Heroes of Might and Magic III (RoE, AB, SoD)

2010-07-25 01:48:28 pm

i searched for H3 but the list was too long ... and it slipped by me, thanks for pointing out.

2010-08-25 07:35:39 am

What I want:

Speedrun through the campaigns of RoE, AB, SoD
what would be legit to do in terms of what rules can be bent, what rules can be broken, and what rules can't be bent or broken? Smiley

I think it's legit to "tweak" the game - fastest movement, no enemy animation, etc. Also, I think at some point some player - AI will be required, is doublecasting allowed?
stuff like that.

I'll try to come up with a list of what I think that needs to be done, and feel free to argue/improve/suggest

thanks

2010-08-25 08:04:37 am

Hi

You can tweak the game as it's on the game options. You can also doublecast if you want.
I don't play HMM 3 for a while so I don't remember well the campain. I remember that a good castle development can make an angel on the first week. With this one, you can rush pretty all you need.
This game is pretty all about luck manipulation. Luckily, if you do twice the same moves, you have approximatively the same result, so you can save and reload to get the best result.
Do you want to make a time speedrun or a number_of_turn speedrun ?

Edit history:

raymond007: 2010-08-25 08:24:07 am

2010-08-25 08:22:42 am

I would like to finish the 3 episodes as quickly as I think I can. I am estimating it took me an average of 40 - 50 (realtime) hrs per episode, but that was without any speed in mind. Overall I finished the episodes about 5 times RoE, Sod, twice AB. But never was speed an issue, this is why i'm "noob" about this whole thing, and I think I could use pointing out certain directions
For instance: Logistics skill is a must - I think. Pathfinding - depends on map; however, mages with massive MP will dimension door up to 10 times, I think, with expert Air + huge Spell Power. But that is by taking a lot of time which will not be available in a speedrun. I mostly play magic-oriented heroes, and used to build them like that, but as I said, I am (as we speak, lol) playing SoD and really tried to play Yog to his full potential. But not even one spell is mind-melting for me, so there must be a spell strategy in mind as well. for quick travel purposes, Wisdom + Air and Earth for town portal/fly/dimension door; not to mention how incredibly good mass slow + mass haste is. On the blurry side, it takes time to build up. However I tend to think by optimizing paths, avoiding backtracking, picking up only the necessary stats ( att/def/spow/know and not chasing every single garden/magic tower that time of 50 hrs ( continuous gameplay, not including reloads (( countless reloads of fights)) and such - wasted time) can be cut down pretty drastically. I have no idea how much time would be considered a good run - right now 10 hrs ( per episode ) seems reasonable. I can make a pretty good picture once i actually start speedrunning.

I would like to finish the 3 episodes as quickly as I think I can. I am estimating it took me an average of 40 - 50 (realtime) hrs per episode, but that was without any speed in mind. Overall I finished the episodes about 5 times RoE, Sod, twice AB. But never was speed an issue, this is why i'm "noob" about this whole thing, and I think I could use pointing out certain directions
For instance: Logistics skill is a must - I think. Pathfinding - depends on map; however, mages with massive MP will dimension door up to 10 times, I think, with expert Air + huge Spell Power. But that is by taking [b]a lot of time[/b] which will not be available in a speedrun. I mostly play magic-oriented heroes, and used to build them like that, but as I said, I am (as we speak, lol) playing SoD and really tried to play Yog to his full potential. But not even one spell is mind-melting for me, so there must be a spell strategy in mind as well. for quick travel purposes, Wisdom + Air and Earth for town portal/fly/dimension door; not to mention how incredibly good mass slow + mass haste is. On the blurry side, it takes time to build up. However I tend to think by optimizing paths, avoiding backtracking, picking up only the necessary stats ( att/def/spow/know and not chasing every single garden/magic tower that time of 50 hrs ( continuous gameplay, not including reloads (( countless reloads of fights))  and such - wasted time) can be cut down pretty drastically. I have no idea how much time would be considered a good run - right now 10 hrs ( per episode ) seems reasonable. I can make a pretty good picture once i actually start speedrunning.

2010-08-25 08:50:07 am

You should train yourself on the first episode and show a WIP. If you drasticaly minimise everything, I think that you can run in far less than 10 hours.

