BtS WAY to easy???

[uberfish]

1) The game shouldn't be balanced around a specific overpowered strategy (Elizabeth globe theatre rifle draft)

2) the AI is coded to spend more on military, if you want it to use it turn aggressive AI on.

3) fast tech + 10% upgrade cost + insane mass units made warring about as fun as banging your head against the wall repeatedly

4) tweaking AI behaviour is better than messing with handicaps and I'm sure Blake is keeping an eye on the feedback for the next patch.

 

[frob2900]

Fair enough, but how hard is it to get the AI to prioritize currency, col, and courthouses? Combine that with getting the AI to USE it troops to expand instead of sitting around serving as a drain on the economy and that would help a lot. The expansion wouldn't have to be by force if there is land left. The point is just to claim more territory and have currency/col/courthouses supporting that expansion.

I think that would be one fairly straightforward fix.

Commerce city specialization and unit production-economy growth balance might be trickier, but I would rather see effort put toward these ends than reintroducing cheat bonuses.

My concern is we will only get 1-2 patches and I would like to see Blake given a chance to develop the AI. Adding mods to make simpler changes is always an option if he is ultimately unsuccessful but modding in complicated changes that he is capable of is certain not as easy, unless he is the one providing the mod.

Yeah, you're probably right about 1-2 official patches. I'm guessing 2.

Is the AI not prioritizing CoL/Currency now?

Would you say that the AI should become more aggressive especially toward the player when land runs out, or should the extra aggression be equal toward the player and AIs.

I've heard some people say that there is a problem with intra-AI wars (and I remember even some Warlords games where extensive AI-AI fighting led to very easy Monarch tech leads), so maybe skew the aggression against the player?

[EDIT]

3) fast tech + 10% upgrade cost + insane mass units made warring about as fun as banging your head against the wall repeatedly
4) tweaking AI behaviour is better than messing with handicaps and I'm sure Blake is keeping an eye on the feedback for the next patch.

Well, of course a behavioral fix is preferable. But how do you make the AI a credible threat to the player without large-ish stacks or a tech edge? Wouldn't it be a pretty daunting task to actually make the AI a good tactician? Or for that matter a strategist? Of course, I was extremely impressed by Blake's warlords Better AI (which I played a lot against), so who knows

 

[Phrederick]

I've heard some people say that there is a problem with intra-AI wars (and I remember even some Warlords games where extensive AI-AI fighting led to very easy Monarch tech leads), so maybe skew the aggression against the player?

Not by much, if at all. I love playing games where there's tons of intra-AI wars, because it feels a lot more realistic and less of a one-against-all feeling. Plus, since the AI is more efficient at taking cities now, perhaps those wars will lead to a more powerful victor more often.

 

[uberfish]

Well, of course a behavioral fix is preferable. But how do you make the AI a credible threat to the player without large-ish stacks or a tech edge? Wouldn't it be a pretty daunting task to actually make it a good tactician?

It does have large stacks, in the late game the AI just maintains stacks of 40+ units sitting around in case a war breaks out. AIs that go on a conquering rampage in the midgame and collect other AI vassals are dangerous, too. However they do stagnate more than before in the event of an indecisive war.

 

[frob2900]

It does have large stacks, in the late game the AI just maintains stacks of 40+ units sitting around in case a war breaks out. AIs that go on a conquering rampage in the midgame and collect other AI vassals are dangerous, too. However they do stagnate more than before in the event of an indecisive war.

Right, but generally those stacks are seriously out-of-date? (Compared to a moderately dedicated tech-beelining player)

[EDIT] I suppose what I'm getting at it is this:

The AI is teching slower because its paying much higher Unit Upgrade+Support (and espionage, but lets leave that out for the moment)
So its tech rate can be (simply) increased by either reducing its peacetime stacks (i.e. only let it build in preparation for war) or giving it some bonuses.

How would an alternative behavioural fix address this? A player would also fall behind in tech because of large stacks if the player didn't have the strategic and tactical insight to use the stacks decisively in a war, trying to gain as much territory as possible, while cutting losses etc. The AI seems to keep its stacks hanging around with no such tactical or strategic insight. And making it move its unit stacks decisively in a warwould be very difficult.

