When Elves Ride Dragons

[Nelphine]

In this situation it's about how the AI weighs declaring war or actually attacking a town.  For example, even when I was at peace or had treaties I would have doom stacks sitting right outside my cities, even well defended ones.  there are times the AI doesn't comment on a unit wandering close to the city, but others where they threaten due to a close unit (I think it's based on if the unit can see their garrison), but does it have any effect on diplomacy?  And is that effect based on the unit or how many cities it can see?

If you have a wizards pact, then any unit within 2 tiles of their city will trigger a warning.  You then have 1 turn to get all units at least 2 tiles away from all their cities, or they break the wizard pact (and you take a diplomacy hit as you are considered to be breaking the pact.)  Note that the AI has NO restriction for how far they can be from the human cities. Instead, they literally cannot attack you or target you with curses if you have a wizard pact. Asymmetrical diplomacy!

Then there's the situation I most recently described, I had a bunch of spiders which were enchanted with resist elements and water walking.  They were able to 1v1 the cockatrices, but obviously weren't weighed as heavily - I assume due to web not counting at all, there was also a stack of focus magic cockatrices which apparently doubled their combat value, but solo cockatrices with no enchants were coming in too.

Strategic combat! Go back 3 pages, and you'll see a thread named strategic combat by me.  It's got a lot of the mechanics of strategic combat in it. However, you might not actually care about that, and instead only care about the overland AI decision making process.  The AI has no idea what 'an ability' is. First strike, or web, melee attack, or magic immunity.  However, strategic combat weighting gives us a (very rough, and occasionally innacurate) idea of the raw stats of a unit (note that in the original version of the game, this was so bad, that AI spearmen could literally kill sky drakes. There are no huge problems like that left; units with 4+ figure and numerous buffs/experience levels are the only real issue left, although I personally think thrown/breath is still weighted too high, while first strike is weighted a bit too low; but that's getting super specific, so ignore that.). This includes all melee/ranged/thrown/breath - but it doesn't differentiate between magic and normal ranged. It includes about a dozen abilities that are 'fairly' straightforward (such as first strike and destruction, but not things like web, bless, or casting skill of a unit). It includes armor and hp, but on a completely linear scale that has nothing to do with number of enemy figures, and so multi figure units tend to be weighted much more heavily than a single figure unit of roughly the same stats.  With buffs this can get excessive (*cough bezerkers which the AI think are stronger than any very rare unit, let alone anything weaker than the very rares cough*).  The AI uses this strategic value overland for .. many decisions - such as, it won't attack garrisons with a stronger strategic combat value than its own stack; the strategic value of units is the basis for the army strength graph, and in turn, this is the basis for a number of diplomacy decisions. (For instance, if the AI is much stronger, its inclined to accept peace treaties; if the strength is roughly equal, the AI will declare war; if the AI is much weaker, it will beg for peace.  Note that while these decisions don't seem intuitive on the surface, the basic idea is that if the AI is strong, then in peacetime, it's liable to get stronger faster than its enemy, so peace benefits it.  If its weak, then war will result in it dying.  And if its equal, then its opponent is probably in the process of becoming stronger, so attack now before its too late.)  This is not the only factor in these decisions; these are just decisions that the army strength graph influences (in some cases it is the primary factor though, thus my bezerker strategy.)

Does having a large number of cities with a small guard each differ from less cities but the same army size?

Not for diplomatic decisions; but for any given AI stack of troops, they don't care about your overall strength, they simply care about finding a target weaker than themselves.  So it depends on whether you're looking at the macro or the micro situation.

How big of a hit is attacking a wandering settler?  Attacking an outpost is instant war, but attacking a settler just leaves them disgruntled - but how is it vs attacking a doom stack?  Is there -really- a hit in relations, or do they just talk about it, and if there's a hit is it based on what they lost/how many/cost/etc.

There is a hit to relations. Repeated attacks on any kind of troops will result in war. I don't know the exact numbers offhand. I don't actually know if doomstacks are more important to the AI in terms of diplomatic decisions.

What situations will cause the enemy to actually attack the town?  Do they have to be some amount over the defending garrison in the numbers?  Are invisible units considered in that?  I find most of the time I end up being the one attacking because I end up with one or two big defensive forces, but can't have huge defenses in every town.

