Megapack Balance

[claymore]

I don't see this being addressed in a hurry, but I think it's pretty important.  I could make a huge list, but here are some imbalances I can think of off the top of my head for Magic:

• The evil dragon upgrade is absolutely worthless.  Expensive, very time consuming, and barely stronger than a regular dragon.
• Have the Hell Gate upgrade depend on having built an Archmage Tower.  Behemoths should be a later stage unit.
• Ghost armours are worthless.  Make them cheaper, or have improved armour, or be invisible, or respawn after a while, or something...
• Summoner's units should be cheaper, summon more quickly, and require summoner's mana/energy to create.
• Remove the dependency of some units/buildings  of a summoner being existent somewhere.
• Have upgrades for the daemon scout to make it faster, and maybe more durable.  Otherwise, I don't tend to use them.
• Wicker Behemoth is a pretty pointless intermediate stage of behemoth creation since it's completely non-functional, and you'll immediately want to upgrade it to Behemoth.

Titi?  Anyone?  What do you think?

[Omega]

Well, let's see, I do agree the evil dragon could be made a bit more powerful, although it is pretty powerful as it is. Perhaps increasing its defensive abilities (HP and armour) would be a better idea?

I'm neutral to having hell gates depend on the archmage tower. I suppose the behemoth should be a later game unit, but I've always found it to be so clumsy that I never use it. I think we should address that first. The biggest problem is that its large size makes it difficult to reach a foe and it takes more damage from splash attacks too. I think adding a ranged attack where it picks up a rock and throws it would be a good idea. However, the AI would also need to be improved as to know when to use the ranged attack instead of the more powerful melee attack (namely if it takes too many steps to reach a foe).

I agree, ghost armours are a bit difficult to get. I think it'd be easier if we simply removed the upgrade prerequisite and just require the summoner's guild to exist. Being invisible or respawning would be overpowered and not currently possible.

I don't think the summoner should be cheaper or faster, but I do think the EP cost of their attacks should be reduced. I can't help but think that we should also be able to set a minimum EP for when a unit is produced. As it stands, all units start with zero EP. So if I pump out a summoner into a battle, she'll definitely die, as she won't be able to attack immediately. This applies to all units. I'd like to see units starting with full (or maybe half?) EP, or perhaps even more preferably, let the modder set a "starting EP" in the XML.

I didn't realize some buildings required a summoner, and agree there's no reason for a building to require a mobile unit. Building prerequisites should be limited to other buildings and upgrades.

By upgrades to the daemon, do you mean morphing the daemon into something stronger (disagree) or an upgrade to increase the stats of all daemons (agree)? While the daemon is intended to be a cheap swarm unit, they do become pretty useless late in the game, and increasing their HP and attack slightly could help keep them slightly useful (though they'd remain a swarm unit, which is still useful for holding back enemies).

As for the wicker behemoth, why not just have the behemoth be built, but make the "be built" model the wicker behemoth model? That would make the behemoth similar to the golem in terms of creation, as I agree, the intermediate stage is just annoying and unnecessary. However, since victory is considered to be when all buildings are destroyed, and buildings are defined as units with a be_built skill, won't we need a way to flag units as not being required to kill for victory? Not sure offhand what's happening to the golem currently.

To be honest, if we're going to do a MegaPack update, I can't help but think we should consider going all the way. Ask for community help in revamping the MegaPack entirely. For example, the Indian faction is mostly retextured models. The icons of some of the earlier factions are in desperate need of an update, and so on. I can't help but think that while MegaGlest keeps getting better, the MegaPack has been stuck in limbo, despite being the flagship mod.

[claymore]

I'm neutral to having hell gates depend on the archmage tower. I suppose the behemoth should be a later game unit, but I've always found it to be so clumsy that I never use it. I think we should address that first. The biggest problem is that its large size makes it difficult to reach a foe and it takes more damage from splash attacks too. I think adding a ranged attack where it picks up a rock and throws it would be a good idea. However, the AI would also need to be improved as to know when to use the ranged attack instead of the more powerful melee attack (namely if it takes too many steps to reach a foe).

I agree that Behemoths are somewhat clumsy to use, but they're much tougher in the current version of MG than they used to be.  Now I would argue that they are most definitely worth creating.  Most of Magic's units seem to be ranged, so it would be nice to keep it as a melee meat shield.

I agree, ghost armours are a bit difficult to get. I think it'd be easier if we simply removed the upgrade prerequisite and just require the summoner's guild to exist. Being invisible or respawning would be overpowered and not currently possible.

