Offensive Drafting

Another move that should improve AI performance came to me recently after I steamrolled an AI using the draft:

Why doesn't the AI do it to mass troops?

Now, unlike the citizen governor, which chooses ludicrous tiles to work when building wealth (and sadly some couldn't do the math to figure out why for whatever reason), the production governor seems to whip OK. I don't agree with its building choices, but that's a lot more complex than what I'm suggesting:

Have it draft units and favor the nationhood civic more...especially in WHEOOHRN. If the governor can know to whip every 10 turns or so modified by the city's food intake vs available superior hammer tiles, I'm sure the AI could handle drafting without stacking draft . AFAIK it doesn't do this in the base game (at least not effectively, or I'd see more unpromoted or just combat I troops in offensive stacks, which I don't see too much), and searching the last 6 pages of thread titles I didn't see anything about it either.

Drafting is very powerful. For maces, muskets, and rifles, the food:hammer conversion is absurd, and with the big empires the AI tends to get, it could get some good use out of it. Even with conservative drafting (no stacking would keep it at a net -1 for drafting...hardly crippling) it could mass 30-60 troops in the early-mid renaissance with only minor concessions to whatever else it builds. Needless to say this would make it more effective, assuming it wants to declare on someone (maybe just make it part of the massing script).

Or is this being deliberately avoided because it's too powerful? I rarely even see the AI in nationhood, let alone doing it.

One thing that might need to be kept in mind... IIUC the Ai gets production bonuses at various levels (or handicaps at lower levels) and drafting bypasses these bonuses (just as drafting for the human player is nerfed on Marathon speed) because it uses food instead. So in effect, drafting is going to be more powerful the lower the difficulty level.

The question is, how valuable is drafting to AIs at the higher levels where the bonuses to production are substantial?

PieceOfMind said:

One thing that might need to be kept in mind... IIUC the Ai gets production bonuses at various levels (or handicaps at lower levels) and drafting bypasses these bonuses (just as drafting for the human player is nerfed on Marathon speed) because it uses food instead. So in effect, drafting is going to be more powerful the lower the difficulty level.

The question is, how valuable is drafting to AIs at the higher levels where the bonuses to production are substantial?

Click to expand...

A better question might be its opportunity cost to the AI. How much is the AI losing in production to draft? Given the fairly ridiculous conversion rates on maces, muskets, and ESPECIALLY rifles, I find it hard to imagine anything but a deity AI could possibly lose out by drafting offensively...and even there I'm not 100% sure it wouldn't get more net units by drafting in addition to standard troop spam.

But as to your point about it being stronger at lower difficulties...isn't that the point of betterAI ---> to make it play better so that it is more effective without its bonuses? Drafting is so powerful that humans have used it to turn a losing game into a somewhat dominant position in the space of 30-60 turns. That's not insignificant, and yet we see nothing from the AI.

Edit: Just as an example, drafting rifles is STILL pretty good on marathon...even ignoring the free "multiplicative military academy"!

Mass offensive drafting is an Immortal/Diety level strategy. Do you really want the default AI using such strats? I'd say yes, but this should really toggle off below monarch, it would be horrible if the AI used this on settler...

phungus, I wouldn't say mass drafting is an Immortal/Deity level strategy. It can be used at any difficulty to great effect.

TMIT, the way you rephrased my question is pretty much what I meant. I wasn't saying the AIs wouldn't benefit from it. I was simply saying the benefit is greater for them at the lower difficulty levels. I'm not using this as an argument to say drafting should ignored - I too would like to see them using it. I'm just saying it's something that should be considered carefully.

The fact a settler-level AI could draft a rifle for the same cost as a Deity-level AI does not sit entirely well with me, and I suspect it wouldn't with Roland either.

And TMIT, believe me I know full well how strong drafting is. Haven't you noticed how much I rave about it, and why I'm pro-Protective, and why I wrote the GTdraft article?

The impression I get with Better AI is that the people designing and making it have hesitated when changes are suggested that purely take advantage of game mechanics. It has been argued many times before that the most fun AI is not necessarily the one that best knows how to exploit the rules. And I'm using the word 'exploit' loosely here - I'm not trying to argue whether drafting is an exploit or not.

jdog et al have mostly been focusing on what could be called tactical improvements and broad strategic goals (IMO). If you wanted to look at the extreme end of the spectrum of things they would oppose, try something like making the Deity AIs rush the human player with their archers. There is no black and white when it comes to suggestions for AI improvements - everything is grey. I'm only saying this because I think there are questions that need to be answered with AI drafting. I'm neither for nor against - I'm on the fence.

what about switching to hurry civs (nationhood, slavery) when losing a war?

