SORCERY Realm

[Nelphine]

Actually, we only want 1-4 (maybe 5?) Spells. So do we have space for more complicated equation?

Something like: 2 spells if 150+25*land size, 3 spells if 250+50*land size, 4 spells if 450+100*land size, 5 spells if 600+100*land size?

(2 and 3 spells are reasonably simple to get - 2 you would get if you had minimal cities and there were 3 wizards, 3 you would get if there are only 2 wizards. But reaching 4 would take a ton of effort and 5 is practically impossible.)

[Nelphine]

On the spell selection: zitro, crusade is only amazing for the ai if they are perfectionist and have altar of battle and have warlord.  It's decent if they have lots of war colleges and do not have warlord and do not have to lots of altar of battle. In most other cases it's (at best) +1 to attack or defend or both. 

Never steal: 
Just cause
Detect Magic

Steal only if already researching SoM and no other options have presented themselves:
Divine Order
Aura of Majesty

Always Steal but only if casting wizard has at least 6 chaos books (and still low priority):
Chaos Surge

Always Steal but only if casting wizard has finished researching all rares (medium priority - the skill bonus is super important for Spell Binding since number if spells is based on casting skill):
Aether Binding

Always Steal (do we want to include an if not in own research list condition? I'd suggest setting Spell Binding to lowest priority for research, and not including such a condition. That way the AI already finished it's own research so doesn't have to worry about stealing something it already was going to get.  While the human can replicate it, I don't think we should base it around that as most humans can't. Finding it in treasure I think is reasonable enough benefit that if the AI misses out on its own research that's ok) (not in order yet):
Fairy Ring
Armageddon
Charm of Life
Time Stop
Meteor Storm
Evil Omens
Doomsday
Suppress Magic 
Seismic Mastery
Survival Instinct (offensive unit bonuses are rare for sorcery)
Life Force (even more power for skill!)

Steal only if already researching SoM (higher priority)
Everything else.

Note: since I'm suggesting making this spell very low priority, other so should already have good spells researched by the time ai research Spell Binding so they have the biggest list to choose from. Thus why I'm using researching SoM as the time to steal the lower priority spells.

[Seravy]

Spellweaver would affect it (unless we specifically adjust it to remove the effect which is doable but not trivial), heroes and AI would not. Basically, it would be the base casting skill you see on the Magic screen.

. So do we have space for more complicated equation?

This will go into the disjunction priority procedure most likely (even though it's not a disjunction spell anymore) which has a huge amount of space. Think 2000+ bytes. Plus half of the existing Disjunction priority procedure would free up, too.

Oh you mean the cost, not the AI. No we probably won't have space there but I haven't checked. It'll have to go into the part where the player clicks a spell and it shows a red box with "you don't have enough casting skill to steal another spell right now".


I prefer to keep it high research priority. The critical part is how the AI will judge whether it can reasonably expect other wizards to research a particular spell or not.

For example,  it has a list :

Armageddon (unresearched, 120 priority)
Charm of Life (unresearched, 100 priority)
Survival Instinct (available, 45 priority)
AEther Binding (available, 20 priority)

and has 1 use remaining, how does it decide whether to take Survival Instinct now, or wait until the others get researched by the opponents? (Assuming we defined the AI cares about waiting if the priority difference is 50 or greater)
Try to guesstimate from the power and research income of the other wizard and the remaining spells they haven't researched? Maybe wait if priority difference> expected number of turns it'll take to research that thing by the other wizard? How do we calculated expected turns? Assume the spell will be picked halfway by the other player, or assume always last?

Alternately, say the AI never waits, unless the spell is below a certain threshold in which case it always waits unless everyone is already researching SoM or has nothing in their unresearched list to pick?
Do note the AI can and eventually will take a better spell even if not waiting, through the sheer effect of gaining casting skill and allowing it to have an extra - so waiting might not be a good tactic, unless the "cost" in skill is too high.
At one spell every 200 skill, the AI can reasonably expect to take another spell, when it hits 600 (it probably already has 400 when getting Spell Binding) but only that one, 800 is far out of reach in most cases.
We might actually want to have the spacing based on this for the formula, probably should have two cutoff points in the 450-650 skill range where the AI is most likely to go through during the later stages of the game after Spell Binding is researched.

[Nelphine]

Not sure if you saw my post about my problems with 200 or 250 casting skill being used as the basis - it was right before zitros list of what should be always steal etc.

[Seravy]

I peeked at the casting skill of the enemy in my game and it was slightly over 400 when they got Spell Binding. So 400 being the cutoff point doesn't seem to work very well for that purpose.

I guess we can leave it at that but make the AI "hold" their pick on "not the best" spells until it reaches 500?