There a description somewhere on the forum (by Solver, I think) of how an AI doesn't even know when a player is moving a stack toward one of its cities. Apparently it blindly assumes that a city is in danger if player units are within a certain radius of the city, irrespective of what direction they are turn-by-turn, moving on the map. Stuff like that would have to be fixed, but apparently it would slow the game down seriously.

(Apologies for the repeated edits, I'm getting rather sleepy)

 

[pigswill]

If the AI AI is improved then I guess its improved at every difficulty level; if the previous AI level bonusses have been reduced then I guess the overall impact is that the game becomes more difficult for people playing Noble/Prince and easier for people playing Emperor+.

Unless the AI actually plays smarter at higher difficulties its going to be unbalanced.

 

[futurehermit]

I don't want the AI to be more skewed against the human. Why not go after an AI? I hated in warlords when Monty would march all the way across the frickin map to attack the human player, which would hamper him economically and divide his empire if he was successful, when he had easy pickin's next door.

For aggressors, if they build up military (they should) then they should attack a close, preferably soft, target while prioritizing currency and col (I don't know if they do now or not). That is how most humans think and play so the AI should do the same.

For peacenicks, they should build up military as well so as not to be the soft targets in question, but these units should be primarily defensive (still mixed stacks though) and they should prioritize expanding until the land is gone. I don't want Gandhi to start wars just because the land runs out, but in that case he should build defensive units at the rate that his tech can support it. Peacenicks should focus on not delaying defensive techs like feudalism and rifling and should focus on stockpiling some dough so that they can upgrade obsolete units once they hit these techs. In short, they don't have to attack, but they do have to tech and have a strong defensive military.

If the AI is going to build up military it can't support financially, it should be using that military to conquer territory. The spoils will support their research and with a larger empire they will have more production potential and--if they don't delay currency/col/courthouses/forbidden palace/cottages--greater long-term tech potential. If the AI is not going to conquer territory, it should only build military at a sustainable rate and should focus on teching, including defensive techs, and upgrades to protect its empire.

 

[frob2900]

I don't want the AI to be more skewed against the human. Why not go after an AI? I hated in warlords when Monty would march all the way across the frickin map to attack the human player, which would hamper him economically and divide his empire if he was successful, when he had easy pickin's next door.

Ok, good point. I do agree that the optimal situation is that the AIs should expand in the way that is best for them.

For aggressors, if they build up military (they should) then they should attack a close, preferably soft, target while prioritizing currency and col (I don't know if they do now or not). That is how most humans think and play so the AI should do the same.

For peacenicks, they should build up military as well so as not to be the soft targets in question, but these units should be primarily defensive (still mixed stacks though) and they should prioritize expanding until the land is gone.

Ok, so up the expansive aggression for aggressive leaders? (and make curr/CoL priorities). But since all AIs are building stacks, any soft target would be an AI with poor land (since less )? If two AIs have equal land will this turn into a slogging match of death?

I don't want Gandhi to start wars just because the land runs out, but in that case he should build defensive units at the rate that his tech can support it. Peacenicks should focus on not delaying defensive techs like feudalism and rifling and should focus on stockpiling some dough so that they can upgrade obsolete units once they hit these techs. In short, they don't have to attack, but they do have to tech and have a strong defensive military.

But maybe a bit smaller military than they currently have? And tech over espionage?

If the AI is going to build up military it can't support financially, it should be using that military to conquer territory. The spoils will support their research and with a larger empire they will have more production potential and--if they don't delay currency/col/courthouses/forbidden palace/cottages--greater long-term tech potential. If the AI is not going to conquer territory, it should only build military at a sustainable rate and should focus on teching, including defensive techs, and upgrades to protect its empire.

That's fair enough. The Curr/CoL/FP thing is good, but the decisive war thing, like the player does (i.e. collect a hugeish army for the war, use part of it up winning, then go into rebuild mode) relies on winning due to tactics. For two equally matched AIs, they may get into the abovementioned slugfests, leading to two stagnated leaders?