3 things. First, the diplomatic situation must either be hostile or war. (Hostile means at peace, but the AI doesn't like you and so is willing to attack or curse you. No, there's no in game way to tell if the AI is hostile or peaceful if there are no treaties. But it's an intuitive process - if no treaty for a while, and he's uneasy, then your rating is probably going down each turn; therefore if its been 'enough' turns, he's probably hostile.) Second, the AI tries to pick a target for every single stack (completely independently; it has no idea that it has multiple stacks when it does this, which makes its overland strategy a little clumsy at times, although Seravy has done a fantastic job so you don't notice it often. You just see hordes of armies all over, and feel threatened.) Picking a target is based on the importance of the target vs the distance to the target.  Built up cities are more valuable than outposts. Large numbers of troops are more valuable than small numbers (based on strategic strength). Fortresses are super valuable. Finally the AI checks if the strategic strength of the target will allow its current stack to attack the target.  So it might want to attack, say, your fortress; but assuming you have a decent garrison, even though the fortress is a valuable target, the AI won't send a weak stack after it, because it assumes it will lose.

This basically* results in the AI attacking what is intuitively the most important target, with a stack that is powerful enough to threaten it.  Less defended targets get targetted by corresponding weaker AI stacks. *Note we've just finished a discussion on trying to determine what exactly the relation between different targets should be, as well as trying to decide just how important distance is.

Outside of the diplomacy thing, does resist elements apply to stoning touch/gaze or does it base it off the creature it's coming from?  (as in, does it get the bonus from an undead gorgon, or are stoning/life drain/destruction attacks not elementally aligned?)

Attacks are elementally aligned.  Stoning for instance is a nature attack, while I believe poison is neutral. I don't know exactly which attacks are affected by resist elements (the resistance part)/bless off the top of my head.  Resist elements (the armor part) should affect all magical ranged attacks; but note that something like ghouls have magical ranged attacks that also include poison - so you could resist the attack itself, but get no bonus against the poison. (This also holds for magic immunity - ghouls can still kill your magic immune bezerkers due to their terribad resistance.)

Does upkeep cost have any bearing on the army rating?  how about regeneration?

No. Strategic value looks at right now, as the assumption is you have enough income to deal with all your maintenance; and regeneration is too complicated to be included in strategic value.

I took a look around but the only thing I easily found said to check the wiki, which doesn't really say much other than 'prodding the enemy wizards makes them dislike you'.

Exactly! Do bad things, they don't like it. Seravy has actually gotten very close to making that statement true, and all the ins and outs don't really need to be defined.

Another is the combat value of + to hit, fire breath and immolation (which I think works like a touch attack?  As in I think it triggers on the 'ranged' round, like firebreathing, and the 'melee' round)

Immolation is too complicated to be included in strategic value. +to hit and fire breath are both in the thread I mentioned.

Just to further clarify: it's not about gaming the system, it's more about which puts us in the better diplomatic position: slingers, war bears or hell hounds?  Then when war happens anyway, which is the better deterrent?

Uhh, they're all super low so I have no idea? Buffed/high level Slingers would definitely be the best due to multifigure problems with strategic value, but without buffs/levels.. they're just bad.

Something else I don't know; what do walls actually do?  They limit avenues of attack, but they also offer some bonus against ranged, right?  It doesn't say what that bonus is, only that it exists if I remember correctly.

I want to say +3 armor against ranged attacks, and either +1 or +3 armor against melee attacks, for any units inside the walls that are being attacked by units outside the walls.  Don't hold me to those numbers. It's noticeable though.

[Nelphine]

Started a Master difficulty game with: 10 Death, Sage Master, Specialist, Halfling. Get me ma Demon Lord!
Very first node I conquered (3 naga/3 phantom beasts; no lairs or cities attacked before this, so literally the first thing I did) got me a Sorcery Book!!!! *so excited*

@Seravy: First AI I encountered was nature with a bit of death. He didn't seem to have many ghouls. In 1405, an AI I never encountered banished the first AI. I'm assuming he had a pile of sprites in his fortress, since he hadn't lost a single city that I had found, nor had his graphs been noticing changing at all, so I don't think the second AI was just killing all the first AI units it could.

[Arnuz]

Catapults: whaaaat? I like them a lot, it's a quick unit to build that takes no maintenance and can be a decent defender against raiding parties.

Tlavoc: I don't remember by heart the spiders' resistance, but did you consider resist elements on them? Once they beat the -3 of the cockatrices' stoning they should eat them for breakfast with their webbing. Webbed cockatrices stay put for 2-3 turns.

