Master of Magic unofficial patch (latest version)

[WhiteMage]

(Then again, I keep getting messages from people who are dissatisfied with current Insecticide difficulty. I have squeezed the fun out of their beloved game, they say, it's too hard now. Anti-motivating.)

Don’t listen to them. It is not too hard yet. But when it will be, then you shall lower the ridiculous 400% bonus for the impossible AI. That should be ultra motivating for you.

Those who can’t handle the difficulty should play on lower level or older insecticide versions.

[Anthony]

(Then again, I keep getting messages from people who are dissatisfied with current Insecticide difficulty. I have squeezed the fun out of their beloved game, they say, it's too hard now. Anti-motivating.)

How dare impossible be impossible Also, how dare the AI actually attack you, as opposed to declaring war and then not doing anything.

[simone]

Thanks!
Don't listen to them... The game is not too hard. And "impossible" should really be "impossible", by the way

Simone

Any (either bug patches or AI).

I was just curious to know if there's anything coming soon.
Thanks!

(Then again, I keep getting messages from people who are dissatisfied with current Insecticide difficulty. I have squeezed the fun out of their beloved game, they say, it's too hard now. Anti-motivating.)

Don’t listen to them. It is not too hard yet. But when it will be, then you shall lower the ridiculous 400% bonus for the impossible AI. That should be ultra motivating for you.

Those who can’t handle the difficulty should play on lower level or older insecticide versions.

[kyrub]

And "impossible" should really be "impossible", by the way

Come on! Every game now has a level called "impossible", that every I-know-the-tricks player can beat.
(That said, impossible MoM is still beatable. It won't be if those changes enter the frame. 400% is outrageous bonus.)


I gave the spells a quick look and there is massive room for improvement.
The basic architecture is not totally wrong, the AI takes acount of its general situation in battle - and tries to protect or overturn its position. Or damage the opposition when the chips are downs. There are even small tricks (hit Possession and run away, nasty). There are a few big failures, AI makes no effort to estimate the damage its spell will do, for instance, although the fonction is at hand. Some spells have NO inner incentive (Lionheart, of all of them; Animate Dead, another one - I think that's proof Microprose knew Animate Dead was bugged, and it would stiil be bugged if not for Asfex). Some are mismatched (would you believe that Warp Wood can overturn lost position?). The protection spells are valued too highly - I already corrected it in previous patches, but there's more to it, it should be situation specific. There could be a bracket for "universal" spells, good for all but totally lost positions - Heroism, Confusion, Prayer, I appreciate other ideas here.

There's another big fish (a bug-fish), the AI takes the spell cost as a factor for casting. A positive factor! This is probably an attempt to avoid AI repeating fireballs when it can doom bolt a unit. (I still think that the mechanics, while partly effective, is doubtful, as it clouds the proper judgment. On the other hand, modding mana cost of a spell to real values should improve AI spellcasting even in 1.31!) The bug: AI takes the spell discount into consideration. So, let me clear that, when it has a nice collection of blue books and a red one, it will avoid casting Phantom Warriors, because they are too cheap, but will summon a Fire elemental instead. When being conjurer, it won't prefer combat summons at all, probably. Huh.

There are inherent problem in targetting: unit is considered to be a target for a Life Drain spell when it has <13 resistance. Maybe that's why Kali keeps them coming without any success. This is easy to correct, it should be something like <7 isntead. Overall, the resistance of target could be more in scope, targetting a unit with 9 resist role should be a last ressort, if no other targets in sight. 1/10 chance is too low. Some other problems are harder to correct: why is Magic immunity targetted at a multifigure unit? Flight, invisibility as well? Haste, Berserk no proper clever targetting as well. These are, however, not easy to target without running into complicated equations and we have a little coding space here (so I'll probably have to make some sort of a dirty compromise).

[Anthony]

There's another big fish (a bug-fish), the AI takes the spell cost as a factor for casting. A positive factor! This is probably an attempt to avoid AI repeating fireballs when it can doom bolt a unit.