I kind of like the fact that only one of the summoner's summonable units is available initially... I'd be for having to research it (though maybe a dedicated "Ghost Armour" technology at the summoner's guild rather than something seemingly unrelated at the library), but for it being worth researching.  I was perhaps being a bit optimistic with the abilities I suggested, but it could certainly be cheaper and tougher.

I don't think the summoner should be cheaper or faster, but I do think the EP cost of their attacks should be reduced. I can't help but think that we should also be able to set a minimum EP for when a unit is produced. As it stands, all units start with zero EP. So if I pump out a summoner into a battle, she'll definitely die, as she won't be able to attack immediately. This applies to all units. I'd like to see units starting with full (or maybe half?) EP, or perhaps even more preferably, let the modder set a "starting EP" in the XML.

This really wasn't what I was getting at at all... I'm talking about the EP cost of the summoner's units, not the cost of her attacks.  Her attack is close to worthless anyway.

By upgrades to the daemon, do you mean morphing the daemon into something stronger (disagree) or an upgrade to increase the stats of all daemons (agree)? While the daemon is intended to be a cheap swarm unit, they do become pretty useless late in the game, and increasing their HP and attack slightly could help keep them slightly useful (though they'd remain a swarm unit, which is still useful for holding back enemies).

I'm thinking of a stats upgrade, though it would be nice if this could be represented visually by automatically morphing all existing daemons into some more advanced form.  Unfortunately, this doesn't seem to be possible yet.  (Also due to this, Golems have to be tediously morphed into Power Golems by hand.)

As for the wicker behemoth, why not just have the behemoth be built, but make the "be built" model the wicker behemoth model? That would make the behemoth similar to the golem in terms of creation, as I agree, the intermediate stage is just annoying and unnecessary. However, since victory is considered to be when all buildings are destroyed, and buildings are defined as units with a be_built skill, won't we need a way to flag units as not being required to kill for victory? Not sure offhand what's happening to the golem currently.

I like the Wicker Behemoth model, it just seems a shame to waste it.  Also, it doesn't make much sense to construct an organic unit like a building.

[Omega]

This really wasn't what I was getting at at all... I'm talking about the EP cost of the summoner's units, not the cost of her attacks.  Her attack is close to worthless anyway.

The summoner's units don't have EP costs. I'm presuming you're referring to the speed value of the produced units (ie, how long it takes to summon a unit)? I'm fine with those, especially since you can have lots of summoners. However, I do think the EP cost of attacks is a serious issue in balance. Let's say that Tech's initial army is wiped out, so the player has to quickly pump out units to defeat the attackers. Archers, swordmen, guards, they can all attack right away. But initiates, battlemages, summoners, they can't. They need EP to attack, and when they are produced, they start with zero EP. Thus, they can't immediately attack a foe, and are slaughtered before they even get off an attack. It's particularly bad with the summoner and archmage, but still highly noticeable with the battlemage and initiate. The factions that depend heavily on EP, namely magic, are therefore at a huge disadvantage compared to the factions that don't use EP. I'm aware it's not what you were initially referring to, but it's also a valid concern.

I like the Wicker Behemoth model, it just seems a shame to waste it.  Also, it doesn't make much sense to construct an organic unit like a building.

Well, we already construct the wicker behemoth, and then turn it into an organic unit, and we construct the golem in a single step. If we use an animation for the be_built skill of a behemoth and stretch that animation to cover the be_built skill (as is already possible), we could replicate the building of a wicker behemoth that then starts to glow and transform into a regular behemoth. No models wasted.

[claymore]

The summoner's units don't have EP costs...

They don't, but I'm suggesting adding them.

Well, we already construct the wicker behemoth, and then turn it into an organic unit, and we construct the golem in a single step. If we use an animation for the be_built skill of a behemoth and stretch that animation to cover the be_built skill (as is already possible), we could replicate the building of a wicker behemoth that then starts to glow and transform into a regular behemoth. No models wasted.

If this is possible, sure.

[Omega]

The summoner's units don't have EP costs...

They don't, but I'm suggesting adding them.

What is the reason for adding them? I'm not sure there'd be an improvement. With the current way that EP works on produce skills, it's pretty much impossible to have an EP cost, since they'd consume the amount of EP for each animation cycle, and will cancel if it runs out of EP in the middle of an animation. We'd require a change to the EP system on such skills in order for such a thing to work.