[Seravy]

Considering Cockatrices worth a lot to begin with in that scale and with Focus Magic they are worth double, that's not all that surprising. It doesn't necessarily means they have twice as many of them, A Cockatrice with Focus Magic is worth as much as 4 (or was it 5?) Berserkers in total army power.

Is there anywhere that talks about this stuff?  I just don't want to pester you too much about things I could read up on.  It seems like something worth knowing since it has a fairly large effect on diplomacy and I always have a low-rated military strength.

One of the strategic combat threads. The astrologer, AI decisions and diplomacy uses those unit ratings.

Stoning touch is worth a lot (as it only appears on the cockatrice with has -4 save), poison is worth something but not nearly as much.

Diplomacy is explained in the Master of Magic wiki - Caster uses the same system as the 1.50 patch although some formulas have been tweaked. (especially the difficulty modifiers)
http://masterofmagic.wikia.com/wiki/Diplomacy

In general if they talk about something that did reduce relation. The message with the highest relation penalty is displayed if multiple things happen in a turn, and the chance is proportional to the amount - if it's a smaller penalty they might not mention it.

Yes, Resist Elements applies to stoning touch and gaze.

Upkeep has no effect on army rating. Regeneration doesn't either.

If you want to know the specific unit ratings, there is a unitrate.exe tool in the mod. If you run that, and redirect the output to a file, you can have a list of the already calculated ratings of each unit type (tho if they gain levels or get buffs the value will be different)


Walls give +3 defense against any type of attack from outside the walls to the inside (including melee). Destroyed walls give +1 defense.

Killing a settler has a higher penalty than killing another unit, but killing 9 other units will be more than 1 settler. Heroes have the highest penalty, I think -30 for each one.

Ranged units have their ranged rating doubled when the AI decides to attack a target or not. So they are better to scare away the AI (and are better in actual combat too). Of course if you have a super strong melee unit that's worth more than double of your magician/bowmen, use that instead.

@Arnuz
I had time to think about the AI advantage being scaled by cities and it's a bad idea.
The AI gets this bonus because it doesn't follow a coherent strategy and can't focus its resources towards a common goal. The number of cities has nothing to do with it.
Meanwhile, needing more troops to protect more cities is already covered by basic game mechanics - the cities themselves can produce those troops and the resources for them. Even Casting Skill is now covered thanks to Amplifying Towers - this is why it exists in the game.
If the problem is too many Nagas and whatever attacking, then it's a problem with the doomstack procedure being able to use garrison troops for offense. We might need to rethink if we really want the doomstack building to fetch garrison units and allow leaving conquered cities. Although if it's causing problems for the player, that's a sign that the units are more relevant on offense, meaning the AI does the right thing by using them that way. In fact, the doomstack(s) mean the player has to be able to commit a large force to garrison the conquered city, otherwise the doomstack can take it back. It's only a problem if the player is razing the cities. I'm still not sure if the current doomstack behavior is the best but for know I have no better idea.

[Arnuz]

The behaviour seems perfect to me. I want aggressive AIs. The problem is just that there are way too many units on attack, given that the AIs don't need to garrison many cities in year 1 to 3 and can abuse the advantage to create doomstacks. So, the game becomes finding a way to kill the doomstacks as they come, and you know they'll come because of the impossibility to garrison your cities as much as the AIs, 1 barbarian city strategy excluded. Ghouls + web/earth to mud is perfect for this because it lets you even abuse the AI cheating advantage by getting hordes of critters that you can banzai into nodes, lairs or other AI cities... Quickly because that maintenance hurts

So, 1 barbarian city avoids the attacks, ghouls abuse the cheating strategy, draconians lol at all those silly critters on the ground and peddle them with arrows, and nomad horse archers run in circles around them. Have I forgotten anything?

Making the cheating city based is one idea, another could be making it year based. Sure, the start is important, but isn't it for everyone? Then we can nerf a bit these 4 strategies, and voila', a lot of many other strategies can become relevant to master/lunatic.

Finally, don't forget that what I'm trying to achieve is the same that you are:

I'd say Normal is for first time experience if you have played the 4X genre before - if not, then Easy.
Advanced is probably good for the next few or at most a dozen games, until you learn the races, spells and realms.
Expert is somewhat difficult - you either need a powerful strategy or know the game and play well to win.
For Master, you need both of the above, Lunatic you need both plus a lot of luck.

Currently, you need luck in expert too, because if those 4 stacks reach you because of no geographical boundaries and you don't have one underdog strategy you're done for - see the halfling Tauron test that I reported.