Obvious things that can save a lot of time :
-make a second (lvl 1) hero for all the ressources and army. Maybe 3 heroes are even better on some points
-Calculate everything you need to build your castle before starting anything to select your route.
-Take the item that allow you to flight through the map as early as you can
-Don't free the way to items or buildings you don't need
-Use massively the autofight (it's pretty easy if you have only angels on you hero for example)

I don't know how many saves you can consider. I think that you can play through a level, save as often as you want, then restart it and make exactly the same game without saving. As the RNG is pretty friendly for (real time) speedrunning that must be the best way to make a "clean" speedrun.

zoonel

2010-08-25 12:54:47 pm

Go play spacechem !

hi,
some quick advices :

-remember you are aiming at the clock so you can waste turns if they are done quick. One exemple of this is a speedrun of one level from homm V, i dont remember which one but the guy was idling in his base for several weeks to build his army then crushed the opponent.
-i dont know how auto battle works, but maybe you need to tweek your hero for it so it doesnt loose too many units.
-with the map editor you can plan your routes well, you can also see the always join enemies .... but i think you already know all the map by heart, you did it 5 times Shocked

-be prepared to spend a lot more hours to make this speedrun then a normal play, planning trying restart redo change tactic .. all this take a lot of time, so dont be in a hurry and do each day a little, patience is the key Wink

Good luck

Edit history:

xsite: 2010-08-25 01:24:49 pm

xsite: 2010-08-25 01:14:46 pm

xsite: 2010-08-25 01:08:01 pm

2010-08-25 01:06:25 pm

The thing with homm3, is that while it's one of my favourite games, if you complete the game in regular SDA fashion a la optimizing for (real) time, it won't be a very interesting run to watch IMO as it wouldn't be a good representation of regular gameplay. The strategy and tactics aspect will be greatly reduced through quick combat, and spamming end turn. If you turn off "view ai movement" a week can go by in several seconds, so just massing units and rushing to an objective can be a legit strategy. Also things like stables, boots of travel, logistics, pathfinding etc will be greatly downplayed, again since you can just mash end turn without a much of a penalty. Not to mention it would be difficult to follow with player movement turned off. It really shouldn't take longer than a few minutes for each campaign.

Just off the top of my head, for the long live the queen campaign, the first level will be about just end turning your way to the enemy's town, since I believe there is very little or no garrison at the beginning. Second level will be about getting angels and perhaps getting the angel wings item and then proceeding to end turning and steam rolling the AI's castles and heroes. The third one will come down to end turning to the gate keeper huts and until you have sufficient forces to beat the forces at the gates or manipulating the mage guilds to get dimension door and just teleporting over the gates to the griffons.

I think it would be far more interesting to see a homm3 speed run which is optimized for days/weeks/months first and real time second.

The thing with homm3, is that while it's one of my favourite games, if you complete the game in regular SDA fashion a la optimizing for (real) time, it won't be a very interesting run to watch IMO as it wouldn't be a good representation of regular gameplay. The strategy and tactics aspect will be greatly reduced through quick combat, and spamming end turn. If you turn off "view ai movement" a week can go by in several seconds, so just massing units and rushing to an objective can be a legit strategy. Also things like stables, boots of travel, logistics, pathfinding etc will be greatly downplayed, again since you can just mash end turn without a much of a penalty. Not to mention it would be difficult to follow with player movement turned off. It really shouldn't take longer than a few minutes for each campaign.

Just off the top of my head, for the long live the queen campaign, the first level will be about just end turning your way to the enemy's town, since I believe there is very little or no garrison at the beginning. Second level will be about getting angels and perhaps getting the angel wings item and then proceeding to end turning and steam rolling the AI's castles and heroes. The third one will come down to end turning to the gate keeper huts and until you have sufficient forces to beat the forces at the gates or manipulating the mage guilds to get dimension door and just teleporting over the gates to the griffons.

I think it would be far more interesting to see a homm3 speed run which is optimized for days/weeks/months first and real time second.