At any rate, I fail to see what an EP cost to summoning would add. It would reduce her ability to attack inbetween summons, but with the length of each summon, by the time the unit is produced, she'd have enough EP to summon another. As well, I worry that an EP cost could potentially prevent queuing produce commands (not sure if this would work). I'm presuming that the intent is to have some type of cooldown between summons? The EP system won't really work for that as it stands, so I'd say we'd need a new feature for cooldown periods between commands (it'd be useful for other purposes too, such as preventing archmages from using their stronger fireball attack as often).

[claymore]

What is the reason for adding them? I'm not sure there'd be an improvement. With the current way that EP works on produce skills, it's pretty much impossible to have an EP cost, since they'd consume the amount of EP for each animation cycle, and will cancel if it runs out of EP in the middle of an animation. We'd require a change to the EP system on such skills in order for such a thing to work.

At any rate, I fail to see what an EP cost to summoning would add. It would reduce her ability to attack inbetween summons, but with the length of each summon, by the time the unit is produced, she'd have enough EP to summon another. As well, I worry that an EP cost could potentially prevent queuing produce commands (not sure if this would work). I'm presuming that the intent is to have some type of cooldown between summons? The EP system won't really work for that as it stands, so I'd say we'd need a new feature for cooldown periods between commands (it'd be useful for other purposes too, such as preventing archmages from using their stronger fireball attack as often).

Well, it was just an idea... but I felt that summonables should be cheaper and quicker to produce, and EP would help to restrict their production.  It also makes some sense that EP should be expended on spells.  Her attack would no longer cost EP.  If there are engine limitations, that's a shame... I wasn't entirely sure what the capabilities of the engine were.

[bexar]

What is the reason for adding them? I'm not sure there'd be an improvement. With the current way that EP works on produce skills, it's pretty much impossible to have an EP cost, since they'd consume the amount of EP for each animation cycle, and will cancel if it runs out of EP in the middle of an animation. We'd require a change to the EP system on such skills in order for such a thing to work.

At any rate, I fail to see what an EP cost to summoning would add. It would reduce her ability to attack inbetween summons, but with the length of each summon, by the time the unit is produced, she'd have enough EP to summon another. As well, I worry that an EP cost could potentially prevent queuing produce commands (not sure if this would work). I'm presuming that the intent is to have some type of cooldown between summons? The EP system won't really work for that as it stands, so I'd say we'd need a new feature for cooldown periods between commands (it'd be useful for other purposes too, such as preventing archmages from using their stronger fireball attack as often).

Well, it was just an idea... but I felt that summonables should be cheaper and quicker to produce, and EP would help to restrict their production.  It also makes some sense that EP should be expended on spells.  Her attack would no longer cost EP.  If there are engine limitations, that's a shame... I wasn't entirely sure what the capabilities of the engine were.

Well you can kind of set the value for yourself if you can change xml files.  Here is an example of the daemon having an ep cost.  I have been messing around with this the past few days and it works really well.

Code: [Select]

<skill>
<type value="attack"/>
<name value="help_skill"/>
<ep-cost value="100"/>
<speed value="80"/>
<anim-speed value="80"/>
<animation path="models/summoner_producing.g3d"/>
<sound enabled="true" start-time="0">
<sound-file path="sounds/summoner_ack1.wav"/>
</sound>
<unit value="daemon" amount="1"/>
<attack-strenght value="0"/>
<attack-var value="0"/>
<attack-range value="12"/>
<attack-type value="piercing"/>
<attack-start-time value="0.5"/>
<attack-fields>
<field value="land"/>
<field value="air"/>
</attack-fields>
<projectile value="false"/>
<splash value="false"/>
</skill>

<command>
<type value="attack"/>
<name value="call_reinforcements"/>
<image path="images/summoner_attack.bmp"/>
<unit-requirements/>
<upgrade-requirements/>
<move-skill value="move_skill"/>
<attack-skill value="help_skill"/>
</command>
_{Edit by Omega: Fixed code formatting}

[Omega]

That's certainly an interesting take on it, using the attack spawn feature instead of a produce command. The major flaw, however, is that the AI presumably won't use them correctly (I'm actually not sure if the AI uses secondary attacks ever, barring attacking units in different fields).

Granted, I'm not entirely sure that the summoner needs an EP cost, given that units like the daemon are such weak units, and an EP cost just further makes the summoner useless for attacking (she already runs out of EP very easily).