Note that at higher difficulties, AI has extra happy, faster growth via food, and faster production.

So drafting gets 'cheaper' as the AI is playing at higher difficulties.

Oh. I wasn't aware the AI got faster food. Is the faster growth proportional to its faster production?

Personally I see no problems having the AI draft at lower levels.

At the moment the AI is set up so ALL strategies are available at all levels.

I would not imagine jdog plans on completely redesigning the AI system so that different strategies come in at different.

Therefore I think there is no need to make a special case for this.

If the AI becomes too good then a simple re balancing of the AI's bonuses/negatives is all that is needed.

Well, PieceOfMind , I think you are not seeing things as they really are ( in spite of understanding your concerns about things don't scaling well with handicaps ). The fact is that the AI is already capable of drafting, in the same way they can whip or $rush , so this would not change a iota of the game rules. The issue is that the AI seems to not be able to draft efficiently ( if at all ), a thing that in it self is a huge handicap to the AI ( I am certain that if the AI could draft as well as it whips ( and they aren't stellar at whipping as well ) most people would have to get down a level atleast ).

PieceOfMind said:

The fact a settler-level AI could draft a rifle for the same cost as a Deity-level AI does not sit entirely well with me, and I suspect it wouldn't with Roland either.

Click to expand...

Talking about this Roland? Talking about me behind my back, huh.

PieceOfMind said:

Oh. I wasn't aware the AI got faster food. Is the faster growth proportional to its faster production?

Click to expand...

Not exactly but close. The difference is about 5% points maximum if you exclude Deity level. For some reason, the deity AI gets a large bonus to production but a far lesser to growth. Check the Handicap xml-file.

I don't know if the AI per era bonuses also apply to growth. I know they apply to production, I just never checked if the applied to growth.



About the drafting and the nationhood civic

This thread has attracted the players who really like drafting which is logical as why would a player who doesn't use it a lot and dislikes it start reading this thread. Still I'm one of those and was just a bit curious what this discussion was about. I tend to read most of the stuff in the BetterAI subforum.

Just because a player really likes using drafting doesn't really make it a good strategy for the AI. For instance, drafting is the most efficient at low population sizes, so if you wish to get that ultimate efficiency of food to hammer conversion, then you need to construct your cities around that idea. This is a special form of city specialisation while the current AI is utterly incapable of any sort of city specialisation. If the AI uses drafting, it uses it in generalist cities where drafting a population point could very well mean exceeding the happiness cap and losing a fully grown town improvement or a mined grassland hill improvement while due to population anger not being able to use two other improvements which could also be towns or mined grassland hills. The human player would generally not use such cities to draft.

The AI is also often closer to its happiness cap because it isn't as good at acquiring happiness resources through trade. This also makes it more damaging for the AI to use this civic.

The drafting mechanic is also very variable in its food to hammer conversion factor. For riflemen, it's good, for mechanised infantry or macemen, it's not so good. The human players who like drafting will tend to use it to get riflemen instead of these other units and thus get the optimum conversion of food to hammers. But to teach the AI that the drafting value in a city is dependant on 'hammer value of unit/population cost of unit' modified by the production bonuses the AI has in the city is hard. (If you have a 225% production bonus on units, then drafting is less attractive than if you have a 25% production bonus on units.)

Next to the drafting bonus, the nationhood civic also offers a nice empirewide espionage bonus. While this can be great for a human player who is using an espionage economy, it isn't that great for the AI who doesn't use an espionage economy at all.

I do think there are a few situations where the AI should definitely consider changing to the nationhood civic and the top one on that list would be the one Cybah already mentioned: when it is losing a war and losing cities or in a situation where it's threatened to lose cities. The AI is often horrible at recovering from losing its attack stack in the starting stages of a war and you regularly get in a situation where the humans attack stack can move from city to city while rarely meeting strong resistance. Because nationhood can get a huge boost in short term military strength at a long term cost (the AI will not be as capable as a human at avoiding a long term cost), the civic is ideal to save a hopeless war where the AI needs troops now. The AI shouldn't consider the long term cost as without using the civic, there might not be an AI civilisation in the long term.

scu98rkr said:

I would not imagine jdog plans on completely redesigning the AI system so that different strategies come in at different.

Click to expand...

He wouldn't need to. All you do is wrap the logic around an if clause for difficulty level. This isn't an economic system that would tie into everything else. Offensive drafting if built right would be set up much like a dagger, where either the logic was active, or it wasn't based on the situation, and you can easily set up a difficult level check for isolatable things like these.