2010-08-25 01:57:15 pm

xsite you make a good point, and I really think about trying at least RoE with full gameplay, as it's difficulty is low and i think with enough luck manipulation I could get it done in decent time. I don't know how much "decent time" will take but I think I will play a normal campaign and try to time myself to see how much time it actually takes, and see from there, I mean, as pointed out, the battles are the most exciting part. As I played before, spending time to build heroes (and I always used one hero to build, even if "top 5 heroes will go to the next level" so in the later parts the hero would kick ass so much that it made things really easy, EXCEPT for the one act when you have to capture enemy castle within 3 months - 90% of time was a pain. But there is a very good resource http://www.heroesofmightandmagic.com/ that I will spend a lot of time in the future to come, and review all H3 info there is. The planing part will take the most, I guess.
Someone pointed out that a good rush strategy is to get Diplomacy and just pick up neutrals as you go, but I don't see it a viable plan. I think logistics is far better here. If I decide to go battles, then wisdom + earth + air is quite essential in my view. The fourth skill for me is usually logistics, but I guess I won't be marching around that much. I think skill strategy will be quite important as town portal makes massing troop so easy as well as defending with the primary hero for the XP/artifacts.
Definitively, this will not be an easy task Smiley

xsite you make a good point, and I really think about trying at least RoE with full gameplay, as it's difficulty is low and i think with enough luck manipulation I could get it done in decent time. I don't know how much "decent time" will take but I think I will play a normal campaign and try to time myself to see how much time it actually takes, and see from there, I mean, as pointed out, the battles are the most exciting part. As I played before, spending time to build heroes (and I always used one hero to build, even if "top 5 heroes will go to the next level" so in the later parts the hero would kick ass so much that it made things really easy, EXCEPT for the one act when you have to capture enemy castle within 3 months - 90% of time was a pain. But there is a very good resource http://www.heroesofmightandmagic.com/ that I will spend  a lot of time in the future to come, and review all H3 info there is. The planing part will take the most, I guess. 
Someone pointed out that a good rush strategy is to get Diplomacy and just pick up neutrals as you go, but I don't see it a viable plan. I think logistics is far better here. If I decide to go battles, then wisdom + earth + air is quite essential in my view. The fourth skill for me is usually logistics, but I guess I won't be marching around that much. I think skill strategy will be quite important as town portal makes massing troop so easy as well as defending with the primary hero for the XP/artifacts. 
Definitively, this will not be an easy task :)

2010-08-25 02:25:16 pm

Well again, what are you optimizing for? Because the *ideal* real time would lack any real strategy and will probably just mass units. If you're not trying to cheese the game, then what are you doing? If you're doing it for fun, that's understandable, but if your goal is to submit to SDA you'll have to be able to classify the run in a category. I think that optimizing for in game days could count as low% per sda rules, but don't quote me on that.

arkarian

Edit history:

arkarian: 2010-08-25 02:29:59 pm

2010-08-25 02:29:49 pm

gamelogs.org

i think sda should just have both categories— game time and real time. the two methods are different enough from each other that i think 2 categories are warranted. sort of like what majora's mask used to have.

2010-08-25 03:01:58 pm

well, I think the best thing to do is (and here I'm drawing inspiration from Diablo II speedrun) a quick as possible (realtime) - quick combat etc, as I think all games are being accounted for in real hours, minutes, seconds. That way, there will be a starting point, and perhaps a challenge. Eventually, it may turn out a speedrun that includes combat. The way I see it, the strategies are completely different and honestly, I am reading H3 info for the past 6 hours and I feel slightly discouraged at my lack of knowledge about the game. For instance, I did not know that terrain type affects some units' speed - centaurs on grass are + 1 faster than on snow, and so on. My point is that I have no idea how to play or how to optimize for pure speed - realtime gameplay, and I thought since I am playing the game for about 10 hrs a week, I might as well attempt speedrunning it - I saw the only tread about running H3 died around 2007 and nobody really put too much interest in running it.
As other games are presented here - from what I saw - they are pure speed - see Elder Scrolls III: Morrowind, The for Xbox/PC (0:04:19) I don't understand a thing *giggles* I never played ES III but I heard it's a game that requires a long time to play.
To get back on track, I will try to amass more knowledge about H3 and attempt to improve my microing skills - but I never went beyond average in this regard, and to do a proper job you need to micro (I think) the gameplay. Maybe it's too late (midnight here) and I'm tired, I'll pick it up the next few days and hope that I will see the path that needs to be taken by one who would venture to speedrun H3 campaigns.
g'night

