Fatigue question

I asked this on the other forum but am not persuaded by the answers. I am trying to figure out exactly when the AI will decide not to honor your spellcast script based on some judgment that the gems or fatigue aren't "worth it." Basically, I'm trying to understand how to ensure that a mage casts a high fatigue/gem costing spell without the AI quietly vetoing it.

Let's say I have a caster with A3 who wants to cast Living Clouds, which requires A4 and has 200 fatigue. If I give him 3 air gems and script him to cast Living Clouds on round 1, will he do it? It seems to me he can use 1 gem to boost to A4, and 2 gems to cover the minimum, and so he can cast it and get 200 fatigue.

But my gut tells me even though he can technically do it, the AI won't let him, because of some amorphous judgment that this is a bad idea and there is a better spell. Is that true? And if so, what is the AI judging based on? I know the AI won't use gems if it considers the enemy too weak to justify them, but (a) how does the AI measure "weak" (I think it's just the size of the enemy army with no assessment of its strength), and ( are there other reasons the AI might not cast this?

Unless the enemy is considered too weak, which is measured by som rather stupid formula, he should cast it.

The script is not abandoned because "there is a better spell", but because either it cannot cast it for some reason, finds no valid target, or the enemy is too weak.

To quote my exact response to this question on Shrapnel:

He should cast it.

The exception would be if the AI decides that your opponent is too weak to bother wasting gems on - I don't know what metrics are used to arrive at that decision.


Casting fatigue is referenced in your manual at the bottom of p.86, under the subheading 'Fatigue,' and using magic gems in casting is referenced on p.89, under the subheading 'Using magic gems in combat.'

As referenced:

Fatigue
Just like attacking in combat, casting spells in battle costs fatigue. Each spell has a listed fatigue cost which a caster incurs when casting that spell. For each skill level in the required path that the mage exceeds the minimum, he incurs 1 / (1 + (mage skill – minimum skill)) of the listed spell fatigue. In other words, having an extra skill level means the mage suffers only ½ fatigue, two extra skill levels means he suffers only ⅓ as much, three extra is ¼, and so forth. Spell casters also incur fatigue equal to their Encumbrance value for each spell cast. This is not subject to reduction by skill bonuses. It is harder to make magical gestures in heavy
armor. Everyone knows this from armor class rules.

Example: Bageroth casts a Fireball on the field of battle. The spell requires Fire-2 and has a fatigue cost of 20. Bageroth has a Fire skill of 3, which is one greater than the requirement, so he incurs only half as much fatigue from the spell. However, he has an encumbrance of 5, so the final fatigue cost for casting the spell is 15.

also:

Mages can use magic gems in combat for two purposes: to temporarily raise their skill level, and to reduce fatigue. This represents the release of the distilled magical power of the gems themselves. Because it takes a skilled caster to keep things from getting out of control, a mage can only use as many gems as his current skill level in that path. By using a magic gem, a caster gains one skill level in that magic path. This can be used both to allow a mage to cast a spell for which he would not normally have the skill, or to reduce the fatigue of a spell, or both. However, a mage may never increase his skill level by more than one by using gems, or gain skill in a path in which he or she did not already have at least one skill.

Example: Bageroth is a Fire-3, Astral-2 mage. He wants to cast Astral Fires in battle. This spell requires Astral-3 (the primary path) and Fire-1. Bageroth could use one Astral gem to increase his Astral skill to 3, which would make him able to cast the spell. He could use two Astral gems, which would not only allow him to cast the spell, but would reduce the spell fatigue by half. He could not use three or more gems because his original skill in Astral magic is two.

and importantly:

The player may script spells for which a mage does not have the requisite skill and place the necessary gems in his inventory. He may also simply include extra gems in the mage’s inventory and the computer will use them to reduce fatigue for spells for which he already has the skill, if necessary. However, the specific use of gems in this manner is controlled by the computer. There is no opportunity for the player to intervene.


Really, the best way to be sure that something is going to work the way you expect it to in an important fight is to test it yourself. Get a debug mod (something like http://forum.shrapnelgames.com/showpost.ph...966&postcount=1 ), create a new game with both players set to 'human,' and use this sandbox to mimic your real-game conditions. See what happens!

Thanks again Kungfoo - I will also try the debug mod. I am just trying to be sure that those are the only reasons the spell doesn't cast.

For (say) a summoning spell or BF enchant, there shouldn't be a "no valid target" problem. And I can control avoiding the "mage doesn't have the path or gems needed" problem.

So I am just trying to be sure the only issue is the amorphous "enemy is too weak to justify spending gems" problem. And it sounds the only way to get a sense of when that kicks in is testing/experience.

I know I've experienced a number of fights where I had a large army up against a single SC, and where the AI did not cast a gem-costing spell I had scripted. My guess is that the AI decided, "you have tons of troops, and the enemy only has 1, so this spell is a waste." Which of course can be disastrous. I am not sure there's any way to avoid that.

My personal guess is that the determinant for the gemuse trigger is the enemy hitpoint value. This would explain common behavior towards thugs, and it is also the method the spell AI uses to pick targets in battle.

You could rough out the formula with Ghost Rider Testing. From experience it's somewhere around 3 ghost riders on top of a big army to cause gem wastage. But, who knows if the existing army factors into it. Fun thing to test. But someone else can figure it out. I'll probably forget about it tomorrow.

I agree with this. I believe from testing it only cares about starting enemy army hp, but I didn't find out exactly how many hp total it needs to see to flip the switch.

I agree with this. I believe from testing it only cares about starting enemy army hp, but I didn't find out exactly how many hp total it needs to see to flip the switch.

