Caster of Magic Release thread : latest version 6.06!

[Seravy]

Thanks.

Next question:

General question regarding late game.  My opponent just cast 4 earthquakes, 4 change terrains, summoned 2 stone giants, summoned a great lizard, cast transmute, cast blizzard, and cast at least 2 more spells (based on the time elapsed when nothing seemed to happen between spells, so probably casting unit enchantments).

I just feel like this gets excessive. Is this kind of absolute mass spellcasting really want we want late game?  Yes they can't focus on things, but when they cast that many spells.. if there are 2 or 3 of them doing so, the entire landscape can change between one turn and the next, which seems more than is desired.  Strategy shouldn't completely change every turn (I think).

Especially since this is in 1417, after a fairly brutal war (my opponents overall strength on the History of Wizards Power graph has gone down 25%), which by your own admission isn't even meant to be the end game yet (that shouldn't come until 1420+).

I believe you are talking about the wizard who owns 35 cities and has Uranus' Blessing on all of them.

This one :
"Ruthless Theurgist, 4 Nature, 4 Sorcery, 1 Life, Archmage (argh), Runemaster (argh Aether Binding), Conjuror, Lizardman."

So we have a wizard with both skill boosting spells (AEther Binding and Uranus' Blessing), playing what seems to be currently the strongest AI template (Sorcery+Nature, seems to come up on top in most of the games probably due to the absurd strategic power of focus magic on cockatrices, or just the high power of nature summons in general paired with low costs), casting at a reduced cost (Conjurer) and having a skill raising retort (Archmage) which makes the AI more willing to spend on skill, and playing the race that expands the fastest (Lizardmen). On top of all that they control about 2 thirds of the world - you said 35 cities, but on fair map size.
Why am I not surprised? They have an absurdly high skill because they are playing a strategy that wins through that, and they are in the position to win. If a human player controlled 35 cities, they'd win too and they'd have lots of skill from Amplifying Towers too.

[Nelphine]

Yes, I'm not talking about how powerful he is - I still think I'll beat him, although it's painful. My question is about the actual amount of spellcasting going on.

Let's assume for a moment that your goal was met, and that games never ended before 1420 (which I think is unrealistic, as I've mentioned). Then most games might go to 1425-1435.

Yes, I have an absurd opponent. But lets say it was a slightly weaker opponent.. 10 years later. They'd be doing the same thing anyway. The absurd opponent.. well.. I don't really want to think about that.

Actually lets think about that. In 3 years, this opponent managed to lose roughly 7 amp towers with uranus' blessing (that's 98 casting skill). However, his casting skill didn't change. So in 3 years, he actually produced 98 casting skill. Any archmage sorcery wizard will build casting skill at roughly the same rate as this wizard. They aren't particularly rare. Maybe not every game, but every.. 4-6 games? probably?

So call it 20% of games have an AI capable of pumping out 98 casting skill in 3 years. That means by 1425, that AI is going to have 500+ casting skill, without amp towers. They'll probably have uranus blessing and amp towers of course, so call it 700+ casting skill.

A less crazy wizard.. will probably have 350 casting skill, and normal amp towers. Call it 500 casting skill.

So in your ideal game, a NON crazy wizard like mine will be able to do exactly what mine is doing. The slightly uncommon crazy wizards will be doing 50% more than that. (Imagine if one of them had spellweaver instead? That's much worse than archmage.)

And then lets look at another AI that 'only' has 250-300 casting skill (which almost all AI will have by 1415, and even the ones that do extremely poorly, but don't die, will have it by 1420). With extreme AI bonus that still lets them summon 2 or 3 stone giants, and cast 2 or 3 other spells. Every turn.

Is that the quantity of overland spells you want in your game? It genuinely feels like the entire overland map can change between hitting 'next turn' and actually playing the next turn. My entire empire can be changed (terrain, minerals, productive cities, helpful nodes, where strong stacks are), my opponent's entire empire can be changed, all in that time. The only thing that won't change is where my opponent's fortress is (of course, including spell of return, even that changes constantly). Without accounting for a single attack made by the AI. The strategic depth of the game gets lost and it becomes 'wipe out whatever you can reach, because anything can be good or bad, so it doesn't really matter what you actually hit'.