[bexar]

ya I have been testing that today and the AI definitely will not use that attack.  It really is not a problem for the summoner.  I adjusted her stats to make it a bit easier. The reason why I give it an EP cost is because she will keep summoning a new unit literally every second.  So if you had say 5 summoners out at a time using that skill you could easily overwhelm every opponent on the map in just a few minutes.  So now the summoner is more like a real summoner should be and does not need to attack.  The way the skill is set up you can have the summoner use that skill on one of your own buildings to create the units.

And you can replace the daemon with whatever unit that you want.  But it is easy to get carried away with the units that you put on there.  Hehe I had fun slaughtering the entire map the first few attempts at that

[MoLAoS]

I actually had a problem with using EP and other costs multiple times when I was adding skill costs/command costs to mandate. Basically the design of Glest was applying costs in a bad way and the way it uses EP because of attack skills was dumb. If you want costs on multiple commands besides attack you'll either need to separate EP costs based on the skill or just directly rework how costs are applied.

[biel]

You can look at my post for an idea balancing magic: https://forum.megaglest.org/index.php?topic=9081.0

I think magic is completely unblanced.

[MoLAoS]

You can have auras in GAE. They are called Emanations and use the effect_type class. Mandate has that as well. However neither of those engines are suitable for multiplayer and balance is something of a waste of time in single player since the computer will always be too dumb to beat a human. Plus you cannot even play Magitech on Mandate anyways.

Adding the effect_type and emanation classes would take prohibitively long for the MG devs. You might as well just use the Attack Boosts.

I should also note that Magic is probably balanced, its just not the same play style.

[biel]

Yes but it only have to be done once and will imporove the modding possibilities. Anyway, are my modifications a good idea?

[D.U.P.A.]

The game really needs some balance changes, starting from ones claymore stated. Also to eliminate oddities like artillery (catapults) having same range as archers. There should be a dedicated person to balance. Good thing that it doesn't need to be a developer, just someone who is good at game, modifying xml it's a simple task.

[Omega]

The game really needs some balance changes, starting from ones claymore stated. Also to eliminate oddities like artillery (catapults) having same range as archers.

That's an interesting point. However, there's a reason for the limited range of the catapult. If the range becomes too great (ie, attackers range > defender's sight), the defender will be unable to see the attacker and will sit around while being attacked. This would require engine improvements to get around (although you or anyone else is welcome to submit patches). And since units cannot attack outside of their sight, the attacker's sight must be at least as large as their range (for example, a unit with a sight of 15 cannot attack a unit 16 tiles away, even if it's within their attack range). If I recall correctly, however, it is possible to attack if another unit has the target within their sight (think: spotter). It's a bit hard to apply this strategy in an actual game, though (I only ever used it for a "super weapon" in AD).

You're probably wondering why we can't just increase the sight of the catapult and be done with it. We could, but the game plays rather poorly when there's a large variance in the sight of units. For example, say I have a cluster of two different types of units, and one of them has twice the sight distance of the other. This means they'll see the enemy earlier and start attacking, while the short-sighted unit sits around doing nothing.

Further related to sight is the fact that catapults are usually used for anti-building attacks. So what happens if some super-sight catapult sneaks around an enemy camp and starts attacking buildings (which usually have poor sight). Most buildings can't attack, so just having a unit detecting when it's being attacked isn't much help (and losing your main building is usually a game ender).

Almost all of these issues could be improved with a better AI. The AI units must communicate with each other instead of working alone. It should understand how to properly attack and defend (avoid leaving military units idle, for example). These kinds of changes would go a lot further than tweaking numbers in the XML files. While the MegaPack is certainly not perfect, it's balanced quite well in human tests. I maintain the primary issue is with our AI.

[D.U.P.A.]

Yes, catapult is very effective against buildings. But I wanted to have them to be used also against massed archers, because of their splash. But their short range means they will focus more on melee units, destroying also my units in the process. This means I need to be concentrated only on catapults, not to mean that I need to push them almost in the frontline to attack archers. There is the same problem with archmages (while being cost prohibitive). If the sight variance is the problem, then there should be a global sight increase of units (or decrease of range of certain units).