well, I think the best thing to do is (and here I'm drawing inspiration from Diablo II speedrun) a quick as possible (realtime) - quick combat etc, as I think all games are being accounted for in real hours, minutes, seconds. That way, there will be a starting point, and perhaps a challenge. Eventually, it may turn out a speedrun that includes combat. The way I see it, the strategies are completely different and honestly, I am reading H3 info for the past 6 hours and I feel slightly discouraged at my lack of knowledge about the game. For instance, I did not know that terrain type affects some units' speed - centaurs on grass are + 1 faster than on snow, and so on. My point is that I have no idea how to play or how to optimize for pure speed - realtime gameplay, and I thought since I am playing the game for about 10 hrs a week, I might as well attempt speedrunning it - I saw the only tread about running H3 died around 2007 and nobody really put too much interest in running it. 
As other games are presented here - from what I saw - they are pure speed - see Elder Scrolls III: Morrowind, The for Xbox/PC (0:04:19) I don't understand a thing     *giggles* I never played ES III but I heard it's a game that requires a long time to play. 
To get back on track, I will try to amass more knowledge about H3 and attempt to improve my microing skills - but I never went beyond average in this regard, and to do a proper job you need to micro (I think) the gameplay. Maybe it's too late (midnight here) and I'm tired, I'll pick it up the next few days and hope that I will see the path that needs to be taken by one who would venture to speedrun H3 campaigns. 
g'night

2010-08-25 03:41:52 pm

You say that you want to be master of air, why ? except haste (wich is not THAT necessary) I don't find this skill necessary.

In the fire skills, I see berserk, armageddon(with immune fire//immune 4th level spell etc...) etc. I think that it's a pretty better skill.

If you master earth, implosion is also really good to clean big groups of ennemies.

2010-08-25 03:58:46 pm

Air has dimension door and fly, which are basically the only spells you would need in a speedrun of this game.

2010-08-25 04:42:39 pm

mmh ok didn't remember theese sorry. There are level 5 so pretty hard to get on the shortest campains... Do we need both of us ? Theese spells, well used, should allow not to kill neutrals (so you don't need diplomacy) but on the other hand you will loose exp and units.
Are you sure that we can learn the two lvl5 air spells in our city ? If I'm right, it's possible to disable some spells from the mage guild. And if so, we will need to build the last mage guild wich is a waste of time(turn) .
I hate not having the game on my comp to check all of this stuff by myself.

2010-08-25 06:39:05 pm

I'd imagine that the way you'd go about getting these spells is luck manipulation. No building is difficulty to get, all you're doing is just spamming end turn until you have enough resources to build it. I would also imagine that leveling your hero isn't even necessary since just macroing and having a unit advantage over the computer is all that's required since the AI usually splits its armies whereas the player doesn't have to.

In all honesty the speed of this run will come down to the player's clicking speed, much like myst, over anything else really.

Edit history:

raymond007: 2010-08-26 12:24:16 am

2010-08-26 12:10:38 am

from
http://speeddemosarchive.com/lang/rules_en.html

"Always remember that speed is the first and foremost priority; side issues such as entertainment are secondary. You can include a neat trick that costs a second or two, but be warned that verifiers will see this as a negative, not a positive. Someone can beat your run later simply by omitting your time waster. If you waste enough time, the verifiers will reject your run outright. Similarly, if a game lets you skip cutscenes or advance through text quickly, then you must do so."

So sheer speed (- on the fun side) however; If anyone ( I hope it's me ) gets to complete a speedrun by the rules, I think can ask for indulgence and perform an "entertainment version"

I will give it a try this weekend, and I'll try to post some results.

arkarian

2010-08-26 02:40:45 am

gamelogs.org

sda would probably allow a game time category. majora's mask used to have one, and afaik the only reason it doesn't now is because there's not enough difference anymore between game time and real time.

2010-08-26 03:11:18 am

I think for H3 there should be 2 categories: quick combat off and quick combat optional. Optional, because in SOD, for sure most games are won by the player rather than the total amount of troops. I know I have won many fights with ridiculously small amount of troops early stages on, and in the "long run" it may prove to be faster than quick-combat as the number of troops lost in a victory would be slower - meaning less time spent building up armies etc. What I notice is that AI will push troops in the middle of the battlefield, while I prefer wait/defend until my ranged troops do full damage, then block them, and so enemy is within full damage of ranged troops and my troops are dealt less than full damage from enemy ranged. Mass slow is a must in these scenarios. However, I think I will try a few runs of RoE on quick combat as I think RoE can be ran without too much spell requirements - thus eliminating luck manipulation for the proper spells, for the hero to pick up wisdom as well as proper magic class. IMHO, fire magic is pointless, I almost never used Armageddon with fire immune units. Blind is just the same and the rest of the spells are useless. Blind is damn good but just as good without fire magic skill.
In terms of time, I think quick combat will also take less time do do the actual speedrun, by eliminating magic and hero buildups, luck manipulation ( 3 layers - wisdom, correct magic class - at the right time, and the right spells in the level) all that combined with no mass morale/luck from computer - I think H3 AI is the biggest cheater of all games, many, many, many times I went Alt-F4 cursing and swearing I will never play this game again. Lol. I love my dread knights not hitting a single double damage/curse until there is a stack of 3-4 of them. But all that is part of the fun, I guess B-)