So it would be like this - note the AI would never steal a spell they can research in the future :


Never steal
Just cause
Detect Magic
Divine Order
Aura of Majesty
Chaos Surge (unless owning 5+ Chaos books in which case medium priority)

Always steal
Fairy Ring
Armageddon
Charm of Life
Time Stop
Meteor Storm
Evil Omens
Doomsday
Enlightenment

Steal if 2 or more unused picks, or skill=[480,600), otherwise wait for better if one exists
Zombie Mastery
Suppress Magic
Survival Instinct
Crusade
Life Force
Seismic Mastery (always if knowing a spell to trigger the EQ)

Steal if 3 or more unused picks, or skill=[540,600) and 2 or more unused picks, otherwise wait for better if one exists
Doom Mastery
Eternal Night
Herb Mastery
Holy Arms
AEther Binding


This would make the AI who likely got Spell Binding around 400 skill to wait somewhat if there are only medicore options, and steal whatever is best at the moment it hits 500, then 600, then 800 skill.
So the first two picks would be used immediately.
We can translate each category to a priority value and set priorities accordingly.

[Nelphine]

Are you rounding skill up or down? As per my post detailing my concerns I'm not sure of that .

[Seravy]

Rounding up.

0-199 = 1 spell
200-399 = 2 spells
400-599 = 3 spells
600-799 = 4 spells
etc


Here is a priority table, which roughly preserves original spell binding priorities but is modified where necessary :

Eternal Night 48 +30 if at least 4 Death books
Evil Omens 140
Zombie Mastery 37
Aura of Majesty 0
Aether Binding 65
Suppress Magic 120
Time Stop 250
Seismic Mastery 100
Fairy Ring 160
Herb Mastery 30 +50 if any other player has Meteor Storm
Chaos Surge 0 +40 if Doom Mastery known
Doom Mastery 70
Doomsday 170
Meteor Storm 240
Armageddon 230
Enlightenment 145
Lifeforce 150
Crusade 180
Just Cause 0
Holy Arms 60
Divine Order 0 -2* DO Need
Charm of Life 200
Detect Magic 0
Survival Instinct 68

Priority of 18 or below would be "never steal".
19-99 would be where the AI uses "Steal if 3 or more unused picks, or skill=[540,600) and 2 or more unused picks, otherwise wait for better if one exists"
100-139 would be "Steal if 2 or more unused picks, or skill=[480,600), otherwise wait for better if one exists"
140+ would be "always steal".

Meanwhile I received a counterargument IRL, basically
"This is worse, with that the AI can steal the spell even if you don't cast it so you can't prevent it from being stolen at all". Unfortunately he has a point.

[zitro1987]

I like your newest priority table,

*I'd probably increase charm of life to the mid 200s as it helps any situation, period.
*Fairy Ring could also be in low-mid 200s, giving them so many top-tier summons.
*I'd probably reduce time stop a bit to upper 100s, given your experiences that AI is not THAT scary with it.
*I'm not sure if meteor storm needs to be higher than doomsday, the latter is incredibly damaging to economy.


0-199 = 1 spell
200-399 = 2 spells
400-599 = 3 spells
600-799 = 4 spells    (I'd suggest 600-infinity, capping it at 4 spells)

"This is worse, with that the AI can steal the spell even if you don't cast it so you can't prevent it from being stolen at all". Unfortunately he has a point.

I thought we were no longer into the thought process of avoiding having the spell stolen. Are we to avoid casting very rare globals simply because they can be stolen? If so, we would therefore avoid researching these very rare globals! (which would be a strategy in this case)

[Nelphine]

I'm assuming that it would be based on base casting skill so that neither hero skill nor spellweaver nor AI overland cost reduction would affect it (making archmage a little more attractive to sorcery wizards). I'm also assuming when you say every 200 or 250, you mean round up (so 251 is 2 spells).

Part of why I'm suggesting 300 is because it's reasonable to assume that a sorcery wizard will have Uranus blessing. That means on fair, assuming 3 wizards survive to very rares, and each controls ~2/3 of one plane, we can expect the casting wizard to have ~25 cities. That's 275 casting skill just from amplifying towers. We should also assume aether binding. That means the human can reasonably expect to have 400 casting skill, and an AI can reasonably expect to have 400-600 depending on difficulty. 

Even at 250, an AI would often start with 3 spells, which seems too high. At 200, a human would start with or be almost at 3 (only needing 401 casting skill for 3).

With only 2 wizards left, that number goes up so that at 250 but high difficulty an AI could start with 4. Moving to huge, a human, at 250, would be pretty close to starting with 4 (550 just from amp towers +150 from actual casting skill). On huge, an AI who controlled one whole plane, could reach 5, which is absurdly powerful.

That seems too many spells to me. The only time anyone ever has 1 is if their empire is small given they've  reached very rares AND they don't have Uranus blessing. I don't think you should balance the spell around that.

Just posting this again. Do you really want most people starting with 3 spells and a significant number of AI starting with 4 spells?

[Nelphine]

On topic of should we do it at all: actually this is true. My suggested 750 casting cost is probably too low. Should be more like 1200 (roughly doubles the cost of any spelling you use via this.)