[zitro1987]

My thought - I like the concept of extreme overland casting skill late in game - your mana/gold has to go somewhere and you need skill to keep up with extreme unit recruiting from 20+ cities. Without Amplifying Towers, this was not possible and I was left getting a gorgon every 3-4 turns while in that time for 1 gorgon, I got 10-20 advanced units.

[Seravy]

That means by 1425, that AI is going to have 500+ casting skill, without amp towers. They'll probably have uranus blessing and amp towers of course, so call it 700+ casting skill.

A less crazy wizard.. will probably have 350 casting skill, and normal amp towers. Call it 500 casting skill.

I usually have 300-500 skill in 1420+ myself. I don't see a problem.
Considering a very rare creature costs 1000 on average, that much skill is necessary to use them effectively. (waiting 12 turns for a single Sky Drake is no longer a viable strategy as was in vanilla MoM)

'wipe out whatever you can reach, because anything can be good or bad, so it doesn't really matter what you actually hit'.

Unless the mentioned 2 Stone Giants happen to appear exactly in the one city you targeted, the rest doesn't matter. So what if they used Change Terrain on those cities? 1-2 higher max pop? No one bases decisions on that. Other city buffs get removed after you conquer it so they matter even less (unless it happens to be a spell that stops you from attacking it - unlikely, by the time you arrive those will usually be on all cities already)
What else...Earthquakes? Those affect you, not the enemy. You should be prepared to use a strategy that isn't reliant on a small amount of production centers that can get wiped out if you fight a wizard that has it and the personality to use it often.

What you talk about would be a reality if cities were all unique but in the late game they are not - all of them are maxed out and are the same thing, unless there are some inferior races like barbarians involved. So in the end the only thing that matters is what your units and the defending units are.

Is it "necessary"? Let's see...the wizard has 35 cities. The AI will want exactly two of their best units per city. They spend one third of their summoning power on this goal, the rest goes to summon on the capital, or the designated "primary summon location". Let's say the best summon the AI has costs exactly 500 MP. With 35 cities, they'll need to be able to summon 3*35*500=52500 mana worth of creatures. Assuming the AI spends half their casting power on summons (fair assumption, the spells are expensive so even if other spells are used more often, summons will still eat the majority of skill), it means they need a total of 105000 casting skill, and assuming we want those creatures to appear in a reasonable amount of time, say, 100 turns, that means the AI has to have 1050 casting capacity = 525 actual skill on Extreme. This sounds about right, and as half of that is being provided by the cities themselves, this scales fairly well with empire size. Large empires might fill up a bit slower on creatures than small empires, but in exchange the "overflow", extras at the capital and primary summon point will be proportionally larger, representing the "larger enemy=larger threat" principle well. Slower empires will fill up a bit faster on garrisons, but have much less leftovers for offensive stacks.
Note that 100 turns is actually way too long, and I'm using rare creatures for this calculation instead of very rares. If anything, skill might be insufficient for the AI's needs, but I used Extreme numbers. On impossible the AI gets both more power and skill.

Think about it this way : for the human player to have a stack of 9 great drakes, they need to summon 9 great drakes. For the AI, they need 2 for each of their cities, 9 for their capital, and 9 for the stack, so a total of, assuming 35 cities, 85 of them, almost 10 times more. Compared to that, the AI overland casting bonus is actually not that high (2 or 2.5 times more casting).

It's a bit of a different story for other spells, but those typically cost so low skill that they can be "maxed" even without any AI bonus. Spells like Famines or Prosperities are limited to one per city and cost like 100-200 each. Total cost is a few thousand to cover the entire map with the spell. Having 200 or 500 skill won't make a difference on those.

Out of the spells you listed the AI used, only Stone Giants are expensive. He could have cast all the others without a skill bonus.

Anyway, one thing to point out : Sorcery is a late game realm. And it's strong against Life. You should always aim for elimination Sorcery wizards before they reach this point when playing mono-life. (most strategies have weak points. A strong Sorcery wizard on the other plane is Life's weakness.)
You're still doing fairly well because he only has 4 Sorcery books. If he had Spell Binding or Time Stop...I think you can imagine what would have happened.

[Nelphine]

Right, and I think the human player with 300-600 casting skill is in a good place. My issue is that AI, in order to summon lots of things, have to be able to cast a ton of spells; but again, just like with production, they are actually good at all other spells now (except maybe unit enchantments).