I think for H3 there should be 2 categories: quick combat off and quick combat optional. Optional, because in SOD, for sure most games are won by the player rather than the total amount of troops. I know I have won many fights with ridiculously small amount of troops early stages on, and in the "long run" it may prove to be faster than quick-combat as the number of troops lost in a victory would be slower - meaning less time spent building up armies etc. What I notice is that AI will push troops in the middle of the battlefield, while I prefer wait/defend until my ranged troops do full damage, then block them, and so enemy is within full damage of ranged troops and my troops are dealt less than full damage from enemy ranged. Mass slow is a must in these scenarios. However, I think I will try a few runs of RoE on quick combat as I think RoE can be ran without too much spell requirements - thus eliminating luck manipulation for the proper spells, for the hero to pick up wisdom as well as proper magic class. IMHO, fire magic is pointless, I almost never used Armageddon with fire immune units. Blind is just the same and the rest of the spells are useless. Blind is damn good but just as good without fire magic skill. 
 In terms of time, I think quick combat will also take less time do do the actual speedrun, by eliminating magic and hero buildups, luck manipulation ( 3 layers  - wisdom, correct magic class - at the right time, and the right spells in the level) all that combined with no mass morale/luck from computer - I think H3 AI is the biggest cheater of all games, many, many, many times I went Alt-F4 cursing and swearing I will never play this game again. Lol. I love my dread knights not hitting a single double damage/curse until there is a stack of 3-4 of them. But all that is part of the fun, I guess B-)

2010-08-26 04:05:04 am

To xsite : You are true if you consider a time speedrun. If you have enough ressources, you can take 5 days building your mage and manipulate it as you wish. In a day/Week/month speedrun, you won't loose 5 days at least for the first campain (you can have the boots pretty early, and the first scenario is pretty linear).
As leveling is pretty slow, you can think about leveling 2 heroes, one for skills like diplomacy and taking the mines you need etc, the other one can be a good fighter. It's a slower start, but as you have 5 heroes at the second part of a campain, your second hero will massively increase the speed of your expansion.

I purpose armageddon because it's a good way to kill small groups, or small hero armies. It's pretty faster than haste// slow everyone.

zoonel

2010-08-26 04:19:30 am

Go play spacechem !

i dont think time speedrun would be as easy as spamming end turn button like xsite believe it is.
Anyway did you choose the patch you are playing with and the difficulty to run? I mention the patch version because without patch you can do gremlin rush / skeleton rush etc...
I like this site for infos:
http://www.heroesofmightandmagic.com/heroes3/
dont forget the forum http://heroescommunity.com/forumdisplay.php3?FID=6

2010-08-26 05:34:14 am

oh snap! i spent 20 minutes writing a post and "sorry, an error occurred when posting
I think I know the way to easily do it. I have rethought the Armageddon and basically i will make every hero a spell caster. Anti-magic your units ( preferably use two stacks only and doublecast anti-magic) then next turn - pow! death and despair! hehehe I don't know why i never thought of this before.
And I don't think building up a town and spamming end turn will get you far, as the AI will play as it normally does; it will just do it quicker. Except the first episode (where you get to build up a hero) most of them you start with a handicap in army size, whereas the AI will have big to massive armies.
But i really dig this Armageddon thing ... I think it can really solve my problem with later stages of the game in terms of speed. the issue is, is it worth it to run all over map to get decent spell power/ knowledge to make it work? in this scenario, the hero build-up in terms of skills are
Earth - town portal, anti magic) slow, shield, stoneskin, resurrection, implosion
Air - Dimension Door, Fly, Haste,
Fire - Armageddon
Logistics - increased movement ( early on, so that it will take less to chase all shrines etc )
Wisdom - required to cast all those spells
Intelligence - see above Smiley

that is 6/8 skills a hero normally gets. most heroes get static defined skills (2) in the beginning, and it might be possible to luck manipulate the rest. (some start with one of the ones i think necessary, and it might be replaced by Sorcery) Crag Hack i fear has a built in skill tree, i think he will always get the same skills, but i need to educate myself on this if it's true or not.
thanks to all for your input, it seems more achievable now happy