[Seravy]

Making it year based makes more sense but it's also wrong.
That favors faster strategies over slower, while we want the exact opposite - as is, fast strategies are already more powerful and slow strategies kinda weak.

The main problem I see here is the AI has to have enough resources to be able to summon up 9 garrison units for the capital as early as turn 15 otherwise "barracks first" - to use the warcraft term for producing military first instead of building a city - kills it. But if it has 9 units on turn 15, it'll have 18 on turn 30 and 27 on turn 45. So if the AI decides to go to war with the player on turn 45, they'll have 2 doomstacks.

Basically, due to fast movement and production times, everyone is forced to go military first or summons first - if they don't, and anyone else does, that anyone else takes them out of the game.

This definitely isn't good but what can we do against it? Fortress Lightning helps but it won't stop a stack of 9 anything decent when the AI only has 5 swordsmen in their capital.

But there is more than just the capital. Lightning only protects the capital and a wizard with nothing but the capital is as good as dead. They need additional forces to be able to keep their outposts in play - either by pulling those units into them as garrison, or more likely, to have them as a threat so the player won't attack. Yes, you can steal that 1 pop city from the AI - actually, all 3 of them - but the price for that, you have to be able to deal with their 10-15 summoned creatures.

...so my point is, unless we figure out how to alter the game in a way that early AI board presence becomes unnecessary, and do that without completely removing early strategies as a side effect, we can't do a thing against those Naga doomstacks.

I agree we have the same goal, but I don't see a good solution to it yet.

(PS : In the base game everything was so slow, this wasn't a problem - by the time the player could have 9 units and reach an enemy capital, the AI had a massive size capital with powerful units because it kept receiving the 300% bonus for all the 50 turns it took to get there. But now you can get there in 15 turns including the summon/production of your stack...)

[Arnuz]

The main problem I see here is the AI has to have enough resources to be able to summon up 9 garrison units for the capital as early as turn 15 otherwise "barracks first" - to use the warcraft term for producing military first instead of building a city - kills it. But if it has 9 units on turn 15, it'll have 18 on turn 30 and 27 on turn 45. So if the AI decides to go to war with the player on turn 45, they'll have 2 doomstacks.

This is exactly what scaled cheating addresses:

• 9 units are necessary, you say? OK! Let's make that the base. AI casts 9 units by t15.
  
• A doomstack is also necessary? Good! That's then also part of the base. AI is given the resources to cast/build 18 units by t15. I'd make this is for advanced, already.
  
• 2 doomstacks needed at higher difficulty? OK! AI has enough resources to cast/build 27 units by t15-24. (I'd say 22 for expert and 27 for master, give or take)
  

We do some testing and assess what's the cheating needed for these numbers, and make that the base K of the formula. Then, any further summoning is based on number of cities or passed turns (I don't care too much which one tbh): so, cheating=D * (K+n), where D is the difficulty factor. Simple and effective, and assures the principle of AI not defeating itself at start.

On the power of early strategies: those are based on units that are strong early. So, again, I'm proposing some small nerfs here or there to address that as well. I don't believe that you can ever achieve overpowered cheating at start that does not make the game a dice roll decided on geographical boundaries, and I hope we all agree that that's bad?

[Tlavoc]

Wizard Pact: Does corruption/raise volcano count when it comes to curses? I'm fairly certain a guy I was in a pact with was tearing up my gold/food tile bonuses in one game. And what about undefended cities? I -think- a guy I had a wizard pact with took one of my undefended towns. Did he just suddenly decide to go to war with no warning? I could have just had no treaty with him (peace) but I could have sworn I had a wizard pact.

Also, thanks for the info on 'staying too close means I broke the pact' - I thought if THEY were the one to break the pact, as in telling me they are breaking it rather than me attacking, I wouldn't take a diplomacy hit.

Strategic Combat: All of that was really useful. It gives me a lot to consider. Going to war and protecting investments is costly. If a couple of hellhounds are better at deterring attacks than a couple of war bears it's good to know.

Attacking Troops: I figured there would be a hit to relations, but I'm curious on if it's a bigger hit if you kill 5 troops in a stack or if it's just the fact that you attacked. For example, if you attack and LOSE without killing anything (flee) is it a diplomatic hit?

Attacking a town: Thanks for the info! And yes, I DO feel threatened when my kingdom is covered in enemy units - sometimes to the point that I can barely move due to the shape of the land. One game I got irritated of my 'allies' squatting on my land so I built a 'wall' of skeletons. I've found killing the lairs asap helps with this.