An extreme (or *shudder* an impossible) AI gets such a huge boost to overland casting skill, that they cast a massive amount of spells. And there are 4 of them.

I'd say it's reasonable that in 1420, if the game isn't close to over yet, you'd be opposed by about 1200 casting skill. Whether thats on 2, or 4, AI, that's still a massive amount of overland spells cast every single turn. By 1425, it could be 1500 casting skill. (In my game, in 1415, i was opposed by about 950 casting skill)

On extreme that's effectively 3000 casting skill. Even assuming the average spell cast has a cost of 200 (it won't, it will be lower than that), that's still 15 powerful uncommons or average rare spells, every turn.

[Seravy]

My issue is that AI, in order to summon lots of things, have to be able to cast a ton of spells; but again, just like with production, they are actually good at all other spells now (except maybe unit enchantments).

And this is exactly why the AI is picking their spells based on personality and chance instead of focusing on the one that wins them the game.

But sure, let's go investigate all types.

Creatures - I already explained this one.
Global Enchantments - You can only cast them once. Irrelevant. Only affects Disjunction, but other AI wizards will cast proportionally more of those to balance it out, and we don't want the human player to reliably prevent these unless they are already winning the game or are playing Sorcery. Dispelling key spells before entering a war will still be relevant.
City Enchantments - Low cost and only once per city. Skill is irrelevant. Dispelling has an advantage as it hits multiple spells at a time and has doubled effectiveness if used by Sorcery wizards, so if anything, more skill just balances it out.
Unit enchantments - The AI is bad with these and they are easy to punish by Dispel Magic if overused.
Item creation - The AI is bad at this too, if anything, it helps the human player who will loot the items after killing the hero that holds them.
Disjunction - How many of these are used is regulated by the amount of dispel priority assigned to potential targets. The AI is holding back a lot in this area because it wouldn't be fun if they used max power disjunction.
Disenchant Area - Only affects Death wizards who should be smarter than to try and drain the power production of a high skill wizard with 35 cities. This tactic only works on smaller empires anyway.
Summoning heroes or champions - Limited to 6 slots most of which already filled by free offers. Skill use is not relevant.
Change Terrain - very low cost and can only be used a limited amount of times, no different from other city buffs. Transmute is the same.
Curses - Almost all of these has a low (100-200) cost and most of them only work once per city - those that do work twice will have much less effect on the second casting (not enough stuff to still destroy). Casting skill is not very relevant on these due to that, and the amount is directly regulated through personality and other choice modifiers. If we don't like the amount, we can change those.

I don't see a problem in any category and I feel I have had this discussion before.

Also, no wizard will have 1500 skill even in the late game, and if you still have more than one strong enemy alive and actively at war with you in 1425 on Extreme, it means something went wrong. You're supposed to either kill other wizards one at a time, or form treaties to make sure they don't all attack you at the same time. If neither is an option and they are all allied, you should consider to play a "wait out" strategy, raising your own power and wait until their alliances break (due to Spell of Mastery or other nasty globals), then strike them out one at a time.

[Nelphine]

I think you underestimate what being able to cast half a dozen city curses every turn can do, or being to cast half a dozen city enchantments. A player cannot ever effectively blanket an entire enemy empire in curses; and AI can do so in 1-4 turns (it probably actually takes longer, maybe 10 turns, to hit every single enemy city, due to random chance). Similarly, for a human player, blanketing an entire empire in Uranus' blessing or something similar is something they have to actively decide to do - here, we have a sorcery wizard casting Uranus' blessing on OUTPOSTS, and it isn't even a particularly bad move.

The balancing around spell costs that has been done so precisely when it comes to humans using a spell is thrown out the window, because AI MUST be able to summon dozens of high rarity creatures in order to use them offensively. I understand and accept why summons need to come that often. Having earthquake, or pestilence (if you don't have consecration), or warp node, or any other curse spell being cast half a dozen times per turn per AI, will very effectively cripple a human player. And they can't possibly keep up with that. Similarly, something like change terrain? This AI has cast it over a hundred times, and its not yet 1420. For this AI, within 1 year, he literally took a max pop 5 desert outpost, and converted it into a max pop 22 high elf capital - and it doesn't have a farmers market yet. That's not something a human player can do, unless the game has already been won. Raise volcano? Same thing. Casting it en masse is extremely powerful, and the cost is balanced around this - but the AI breaks that balance.