How does a behavioral patch address this, unless the AIs are slightly skewed against the player? (I am not saying it HAS to be addressed, certainly its realistic that nations sometimes might ruin themselves in a long war, but some people blame the AIs lessened performance on this)

I know bonus modifications seem like a cop-out, but remember, the bonuses were heavily lowered for BtS, and the BtS AI might not have been "battle-tested" to the extent it is now when those bonuses were lowered (I'm not saying they didn't playtest, I'm just saying there are even more playtesters now), so I don't see it as a complete surrender if some of the bonuses have to be tweaked.

 

[Via]

Seconded to gettingfat and uberfish. I feel exactle the same.

AI techs slow as hell.

I too have found that ignoring early espionage(not mentioning that one great infiltrating spy), skipping medieval warfare and rushing to rifling is very effective.

One great spy can practically allow the human player to steal 5-6 early techs.

AI does declare wars against each other more often(and more cheaply, if bribing - I think it has something to do with the fact that AIs are more ready to war with all these large stacks).

I play normal game speed, small/normal maps, Emperor/Immortal difficulty. Won most of my 8 BTS games (6emp/2imm) with tech-lead rifleman/redcoat warfare for domination victory. 1 was cultural 1 space race and 1 diplo victory.

 

[Via]

1 more thing - AI should not move the slider for culture so early when pursuing cultural victories! All it does is, that they fall behind in tech and allow you to take their 1-2 legendary cities with your Back To The Future army.

 

[justkilled]

just some random thoughts:

Espionage: well it seems that the AI can handle it to some degree in war times (got my share of poisened water supplys and destroyed buildings). But wont use it in peace times. Thats no problem usually, but it becames one IF the AI investes a good part of it economy in spying. Since you can make good use of Espionage for tech stealing it schould be trained to do so from their near allys (because of distance modifiers / open borders etc). Maybe the tech pacing becames better that way.

Upgrades: I was always wondering why the AIs nearly "autoupgraded" everything when money was there. Granted they get a good discount, but it would be good (economy wise) if the AI would keep their archers for citys in the hinterland instead of paying good coin to upgrade them every time a new wartech arives. It would be optimal if the AI could keep some $ around to do the "hurryupgrades" we use (getting more modern troops out of your old axes in an emergency) instead of directly investing it. But thats daydreams probably .

Flow of an army: the AIs have a steady increase of their military might (mass upgrades aside). Humans might increases more in bursts just before or at the start of a war. Perhaps the warmongers could learn this technic. Growing econ and than suddenly switch civics and start to mass produce/whip/draft a huge army with the just archived military tech for a direct hit... go go Blake go go

 

[futurehermit]

How does a behavioral patch address this, unless the AIs are slightly skewed against the player? (I am not saying it HAS to be addressed, certainly its realistic that nations sometimes might ruin themselves in a long war, but some people blame the AIs lessened performance on this)

Well, when making a DoW decision, the AI should make a power-graph comparison. "Am I sufficiently more powerful?" If yes then attack, if no then don't attack. If no, continue to build up troops following more of the peacenik approach (i.e., don't stagnate economy) but building offensive troops moreso than defensive troops.

Something like that

 

[frob2900]

In general I'd like to point out that the BtS AI handicaps discussed above (50% unit supply, 50% unit upgrade and 80% inflation on Monarch+) seem to be exactly the same as the optional handicap changes supplied with the 01-13-07 Build of BetterAI (i.e. way before BtS was even in the pipeline).

Now, I've played a lot of warlords with the january version of BetterAI/warlords, and the AI performs well. It performs extremely well with BetterAI and the original Civ handicaps (terrifyingly well, actually), and about the same but perhaps slightly better than vanilla with BetterAI and the modified handicaps. So I am confused about what could be causing the extreme drops in AI tech rate that people are reporting in BtS? Espionage?

Again, I'll reiterate my in-principle agreement with those posters who think that behaviorial fixes are better than handicap tweaks, but on the flipside I don't see handicap tweaking as a complete cop-out. If it removes the problem of really slow AI teching then it's a good fix.

Well, when making a DoW decision, the AI should make a power-graph comparison. "Am I sufficiently more powerful?" If yes then attack, if no then don't attack. If no, continue to build up troops following more of the peacenik approach (i.e., don't stagnate economy) but building offensive troops moreso than defensive troops.

