[1.5] AI ideas for sharpening its sword (Hero usage and traits)

I think it would be an improvement to the AI if they couldn't buff themselves in combat.

don't think a global ban on tactical buffs would help the AI much. it's a bit ridiculous when you're already steamrolling and the AI champs insist on buffing when half their army is already slain in turn 1, but at that point it doesn't really matter anyway. not wasting a few points of mana and instead do some insigificant damage to your killer stack probably wouldn't turn the battle around.

In almost all cases, a buff isn't worth the turn and mana spent casting it.

Bloom's Taxonomy shows the hierarchy of thinking processes.  The highest that anyone has ever been able to code into AI is Analysis.

Synthesis and Evaluation are still solely in the domain of humans as of now.  Evaluation, in particular, is what you really need in order to be able to determine whether it is worth the mana and the turn to cast a buff and the requisite synthesis framework needs to be in place before that can even be attempted.

I really don't know how the AI is programmed to decide about casting a buff or not, but it may not even be as high as the Analysis level.  It may just be in the Application level.  If so, that just puts one more step between Evaluation and where the AI is.

The value of a turn is really high and the value of a buff is variable and potentially really low.

That is the crux of the problem, if the AI can't be programmed to determine when the value of the buff exceeds the value of the turn, then much of the time it will sit around doing useless things.

If the AI wants to cast a buff outside combat when it doesn't waste critical turns, I am all for that.  The value of the buff may still be low, but at least it won't waste the most important resource in the game (turns).

I also think that the AI shouldn't be able to spend more than half of its mana generation on strategic buffing, but I don't want what's left to be spent on tactical buffs either.  The spells the AI should be using with its mana in combat are the ones that players use.  The ones with dramatic and mostly immediate effects.  

Buffs just don't fit the bill and I would hazard a guess that very few players use any buffs other than haste and only then because it is the only buff that doesn't cause a loss of turns (long term) and indeed causes you to have an increase in the single most important resource in the game (again, turns).

I don't really understand this reasoning. Some of the buffs (Giant form, Diamondskin) are quite powerful. The AI in my current game likes casting Cloak of Fear, which effectively means my melee units can't damage it for five turns. Focus has its place as a damage multiplier, as well.

I normally agree in games like this that movement and speed are enormously powerful, but as long as you move before casting a buff, you don't actually lose a turn, especially in the first turn when you probably can't attack in melee anyway.

I would agree that the current version of the AI maybe casts buffs a bit too much. This is essentially a tuning question though, it used to barely cast spells at all.

I would also agree that the AI should aim to keep a decent size store of strategic mana, especially if it is aiming to cast the Spell of Making. So in other words don't cast spells if it has less than 250 mana (or whatever).

I guess I can't remember the last time I had an Earth Archmage so I think I have only ever cast Diamondskin once and Giant Form never.  Also Cloak of Fear never.

I tend to play humans most often, so getting into melee isn't usually a problem.  If not I am usually playing with horses, tireless march, and so on.

I would rather see the AI casting something offensive if it can't get to where it can attack rather than having it buff itself.  A successful Slow on a random target is going to be better than a Flaming Blade most of the time.

It's not that I don't want it to cast spells, its just that buffing is almost always a waste of precious combat turns.

Buffing is highly situational.  First off you need to determine whether you can wipe out the enemy before they get tot strike.  If not, you need to determine whether the buff will shorten the time it takes to obliterate the opposition (including the turn to cast the buff).  Also you need to determine if there is an enemy unit (usually a hero) that REALLY needs to be killed fast, if so, do you have the capability to do it (like shadow shift it into the middle of your army, or using the Kill spell).

If death can't be ascertained in 1 turn, you start looking at two turn tactics.  Maybe do huge AOE damage first, then buff your soldiers with your second turn (this is actually very effective), and let your soldiers mop up the dregs.

Overall tactical buffs are weaker than tactical debuffs/damage, however there are some beautiful ones that can really turn the tide in the right situation.  Cloak of Fear is AWESOME.  never underestimate it.  Spell check or no attack is nuts. Giant form (double attack no penalties), can make that armour they are wearing suddenly seem a lot thinner than it was (especially if you're not playing as evil and can cast mass curse).  Diamond skin on that assassin you just shadow shifted to the back of the other army, letting your army kill the poor (usually hero) unit that is now in the middle of your army.

Buffs and debuffs all have a place.  But they are very situational.  My post was really about just optimising heroes' traits, which should be easier from an AI perspective (i think? ).  Tactical battle AI wasn't touched on because that is a completely different kettle of fish (imagine trying to code when a unit should step forward one square past an enemy unit....... very difficult for AI, very easy for human).

So you basically play the 'same' template to defeat the ai as I said in another thread. point proven. People find a template that works and they tend to use it over and over. that's not outthinking the ai that's just outthinking the mechanics of the game over and over. Play with random skills and traits (roll them up or draw chits) and don't use those mages every chance you get....do something else. I bet you won't win so easily or hands down then. :))

Good job proving that I am not an idiot.  You did that very well.  Thanks.  

I never was trying to argue that I am an idiot or that I play like an idiot, but it is at least helpful for you to verify that I am not and I don't.  

We all very much appreciate your efforts.  Run along, it's time for adults to talk to adults now.

The key thing in all of this is to try to maximize outcomes.  That means you want to win while using the fewest possible resources in the process.

If you have the choice between attacking (uses 0 mana) and buffing (uses > 0 mana) and the outcomes of both actions are the same (you win) then it is strictly better to save the mana and just attack.

This is the primary "fighting" stage I am talking about here, when victory is in question.  Once there is no question and victory is essentially guaranteed then for some people ending the game as quickly as possible becomes the primary concern.