[biel]

A great improvement to the AI would be that units that have ranged & splash attack order targets by priority:

• The units that can deal more damage to your team (+ priority)
• The units that if get attacked by the splash damage, the attacker would damage itself (- priority)
• The units with larger size (+ priority)
• The units with more armor. The armor is calculated for each attack (or i'm wrong) so if the attack deals lots of damage, is better to attack armored foes. The less armored units can be killed by weaker or more machine-gun units. (+ priority)
• The units with more benefit from damage multitplier (+ priority)

In fact, that can be extended (with some tweaks and addings) to all attack types and strategies for the AI. It would be great that when a unit has multiple targets to choose get in sync with all units in range to deal the highest damage possible calculating the damage each unit will get for all possible cases and to eliminate enemy units that will deal more damage first (eliminate a catapult first than a swordman but if there are 8 swordmands that a catapult can hit all with a shoot, then the catapult focus swordmans).

Do you get my idea?

[MoLAoS]

I actually wrote part of the special AI for my project to calculate damage on targets. It both picks the optimal move damage wise including mana/energy w/e and which target is best to hit. It doesn't even need strings to handle this, it just plows through the damage resistance list because all damages have a resist and the list is the same for all units.

I think MG uses a different armor damage system though. But its definitely possible to add a lot of this stuff. The most cpu intensive part of attacks is building the list of attackable units I think, so adding priorities shouldn't slow it down too much.

[biel]

I haven't looked at the ai code but i think the CPU AI already creates a list of attackable units. If it dosen't, how does it choose target if there are multiple?

[MoLAoS]

I haven't looked at the ai code but i think the CPU AI already creates a list of attackable units. If it dosen't, how does it choose target if there are multiple?

I didn't say it didn't create a list. I said that was the most CPU intensive part of attacking. Calculating a list.

[biel]

A great improvement to the AI would be that units that have ranged & splash attack order targets by priority:

• The units that can deal more damage to your team (+ priority)
• The units that if get attacked by the splash damage, the attacker would damage itself (- priority)
• The units with larger size (+ priority)
• The units with more armor. The armor is calculated for each attack (or i'm wrong) so if the attack deals lots of damage, is better to attack armored foes. The less armored units can be killed by weaker or more machine-gun units. (+ priority)
• The units with more benefit from damage multitplier (+ priority)

In fact, that can be extended (with some tweaks and addings) to all attack types and strategies for the AI. It would be great that when a unit has multiple targets to choose get in sync with all units in range to deal the highest damage possible calculating the damage each unit will get for all possible cases and to eliminate enemy units that will deal more damage first (eliminate a catapult first than a swordman but if there are 8 swordmands that a catapult can hit all with a shoot, then the catapult focus swordmans).

Do you get my idea?

Anyway, (question to the MG developers) is this going to be implemented in the game AI?

[D.U.P.A.]

Definitely smart onagers, like in age of empires, would be really needed. Currently it is a pain controlling them, because they seem to attack melee enemies first, slaughtering your own troops. I think that auto attack should not target any enemy in case it could harm your own units.

[Ishmaru]

I haven't looked at the ai code but i think the CPU AI already creates a list of attackable units. If it doesn't, how does it choose target if there are multiple?

Don't know if this behavior has changed but at one time the AI was upgraded to prioritized warrior (units able to attack) over non fighting units, as long as they are in its own attack range (not sight range). So AI wouldn't ignore attacking a hostile enemy right next to them, because they were attacking a farm. This only works if the AI does not have to move to attack a hostile outside its attack range, even if its only by 1 square.  The way AI is set up once it starts attacking it wont move from its location until there are no more hostiles in attack range, Then it will move toward the spot of its last target and engaging any other hostiles within its sight range.
Once again I'm not certain this was changed.

If I recall correctly, however, it is possible to attack if another unit has the target within their sight (think: spotter). It's a bit hard to apply this strategy in an actual game, though (I only ever used it for a "super weapon" in AD).

Yes its very possible for a player to do this, however the AI seems to be able to skip the spotter and attack even from beyond its vision. Doesn't bother me much as players get to see the units attacking from outside our vision.

Almost all of these issues could be improved with a better AI. The AI units must communicate with each other instead of working alone. It should understand how to properly attack and defend (avoid leaving military units idle, for example). These kinds of changes would go a lot further than tweaking numbers in the XML files.

Agreed.

[biel]

My idea on how to do this is create a class "AttackSimulator" or a thing like this in the AI that tries all combinations and gets all resulting hp and ep for your allies and enemies for all units in all target combinations and finally returns the best target for each unit controlled by the AI.

This seems to be very CPU costly (especially with large amounts of units) but i think it isn't necessary to calculate all the variables exactly as float / double  data types. An integer calculation should be fine.

I want to (try to) learn C++ in-depth so if anyone tells me how the .patch files and all about that works, i'll try to do this myself.