The AI can use city curses, city enchantments, warp nodes, raise volcanoes, or any other similar spell at least as effectively as the human player (in terms of intelligence) and gets better results than the human player (due to bonuses). Yet they can also cast them dozens of times more often than the human player.


I'm not saying stop them from using these spells. I'm saying, if they're already as good at using them, and benefit MORE from them, than the human player, why should they also be allowed to cast them far more often?

It sounds like this is what you want. But, that pushes the human even more into 'the only way to war with an AI late, is to constantly do fortress spikes, that the AI can never replicate, because the AI can't play well with units on overland.. but in every other way overland, the AI is numerically so superior to the human, that to NOT do fortress strikes is to die.'

Note: The biggest problem with this is not sorcery, or death, or any realm. Its specifically a multi-realm problem, where one realm is sorcery. Sorcery gets Aether Binding very quickly; and if it gets Uranus' blessing, it can then go crazy with the nature/chaos/death spells that are actually super dangerous. Humans can't replicate this - they have to give up too much to play multi-realms as effectively.


On topic of Disjunctions - by 1415, this AI can cast a max strength disjunction in less than 1 turn. You can't spell blast it. Since he has Aether Binding, even if he doesn't use max strength, he still probably kills your global in one attempt. Even if it takes him 2, there's still nothing you can do about it. You noted AI don't use max strength disjunctions. The reason is because it IS overpowered. But ALL spells that have targets in roughly the same number as the human player would have targets, are also overpowered for the same reason. The AI can spam them. Just like max strength disjunctions, all spells that target globals, or cities, or nodes, or anything else where the number of targets are limited, are going to be overpowered due to the nature of the AI cheating bonuses.


I would like to suggest that, just like with production, only military spells (so, summons and unit enchantments) get the discount. The AI IS as good as the player with non military spells, and city buildings, and even settling. It doesn't need to have the bonuses in those areas. (Sure it can have some; it doesn't need to be twice as good, or more, as the human player.)

[Seravy]

Actually there is one thing I forgot about city curses.
The AI cannot cast them AT ALL unless they are being hostile with you.
If you play the game as intended - eliminate those wizards who are hostile and don't anger those who aren't - no one will cast any city curses on you except during war. A prolonged war of attrition definitely should benefit a Chaos or Death wizard, I see no problem there.
Oh and one more thing. Most curses can be prevented. There is Consecration and Spell Ward if the prolonged war is unavoidable and the enemy is Chaos or Death. If they are Nature, you can do Flying Fortress, it prevents Earthquakes. Drain Power isn't strong enough to win even if used once every turn. Corruptions can be cleaned. And volcanoes, well, they are a pain but you can revert them to mountains and gain a lot of extra production, plus it takes 16 of them to destroy all terrain belonging to a  single city.

Yes, there are cases when you can't prevent the spells, and can't avoid the war. For those you can either win it fast enough, or lose the game. That's what curses are for, afterall.

However, how often curses are cast depends on "curse priority" relative to other priorities.
It simply cannot be so high that the AI spends more than a quarter of their skill on curses, as they cost a low amount of MP. That would require, like, 10 times the normal priority as it has to compete with 4-5 other categories out of which some have massively more expensive spells so whenever they are picked, they eat that many times more skill.

'the only way to war with an AI late, is to constantly do fortress spikes

No, the only way to fight the AI late is to have spells that can allow you to do so. Life doesn't work against Earthquake. If it was any other curse, you could use Consecration.
Oh wait, I know! To fight Earthquake the best tactic is to not raze cities so you always get back more buildings than what was wrecked. Yes, you need to be able to defend them to do that. Buy units in them for gold - you should have plenty as your other cities have high population and no buildings to maintain. Guardian would help you immensely in this scenario.

...I can't help to think the problem is you simply weren't in the position to win this game to begin with. Let me see...

"I end up with 12 cities. Jafaar has 11, Merlin 21, Raven 35."
Yes, this is the problem. You control 15% of the world and fight against the other 85%. That's not a war one can win on this level of difficulty, curses or not. You should have more cities by this time.