In the fighting stage, being able to save X mana in fight Y and use it in fight Z can make a difference so if you could win with no buffs then it doesn't make sense to buff.  It is only when the buff causes you to use fewer resources in total over the course of the battle that it becomes worth using.

Something like Giant Form rarely leads to fewer resources being used.  It costs mana and it doesn't deal damage immediately. It is only 2 unit-turns later that you do better than breaking even.  More if you count the mana used.

When you abstract it all out to attack and defense ratios, the flaws become more obvious.  If you say, for example, that you have an army with 500 offense and 300 defense, is it really worth it to spend a turn in battle and to spend mana to temporarily make that a 550 offense and 300 defense army?  You have to have a pretty small army for the extra offense to dramatically impact your total damage output by doubling one unit's output.  The time when Giant Form is most effective is when you have an army size of 1 unit.

The game, though, doesn't really reward using small armies.  Armies are made for her pleasure, the bigger the better.  The whole nature of the game works to minimize the effectiveness of Giant Form.

I agree with you completely Raidin.  Win with the least resources.

However, losing units, or having to wait turns for units to heal is also an economic loss. These need to be factored into the equation too.  And in my mind these two facts is what generally makes buffs actually good to cast (in certain situations).

I don't know if I mentioned it above, but I rarely cast spells until I can enchant heroes with Mantle of the oceans.  That enchantment completely changes the resource cost aspect of what to do with your turn (buffs become a whole lot more cost effective).

I've found giant form best to use on Hero warriors with already insane attack levels.  It can get high enough to 'break' the armour of most armies.

@Sanderson, I play strategy games to find synergies in the game mechanics and exploit them.  I posted here so Brad(frogboy) could see someone elses ideas of synergies with the current game mechanics, and he can or not decide to have the ai use the same synergies available to us meager humans.

@Raiddinn:

I used to think exactly how you do, however I do feel that the spell 'Haste' is an exception to your rule and deserves mention here.  I seem to often encounter situations lategame where I have 5+ air power from shards/wonders.  At this point, imo, haste (generally on your most powerful hero or unit) becomes a very worthwhile buff if you expect the fight to last more than a few more turns.  It costs very little mana but can increase your initiative by more than 50% and thus give you many more of that most valuable resource -- turns.  The non-mana cost is only one turn, often less than one turn if you aren't able to get to the enemy instantly (or if you can only reach targets that aren't of particularly high value to attack). I often find at this point that when my hero casts haste it gives him another turn immediately due to his increased initiative, making haste becomes essentially free turn-wise.  I feel that 3-5 mana is often a worthwhile trade in exchange for the tons of HP I am saving by killing the enemy so much more quickly. 

It bears mentioning that if the enemy only has one (or one significant) unit and my water power is equal to or greater than my air power I will instead slow that unit.

- WK

I don't know if I mentioned it here or only in other places, but I have singled Haste out as one of the few really useful buffs because it often pays itself off in turns and because the mana cost of it is very low.  It doesn't suffer from a lot of the drawbacks that other buffs suffer from.  In that respect it is completely different from things like Giant Form.

More whole turns is good in a way that more partial turns (Giant Form --> Attack twice in the same turn, effectively) can never be.

Really interesting discussion. Other than Haste I don't cast buffs very often, but that's mainly because I tend to play with Death/Air/Water mages. Life has Growth, Fire has Focus (incredibly cheap and +25% damage is probably worth using a turn). It's Earth who has the best buffs though. I agree the trade-off is between mana cost and other costs, particularly healing units or worse, having a high level unit die. In an average late game battle I might cast only Haste at a cost of 2 mana, and otherwise use free skills, because I can win the battle with minimal damage without casting anything else. In a more challenging battle I will decide whether I want to use a more expensive spell on the principle that it will mean I use less of other resources, i.e. I can kill the enemy faster, taking less damage myself. My play style means that a more expensive spell is generally Raise Skeleton Horde or Blizzard, but if you had an Earth mage it might be Giantform or Diamondskin. I'm not a huge of Diamondskin, the effect is over very quickly, and the spell is expensive, but Giantform is good.

I agree Mantle of the Oceans is important for getting the cost of spells down in general. In my case it makes a massive difference to the cost of spells like Blizzard and Kill. I'd always take a Water hero if I don't have Water already, just to get Freeze and Mantle of Oceans.

The +25% damage Focus spell pays itself off in 5 unit-turns.  It takes 4 turns for 25% extra to equal the one turn you wasted doing nothing and the fifth turn puts you 25% of one turn's worth of damage ahead.

I have never cast that one before and I doubt I ever will.

Yes, but it can still be worth casting if you have limited mana or nothing better to do.

E.g. melee hero moves towards enemy, not in contact with enemy, can choose to either use throwing knife or casting Focus, (not a Mage so casting ranged damage spells is ineffective and expensive), I'd argue casting Focus for +25% Attack for the rest of the battle is almost certainly the correct action.

or: I have 50 mana available; my best spell is Fireball which I cast for 30 mana in one turn. I would normally use Fireball as my first action to take out as many of the enemy ranged troops as possible. However I can only cast it once; provided I can act quickly enough (i.e. before the ranged troops act, which is perfectly possible with Haste or Command) it makes sense to cast Focus first to maximize the damage.

That's quite a specific case, but it's only an example. The principle applies to any battle which you expect to run out of mana, which I do quite regularly in close battles at the start of the game. Those battles tend to last a long time as well, so buffs which last indefinitely (potentially for ten or more actions if I have high initiative and it's a close battle) become much more worthwhile.

Here are my suggestions, as well as a new ready to use mod to try: https://forums.elementalgame.com/450250/page/1/