Something like that

Yeah, ok, but I do think the power graph is already a major part of the decision making process of the AI. The problem with runaway AI-AI wars (I'm guessing) might be that both AI's go into an arming spree:
-One of them might want to invade, and starts building more troops to increase power, whereas the other starts spamming defenders.
-Or both are arming offensively at the point that one actually invades.

Since both have such large power graphs, the war drags out, mass loss of tile improvements (due to inane maneouvering of stacks and a great love of pillaging) and mass wasteage due to bad tactics, loss of commerce to due WW etc. etc.

They are probably also espionaging each other to oblivion with forment unhappiness and poison water.

The only thing that can break the stalemate is superior tactics by one AI (assuming tech parity). This is what breaks the stalemate between the player and the AI (How many times have you invaded with less power and still made significant gains? I know I have done that a few times). But if both AIs in an AI-AI war are using the same improved tactics then the stalement is principle still there.

 

[futurehermit]

That's true. But I don't know that that is the problem across the board. In some cases frankly I would be ok with this happening. Let's use the example of multiplayer. If one civ starts massing an army will the nearby player mass an army as well at the cost of tech or try to keep teching peacefully? They will mass an army at the cost of tech. If the aggressor won't let it go they WILL stagnate each other.

So, I don't see a problem with that tbh, as long as it isn't consistently destroying the tech pace in each game. If 2 civs get locked in a showdown like that and it destroys them frankly that is something that could plausibly happen over the course of 6000 years of history.

So, I guess the question that factors in here is how pervasive is it and is it the prime candidate?

Maybe a series of smaller tweaks from Blake would be the answer. I think everyone agrees the AI is spending too much on the espionage slider OR that the espionage slider doesn't give enough bang for the buck (meaning if players skimped on it, they would pay the price!).

So, maybe if this was addressed first we could get that out of the way and turn to other prime candidates.

I'd prefer a kind of search like this rather than inserting cheat bonuses that mask what the real problems are

 

[sveint]

Just looked at the savegame. Aggressive AI and you (the OP) are toast.

Any normal-setting game on Monarch+ that doesn't use the "real" AI doesn't count.

In other words, the options are there if you think the game is too easy.

 

[FrancisMarion]

For me at least

Emperor + Aggressive AI = damn near impossible
Emperor + Normal AI = fairly easy

It's hard to tech when you've got to build troops from every city to compete.

 

[frob2900]

For me at least

Emperor + Aggressive AI = damn near impossible
Emperor + Normal AI = fairly easy

It's hard to tech when you've got to build troops from every city to compete.

But then again, AFAIK the aggressive AI is skewed against the player. I.e. the AI prefers attacking the human player over other AIs. "Always War" would be another way to make the game harder.

 

[Unconquered Sun]

Always war will be a way not to discriminate against the human player, but it shuts off half the game options. Haven't tried Agressive AI on yet.

The powergraph is stupid. Part of it is composed of stuff unrelated to successful warring. The two things that count in war are: 1) units 2) unit production capacity.

 

[frob2900]

Always war will be a way not to discriminate against the human player, but it shuts off half the game options. Haven't tried Agressive AI on yet.

The powergraph is stupid. Part of it is composed of stuff unrelated to successful warring. The two things that count in war are: 1) units 2) unit production capacity.

No, always war doesn't discriminate (Aggressive AI does), but "always war" cuts outs chunks of the game, like you said. In that way it is "non-standard", just like aggressive AI (which introduces an unrealistic bias against the player).

I thought the discussion was about flaws in AI behaviour that sometimes make e.g. emperor/immortal/deity easier to play than they were in Warlords.

AFAIK the powergraph is deterimined by the "soldiers" demographic, which is computed by adding together the soldiers in each unit you have (modern units have more soldiers) and each barracks (walls, castle, stable???).

 

[FrancisMarion]

The Aggressive AI in BTS is NOTHING like the Aggressive AI in Warlords. It is no longer simply more biased towards the player. It should be named Oppurtunistic AI.

http://apolyton.net/forums/showthread.php?s=&threadid=168663

There's the link where Blake describes exactly what Aggressive AI is in BTS.