oh snap! i spent 20 minutes writing a post and "sorry, an error occurred when posting
I think I know the way to easily do it. I have rethought the Armageddon and basically i will make every hero a spell caster.  Anti-magic your units ( preferably use two stacks only and doublecast anti-magic) then next turn - pow! death and despair! hehehe I don't know why i never thought of this before. 
And I don't think building up a town and spamming end turn will get you far, as the AI will play as it normally does; it will just do it quicker. Except the first episode (where you get to build up a hero) most of them you start with a handicap in army size, whereas the AI will have big to massive armies. 
But i really dig this Armageddon thing ... I think it can really solve my problem with later stages of the game in terms of speed. the issue is, is it worth it to run all over map to get decent spell power/ knowledge to make it work? in this scenario, the hero build-up in terms of skills are 
Earth - town portal, anti magic) slow, shield, stoneskin, resurrection, implosion
Air -  Dimension Door, Fly, Haste,
Fire - Armageddon 
Logistics - increased movement ( early on, so that it will take less to chase all shrines etc )
Wisdom - required to cast all those spells
Intelligence - see above :) 
that is 6/8 skills a hero normally gets. most heroes get static defined skills (2) in the beginning, and it might be possible to luck manipulate the rest. (some start with one of the ones i think necessary, and it might be replaced by Sorcery) Crag Hack i fear has a built in skill tree, i think he will always get the same skills, but i need to educate myself on this if it's true or not. 
thanks to all for your input, it seems more achievable now :happy:

Edit history:

trex7g2: 2010-08-26 06:52:57 am

2010-08-26 06:51:56 am

Be carefull about magic mirror : ENNEMY spells cast on targeted, allied troop has a 40% chance of being redirected to a random enemy troop. And I am not absolutely sure that AOE are reflected. But you can use :
protection from fire "Damage from Fire Magic spells is reduced by 50% for all allied troops"

units :

Gold Dragon "Immune to spells levels 1-4"
(War) Unicorn "Aura increases magic resistace of all adjacent creatures by 20%"
(Battle) Dwarf "4(2)0% magic resistance"
Efreet (Sultan) "Immune to fire"
vampire lord "resurrects members of their own stack by restoring health equal to the amount of damage they do to living enemies"
Black Dragon "immune to all spells"
Fire Elemental "immune to all mind and fire school spells"
golems :"damage from spells reduced by xx%"
Archangel : "Resurrect dead allied troops once per combat" (really efficient if you make 2 groups of them : even if you loose 1 or 2 of them, they will resurect each other and you will be able not to loose any)

spells :

ANTI MAGIC : Target, allied troop can only be affected by level 5 spells at advanced earth, immune if master (really cool with doublecasting)
protection from fire "Damage from Fire Magic spells is reduced by 50% for all allied troops"

You should remember that for your last fight, you just need to survive, even if you sacrifice 99% of your troops on your own spell (for example, if you have only fire elementals, you can finish the enemy with you armageddon, you will win even if you loose all your other troops

Be carefull about magic mirror : ENNEMY spells cast on targeted, allied troop has a 40% chance of being redirected to a random enemy troop. And I am not absolutely sure that AOE are reflected. But you can use :
protection from fire "Damage from Fire Magic spells is reduced by 50% for all allied troops"

units :

Gold Dragon "Immune to spells levels 1-4"
(War) Unicorn "Aura increases magic resistace of all adjacent creatures by 20%"
(Battle) Dwarf "4(2)0% magic resistance"
Efreet (Sultan) "Immune to fire"
vampire lord "resurrects members of their own stack by restoring health equal to the amount of damage they do to living enemies"
Black Dragon "immune to all spells"
Fire Elemental "immune to all mind and fire school spells"
golems :"damage from spells reduced by xx%"
Archangel : "Resurrect dead allied troops once per combat" (really efficient if you make 2 groups of them : even if you loose 1 or 2 of them, they will resurect each other and you will be able not to loose any)

spells :

ANTI MAGIC : Target, allied troop can only be affected by level 5 spells at advanced earth, immune if master (really cool with doublecasting)
protection from fire "Damage from Fire Magic spells is reduced by 50% for all allied troops"

You should remember that for your last fight, you just need to survive, even if you sacrifice 99% of your troops on your own spell (for example, if you have only fire elementals, you can finish the enemy with you armageddon, you will win even if you loose all your other troops