks with units like greatswords and cannons in them.

Or take factions like some of the random dwarves sitting in the mountains. They go from 3/4th of a stack on normal (about what you'd expect from passive income, dwarf warriors are 120 I think so you can barely afford a stack of 20 of them and ai will often have a few better units), to about 1.5 to 2 full stacks on very hard. Passive income alone (they often don't even have an income building) doesn't give 2 stacks to the dwarves.

So either the ai gets more passive income, or they get lowered upkeep. Considering how the dwarves can afford like 5 armies off their starting province, I think lowered upkeep makes a lot more sense. And most people have claimed the ai has lowered upkeep.

I'll admit that I haven't done some experiments to prove this is the case. So maybe I'm wrong here. But either way the result is the same. The ai can afford a lot more armies than the player in the same situation.

Although I think the whole "ai cheats" thing largely ignores how much better the player manages their economy, which is generally enough to more than make up for the ai bonuses. Player is generally meant to be in the top 10 strength rank out of a ridiculous number of factions. If ai cheats were as powerful in practice as they are in theory, this would never be the case.

[Cacomistle]

Messaggio originale di Chief Minister of Gaming:

Messaggio originale di Hrodh:

And there's the crux of the matter: what plays do you take? Multiplayer, campaign? Campaign on Legendary, normal, easy? Min-Maxing, cheesing, lore-friendly? Too many variables...
Would be interesting to program, but probably too much.

Skitterleap. I am the new Programmer.

- I would scrap the AI difficulties as are. Level playing field. Same rules. Having the same rules, also makes many things more easier in the balancing, as we also have had issues with attrition/corruption and other functions.

Then I would reapply the same tier of difficulty, emulating real play. The many variables you speak of seem overwhelming, only because you do not have a scripting platform to shadow play.

Once you have the shadow script, it does the work for you.

I can't tell what you're talking about.

It sounds like you're saying you'd scrap the current system and redo the difficulties. And then from there I don't understand the rest.

Like what is a shadow script? I haven't programmed games so maybe I don't know the term, but I googled it and nothing shows up. That doesn't sound like a machine learning ai (which is basically the only way they'd make something better than a player, because spending a year or 2 making a truly competent ai sells worse".

Also do you mean the many variables he's talking about seem overwhelming. There are a ton of variables that are just part of the game. You can't ignore everything complex.

I'm assuming here English isn't your first language, so maybe that's leading to my confusion. But either way, I feel like you're vastly underestimating how hard an ai is to make.

There's a lot of improvements they could make, I think a large percentage of them would even be easy to make. But the campaign map is just something the ai is almost bound to be bad at. Even if it were like a machine learning ai. Like say you train an ai against itself, and then it goes against a player. And it turns out, players are better at winning battles than auto resolve is. Suddenly, the ai is just suiciding all their armies in what should be slightly favorable battles for them.

Like maybe they could get the ai to be challenging without cheats. There's 100+ ai factions on the map after all, so if they were even skill to the player the player would basically never win. But there are too many variables in this game for the ai to turn into that sort of unbeatable ai, unless you give them perfect micro in battles (which would be incredibly frustrating). A machine learning ai would probably start figuring out to do stuff like corner camping that nobody wants to play against. And a scripted ai is always going to end up having some sorts of abusable flaws, because nobody wants to sit through testing to go through every single one of the thousands of options players have to try and break the ai.

Basically, its not as simple as you're making it sound (I think, I still can't fully tell what you're talking about). Although I do personally believe it wouldn't be that hard to make an ai that isn't worse than the average player even with cheats that make them like actually twice as powerful (I think even new players would be able to beat a legendary campaign more than 1/100 times you'd expect based on the number of major factions in game).