Anyway, here is my main concern with offensive drafting. I've really only seen it used by diety/immortal players using the game mechanics to do an insane rush on the AI that doesn't seem like it should be possible. That or Emperor level players emulating the strategy. Also when ever I've read about it, it seems very gamey, an abuse of game mechanics really.

I've been following the BetterAI forum since it's inception, and the only other time I've been against incorporating an AI behavior was the suggestion to have the AI declare war on the "winning" player. Also seemed gamey to me.

Anway the Offensive Drafting tactic is a legitamate strategy for sure. But I just think players that are playing below Monarch level shouldn't have to face it. Using it correctly a player can create and reinforce an insanenly large attacking force, that isn't comperable to any other means of production. Springing something like this on a noble player will likely make them throw their hands up in frustration. I just don't think making the AI pursue this tactic at lower levels would improve the experience of the game for players plaing at Prince or Below. If you're playing that level, you're not using abusive strategies, and I don't think they want to play against opponents that are either. As for Monarch and above, it might very well be a cool idea to have an offensive drafting logic kick in at that level.

I understand the argument of keeping it off the lower level players, but unless the high level AI were dumbed down further, offensive drafting would make some games insurmountable. Picture an immortal or deity stack, then take away like 4 non-rifle or infantry units, but add 30-80 rifles or infantry and stick that in their SoD.

cap wouldn't be a big hindrance as long as the AI doesn't stack draft anger - nationhood makes barracks give +2 so after 1 draft you're only -1 from other civics.

But if drafting is so strong that the AI can't do it or the game wouldn't be winnable to humans (or extremely difficulty), maybe it's an issue not for the better AI team but for patchers .

IMO siege initiative still rules this game (and if one wants to make a truly tough AI to invade, have it sit back if declared-on and have >10 siege in a counter-SoD...that would make invasions extremely obnoxious), but drafting is one mechanic that comes close to it...

If the changes would make the AI "too good" at drafting, then the following places in handcap.xml bonsues could be lowered to compensate:

1. AI unit supply cost could be increased.
2. AI unit upgrade cost could be increased.
3. AI bonus against Barb units could be reduced.
4. AI WW could be increased.

r_rolo1 said:

Well, PieceOfMind , I think you are not seeing things as they really are ( in spite of understanding your concerns about things don't scaling well with handicaps ). The fact is that the AI is already capable of drafting, in the same way they can whip or $rush , so this would not change a iota of the game rules. The issue is that the AI seems to not be able to draft efficiently ( if at all ), a thing that in it self is a huge handicap to the AI ( I am certain that if the AI could draft as well as it whips ( and they aren't stellar at whipping as well ) most people would have to get down a level atleast ).

Click to expand...

No, no, I'm aware of the game rules for the AI in this case. Perhaps I didn't really phrase it properly - I'm not concerned that Settler and Deity AIs can draft for the same cost (though it turns out they can't anyway). I'm concerned that if they make heavy use of it, it would boost the Settler level AI more than it would the Deity level. Since I've been told the AI bonuses to food and production are similar at most levels, this concern is lessened.

By the way, I'd be inclined to give the AI more encouragement to draft if the unit it can draft gets free promotions, or is a good unit anyway. Protective and Aggressive leaders should be more likely to draft gunpowder units, and Agg leaders more likely to draft melee units. I'm not sure if any archery unit can be drafted but I don't think any of them would be good value anyway.

I would usually suggest that drafting only be done from cities with barracks except in emergencies.

As has been mentioned, drafting comes with making use of the Nationhood civic. Just how beneficial is zero upkeep cost to the AI if it gets a large reduction anyway (I'm not sure on the discount) and how useful is 25% if it doesn't know how to use spies or an EE properly?

AIs that use Free Speech tend to do pretty well in the late game. And if workers are (hopefully) improved to the point where they are not frequently building over towns, leaving alone FS may be a substantial opportunity cost.

I don't really support changing tactics with difficulty level. I'm still a believer in game-difficulty being more a slider of various handicaps - not the AI tactics or intelligence.

How many low-difficulty players are going to complain of superior AI tactics if they are seeking a Better AI mod in the first place?

AIs that use Free Speech tend to do pretty well in the late game. And if workers are (hopefully) improved to the point where they are not frequently building over towns, leaving alone FS may be a substantial opportunity cost.

Click to expand...

AIs that are twice the size of the average empire also tend to do well .