Elemental Aligned Attacks: I wasn't sure if the attacks were aligned based on the attacking unit or the type of attack. I assume fire breath is always 'chaos' even if it comes from a draconian? And resist elements says +4 defense against ranged magical attacks, so I assume it gives bonus armor against true magic ranged attacks, so should give bonus defense against ghouls (resist elements swordsmen should dominate them with the large shield bonus).

All the ins and outs don't need to be defined: I agree we don't really need a LOT of detail, but having an idea of how far you can push the opposing wizard is useful - and the info about strategic combat being the primary thing they care about.

War Bears/HellHounds/Slingers: The reason I brought them up is because they all have the same upkeep, but based on the info (I'll be looking at the strategic value thread after this) it would seem hellhounds are more valuable - 4 figures with decent attack and fire breath. Even at low rating, when you stack up 9 of them vs 9 of the others the difference will start to add up.

Wall: Thanks for the info. I had always thought walls were useless unless you had a strong 'defender' at the gate. It never gave numbers so I assumed 'protects your units' was just about the gate thing.

Arnuz: Yeah, once the guy declared war I started stacking resist elements on all the spiders, and they DID murderize the cockatrices, but 1v1 when he was spamming ice bolt made them take a lot of damage each combat, and the guy had TONS of mana crystals so he could just keep it up forever! It meant I took out something like 7 cockatrices, some with focus magic, without losing a single spider, but my arachnid navy was beaten to hell so as they just kept streaming in I couldn't do much.

Seravy: Thanks for clarifying all of that further! Ranged units = double is worthwhile, generally they get 2 'free' shots before melee engages anyway.

Doomstacks and Cities: I currently raze most cities I conquer because I simply can't garrison them in a timely manner, and since the AI tends to have a million units running around overland they just wander back in and take it back. About a million units running around... As things get later in the game, the time between turns takes FOREVER. Any suggestions on how to fix that? I assume it's something on my end like some option or setting I've missed, but even 'not watching enemy movement' just means I see the non-moving screen for a while as the AI moves everything. I'm putting a point toward chaos for overland meteors - killing all the overland units is a great way to speed up the game.

Addressing the strategies; Is it possible/would it help at all, if the AI units/spells had higher upkeep? It would let the summoning/build cost remain the same, but they wouldn't have huge overwhelming armies AND huge amounts of stocked resources (seriously, doing really well, taking a neutral town with a gold and silver vein VERY early only to find the opposing wizard with double your army and triple your resources is a bit annoying). It might also mean taking out a large city would actually affect the AI player, rather than what it is now where you raze three towns and the only effect is there are now settlers scurrying about again.

I once tried draining a wizard's resources through sending a bit of fodder at his units one at a time so he would cast spells, hoping I could drain enough that he wouldn't be summoning/casting when I arrived with a larger group, only for a single town to be generating enough income, on top of a huge number of ghouls/focus magic to be able to make up for more than all the magic he was tossing around.

[Nelphine]

Problem with maintenance is that the AI doesn't know it exists. So the 'attack with mass numbers of single spearmen to drain mana' was a huge issue in.. Some earlier version of the game, because the AI wouldn't recognize it as a problem.

Similarly, the AI won't ever change its casting priorities - it casts spells to be a threat, so if its maintenance is too high, it won't stop casting. It will just make the maintenance worse.

That being said, we have 2 issues - one, the AI economy is too robust, but where should the line be? Two, the AI summons too much, but again, where should the line be?


@Seravy: what if we make the cheating bonus not increased based on cities, but rather, decreased based on number of units NOT garrisoning cities/nodes? Ugh, starting to get too complicated again. (I was thinking something like each city would allow ~9 offensive units (and fortress counts as two cities) after which cheating bonuses would start to decrease.

[Tlavoc]

Just read through a bunch of the diplomatic stuff, I'll have to look over it again later because there's no way I'm retaining all of that in one read-through. It looks like if you are within 2 tiles of a wizard pact city for too long they'll break the pact, but it does count as -them- breaking the pact, which reduces relationship with them but no permanent damage to hidden relation (so long as I'm understanding/remembering correctly)

I was just trying to think of a way to reduce the stupid amounts of resources they build up. Trying to bribe a wizard on expert pretty much just doesn't work because they have so much money. You have to give them spells.

So... if it has enough summons/enchants to be at -20 mana income, it will continue to summon more critters, which kills the ones it currently has?