Or to just make the AI favor big spells over small. The thing is, fireball is not necessarily the wrong decision; against a 6-counter unit with defense of 7 or lower or a 4-counter unit with a defense of 4 or lower a 39 point fireball will average more damage than a 40 point doombolt. On the other hand, fireballing a single-counter unit (which I've seen the AI do) is always dumb.

Ideally some part of the AI logic could be baked into spells.lbx, to simplify modding. Things like the bug with replacing wind mastery wouldn't happen if the AI was filtering based on the spell type rather than an internal list.

[Asfex]

life drain: if the caster is unit, it does not count positive effect of healing by this.
Ai underestimate the positive effect of lose resistance. it will not try to use warp creature then life drain, or confusion. it will not protect the unit by bless against a death mage. Ai do not count the enemy spell prefer. It will not protect death units from dispel evil, or by resist elements against petrify army. It will not count the enemy books at all! It will not save mana for the cheap counterspell for late combat.
It uses spells with all buffs if possible. so, cast "call lighting" to empty enemy mana pull via disenchant is an effective trick.

But what is a goal? the combat ai behaviour may be improoved endless. I prefer the separate script for this holy idea. for true -impossible players only.

[coldsteel]

If you ever think of a way for the AI to always spell-target, and shoot at, the human-player wraith(s) *first* among a big stack of undeads until the wraith(s) is gone -- put that in! It would greatly protect the AI from the 11-death-book spoiler strategy, which is still the best way for humans to win on Impossible level

[Anthony]

If you ever think of a way for the AI to always spell-target, and shoot at, the human-player wraith(s) *first* among a big stack of undeads until the wraith(s) is gone -- put that in! It would greatly protect the AI from the 11-death-book spoiler strategy, which is still the best way for humans to win on Impossible level

The problem with wraiths isn't really AI targeting; it doesn't help, but a unit of wraiths is pretty much unbeatable early game even with perfect play (in my mod, I'm considering reducing counters on wraiths and death knights from 4 to 2).

[coldsteel]

If you ever think of a way for the AI to always spell-target, and shoot at, the human-player wraith(s) *first* among a big stack of undeads until the wraith(s) is gone -- put that in! It would greatly protect the AI from the 11-death-book spoiler strategy, which is still the best way for humans to win on Impossible level

The problem with wraiths isn't really AI targeting; it doesn't help, but a unit of wraiths is pretty much unbeatable early game even with perfect play (in my mod, I'm considering reducing counters on wraiths and death knights from 4 to 2).

Yes, wraiths are the toughest - but the computer could still beat them sometimes if it aimed its magic-shooters, alchemy-equipped-shooters, and magic spells against the wraith instead of those lowly undead hell hounds and undead spearmen

[kyrub]

The problem with wraiths isn't really AI targeting; it doesn't help, but a unit of wraiths is pretty much unbeatable early game even with perfect play (in my mod, I'm considering reducing counters on wraiths and death knights from 4 to 2).

Yes, wraiths are the toughest - but the computer could still beat them sometimes if it aimed its magic-shooters, alchemy-equipped-shooters, and magic spells against the wraith instead of those lowly undead hell hounds and undead spearmen

For me that's the way to avoid, modelling the AI on an extreme, unbalanced feature of the game. This way, we finish with an AI able to stop Wraiths or an Invulnerable Guardian spirit or Pathfinding Basilisk or ..., but in 95% of other cases the AI completely fails to do its duty and it'll be worse than vanilla.

11-book strategies are extreme cases, they create massive (albeit enjoyable) havoc in game balance. The only good way to deal with them is to mod their mechanics or - most simply - not play with them. Nota bene: if you dislike AI playing with them, like Jafar and his Storm giants, mod the original 11-books wizards in Magic.exe, and the probability of appearance of an 11-book AI wizard will steeply drop.

There are valid ways to partly counter 11-book strategies, and I have used some in 1.40 already (invulnerable GS is harder than in vanilla; also, prayer is not bugged in favour of the player anymore!). For 11-book wraiths, there's only 1 idea I have that could be implemented (and that would not harm AI seriously), and that is DISPEL EVIL as a tech priority for white book's player, when there's a Death Human player around. Dispel evil, uncommon spell, will take a whole wraith unit down in one turn. - But that's colour limited.