I agree with your points however.

If the changes would make the AI "too good" at drafting, then the following places in handcap.xml bonsues could be lowered to compensate:

1. AI unit supply cost could be increased.
2. AI unit upgrade cost could be increased.
3. AI bonus against Barb units could be reduced.
4. AI WW could be increased.

Click to expand...

I'm all for making the AI bonuses increasingly similar to the human as the AI is improved, but doing it in one go like this is precarious and likely quite difficult to balance.

"He wouldn't need to. All you do is wrap the logic around an if clause for difficulty level. This isn't an economic system that would tie into everything else. Offensive drafting if built right would be set up much like a dagger, where either the logic was active, or it wasn't based on the situation, and you can easily set up a difficult level check for isolatable things like these."

Im not saying it wouldnt be possible or even difficult. Im just saying it would be strange having one and one only AI tactic not available at lower level.

I think if jdog planned to go down the route of removing AI tactics at lower levels then he should need to consider other tactics. I just think it would be easier and a better use of time to rebalance the bonuses if we find this needs to be done.

Also what level would you put it in ? Frankly the actual levels of difficulty can vary enormously depending on what sort of game you set up ie barbarians/tech trading/size/number of opponents.

Personally I think it would need to be in at least at prince if not noble. Generally I think most people agree its much more entertaining to play an intelligent AI with less bonuses than a stupid AI with more bonuses. In fact I think thats the main point for including everything at lower levels.

"its much more entertaining to play an intelligent AI with less bonuses than a stupid AI with more bonuses"

For the most part that's true. But offensive drafting is kind of a whole nother level of tactics, that is borderline abusive. I just don't see a noble player finding it fun to have the AI run over them with massive infantry stacks.

Anyway I'll quote myself again since you didn't adress my point:

phungus420 said:

Anyway, here is my main concern with offensive drafting. I've really only seen it used by diety/immortal players using the game mechanics to do an insane rush on the AI that doesn't seem like it should be possible. That or Emperor level players emulating the strategy. Also when ever I've read about it, it seems very gamey, an abuse of game mechanics really.

I've been following the BetterAI forum since it's inception, and the only other time I've been against incorporating an AI behavior was the suggestion to have the AI declare war on the "winning" player. Also seemed gamey to me.

Anway the Offensive Drafting tactic is a legitamate strategy for sure. But I just think players that are playing below Monarch level shouldn't have to face it. Using it correctly a player can create and reinforce an insanenly large attacking force, that isn't comperable to any other means of production. Springing something like this on a noble player will likely make them throw their hands up in frustration. I just don't think making the AI pursue this tactic at lower levels would improve the experience of the game for players plaing at Prince or Below. If you're playing that level, you're not using abusive strategies, and I don't think they want to play against opponents that are either. As for Monarch and above, it might very well be a cool idea to have an offensive drafting logic kick in at that level.

Click to expand...

Is the AI currently drafting? If yes, when?

If the changes would make the AI "too good" at drafting, then the following places in handcap.xml bonsues could be lowered to compensate:

1. AI unit supply cost could be increased.
2. AI unit upgrade cost could be increased.
3. AI bonus against Barb units could be reduced.
4. AI WW could be increased

Click to expand...

This would make the game a lot easier if the AI is NOT at nationhood civic. And most of the AIs NEVER choose nationhood.

Phungus

Im still not convinced. Generally most people play on a level were they are near the top of the score board and the most difficult phase tends to be the start phase. By the later game the human tends to start to pull away from the AI. This has been one the criticisms of CIV since time immemorial that the end game tends to be boring. The game is won generally in the early / middle stages and the AI can never come back at the end and in fact the end stages are thus BORING.

At lower levels the AI generally wont be getting Nationalism too early in the game. Therefore Frankly anything that improves the AI at later stages even if it gives a novice player a surprise is a good thing in my book.

The key statement in your post is this

Using it correctly a player can create and reinforce an insanenly large attacking force, that isn't comperable to any other means of production.

With the key words being "the player" even if the AI can use offensive drafting I still very much it will be able to use this force in the same way a human player can.

The current problem with the difficulty level is that the human tends to pick a level where they will fall behind to begin then slowly catch up and then often easily beat the AI in the end stages.

In an ideal game you and the AI's would remain completive all through the game therefore increasing their abilities in the middle/late game is very important.

Another problem with your argument relates to Multiplayer, you suggest no drafting below Monarch. However in Multiplayer the AI gets Noble bonuses and the humans bonuses change depending on the level they choose hence all strategies must be available down to noble level.