Haven't actually played Extreme yet on the current version but if you don't control at least 30-40% of the world in 1415, I don't see how you're expecting to win the game through military against a 3 way alliance.

I would like to suggest that, just like with production, only military spells (so, summons and unit enchantments) get the discount.

It doesn't work that way. Curses are cheap and represent a small fraction of skill spent. Paying more would make no difference.
How do I explain this.
Let's say the AI has X skill and divides it between 4 types of spells. Each of these have priorities but for simplicity, let's also assume all these priorities are equal.

So X=A+B+C+D where each letter is the typical cost of a spell in the category.
Unit buffs are cheap, let's say 50. City spells are 150. Summons are 500. Curses are also cheap, 120.

So X=50+150+500+120=820. The AI needs 820 skill to use exactly one curse.

You suggest that D should be twice as much (not receive the halving discount).

In that case, X=50+150+500+240=940. The AI needs 940 skill to use exactly one curse in this way.

How much of an improvement is that? The AI is casting 13% fewer curses. Is that good? Well, it's something but does it matter? Not much.
What are the side effects? You guessed right, the AI casts 13% fewer creatures, unit buffs and city spells as well. Do we want that, definitely not.

Now, priorities in reality aren't all 1 but it's fairly close to it. And curse usually has less than the other 4. Even with a Maniacal wizard that has access to 3-4 unique curse types to boost it further, it'll be like, at most 50-100% higher than others. That's not going to make up for the cost difference of a 500-1000 mana summon and a 100-200 mana city curse. And the fact that there are actually 8 or so categories, not just 4.

(and btw for this same reason the production suggestion is meaningless too. Aside from corner cases, cities will build units and buildings both, so having a different production rate achieves nothing - the buildings will still slow down unit production because the unit production has to wait for the building to complete before it can start. And since RNG is involved, if just escalated the problem of bad RNG - if many buildings are picked in a row, it'll hurt the AI more than if production is equal on everything.)

[Nelphine]

So, my opponent has 1000 effective casting skill. That means he does an earthquake every turn. Something isn't working with your formula then, otherwise he wouldn't be casting an average of 3 per turn, and 1 turn hitting 6.

So, if that's the case, then any idea why that would be happening? (He's ruthless, not maniacal as well.)

And again, life is good at winning military campaigns. That's why I STILL expect to win. I've already gone from controlling 12/82 cities in the world (~15%), to controlling 11/24 cities (~45%).

It's not a walk in the park, and things like earthquake is annoying, but I'm not worried about this from a 'this is too powerful' point of view. It's just a crazy amount of stuff.

[Seravy]

So, my opponent has 1000 effective casting skill.  That means he does an earthquake every turn.  Something isn't working with your formula then, otherwise he wouldn't be casting an average of 3 per turn, and 1 turn hitting 6.

So, if that's the case, then any idea why that would be happening? (He's ruthless, not maniacal as well.)

Let's see...(will only list the important modifiers)

Summoning :
120 base
+10 per empty hero slot
+30 if Magic Spirit needed
+50% if conjurer
+30 or +100 if behind in military somewhat or significantly compared to human (probably neither applies)
+80 or +150 if high or top tier summon is available (Stone Giant is neither )
+50 if Militarist or Expansionist

Curse
+125 for Ruthless
Base (100-Relation-Hidden Rel*2+50*(war)*(1+Difficulty))/4, assuming "Hate" relation and war, this is 100.
+60 for each Divine order in play
+50 if Expansionist
+25 if "Blizzard type" is known
+40 if "building removal type" is known = Earthquake
+50 if Dispelling Wave or Time Stop is known (yes those are curses for the AI category)
I assume the other types will not be relevant, the wizard has no Death or Chaos books to have them.

Maybe part of the problem is, Stone Giants with Conjurer only cost 360 and Basilisks etc have enough priority together to get selected half the time, resulting in casting even cheaper (below 200) cost summons. So actual spending on average per summon is ~250 instead of the typical 500.

Also we tweaked summon priority when you asked for more unit buffs. It was most likely more than 120 base originally, I think something like 200 unless the circle was at the capital. So it was almost halved...two weeks ago.

So in short, he does that because he didn't get Gorgons and spends only half the expected on summoning spells. If he gets a very rare summon, it'll reduce the amount of curses to like a quarter of the current.

...anyway, I really need to go sleep now.