I'm not sure about this. Two GRs are enough to trigger spells. But I seem to remember battles where three were needed to trigger gem usage.

Of course I could be blowing this out of my ars as my memory tends to play tricks on me.

That's what I remember two E. 2 usually works but sometimes doesn't.

I'd say it's definately not just the HP value that decides whether gems are used or not. As magic items and mage paths are also factored in. Plus from experience and lite testing I think the difference in HP is the bigger factor rather than just the total HP of the weaker side. (for example, as seen with it sometimes taking 2 or 3 GR to trigger gems use). But HP alone doesn't determine it, I'm fairly certain of that. I tried to figure it out some time ago but abandoned it due to lack of time and a growing lack of interest. Just did a few quickies this morning though to highlight some points on this topic.

The gem use trigger can be seen in the log file, and is covered at the start of battle with some short calcoverkill figures. Here's the test data I gathered this morning.


Tests were an Indy commander(s) with 5stars* vs 40 LA Aby PD and a F2 mage scripted to cast Willy Wisp.

Test 1 - When just 1 Indy Commander attacked (bold text is directly from the log. Other text is indicating what I think the log entries mean)

calcoverkill player 58 --- 58 is the nation id. In this case LA Abysia
calcoverkill 2091 vs 37 --- 'power' level calculations for defender vs attacker
Overkill for player 58 (2091 vs 37) --- This seems to indicate that the dominant naton in this battle may NOT use gems. ie. it has declared an Overkill for player 58
calcoverkill player 42 --- 42 is the nation id. In this case MA Jotunheim
calcoverkill 37 vs 2091 --- 'power' level calculations for attacker vs defender

Brief explanation - The 58+42 player values indicate the nations involved, so in this case LA Aby and MA Jotun respectively. The 2091 is the total 'power' of the defender (LA Aby, 40 PD+F2 mage) and 37 the total 'power' of the attacker (the lone 5star Indy commander). Not exactly sure what this 'power' value (if that' s even the right name for it. But it'll do for now) consists of, but it changes when items and magic paths are added to the commander. And wouldn't be surprised if every stat of every unit on the battlefield is factored in (tests would show it, but someone else can do those if they care). No gems were harmed in the running of this test.


Test 2 - Same Indy commander as Test 1, but this time loaded with SC gear.

calcoverkill player 58
calcoverkill player 42
calcoverkill 47 vs 2091

Brief - Here two lines of log calculations have gone, specifically the ones relating to player 58 having an Overkill declared, meaning gems can and were used in this battle. Note that the 'power' value for player 42 is only 47, because this becomes relevant in the next test. Gems WERE harmed in the running of this test.


Test 3 - The same Indy commnader as previous, and his twin brother. No items.

calcoverkill player 58
calcoverkill 2091 vs 74
Overkill for player 58 (2091 vs 74)
calcoverkill player 42
calcoverkill 74 vs 2091

Brief - Here the two lines of log calculations are back, and as in Test 1 an Overkill is declared for player 58. But note that the total 'power' level for player 42 is 74, which is above that seen in Test 2. Yet there was an Overkill declared in this Test with a higher value, and no Overkill declared in Test 2 with a lower one. This seems to indicate that items have a higher weighting when deciding gem use, but also a weighting that is not 100% clearly visible in the 'power' calculation/figures. ie. The SC items added a 'power' value of 10, and that 10 was enough to trigger gem use. Whereas the 'power' value of 37 added from the duplicate commander was not enough to trigger gem use. Cheers for the transparency Johan! No gems were harmed in the running of this test.


Test 4 - Same Indy commander, but this time empowered up to W1

calcoverkill player 58
calcoverkill 2091 vs 39
Overkill for player 58 (2091 vs 39)
calcoverkill player 42
calcoverkill 39 vs 2091

Brief - Overkill declared, so no gem use allowed, despite a mage being the attacker. And the empowerment seems to have added '2' to the value of the attackers 'power' level, so mage paths do seem to be factored in when deciding gem use. No gems were harmed in the running of this test.


If anyone wants to do the huge donkey work to figure out something meaningful from the Overkill calcs and some sort of semi-precise guideline as to when gems use is triggered or not, then please be my guess. No way I'm doing it though. But the above does seem to indicate that it's not just "HP of the weaker side" that triggers gem use, which is what a lot of players currently seem to think, and has been suggested on this thread. (and why I posted. ie. to clear this issue up a bit)


* It's a very old and long running multi purpose test game I use. Hence the commanders have 5 exp stars. Could have bought new ones but I'm lazy.

That's three people that have noticed the rough 2-3 GR casts to burn gems on a good sized army. So that puts a sort of basic idea into it. I'd say that's for armies around 10-20 magi and 200 troops.

It's a baseline.

For a calcoverkill x vs y (with x being "my army" and y being "their army" from the perspective of the calcoverkill player), there are two things that determine if overkill is triggered. Here's the basic one:

Easy, right? But as you have already said, it isn't so simple. However, the next part only deals with enemy commanders. If any enemy commander makes this check, then the overkill calculation is aborted with no overkill. This overrides the previous check.

Unfortunately, it is ridiculous. Like, lots and lots of checks of unit properties and who the fuck knows what else. So I don't think I'm gonna try to go through and figure out exactly what they all mean, sorry. But now you know it's the presence of any one particular commander that causes there to be no overkill, which makes sense. I believe each commander is considered individually, so groups of commanders just under the threshold should still trigger overkill. I'm also pretty sure that it doesn't take your army into consideration during this check. Parts of it involve what appear to be summing things up about the unit, but other parts are more looking for particular masks and such to make an absolute result. So I imagine certain unit properties/equipment will cause there to be no overkill, although possibly in conjunction with some "power level" score being over a threshold. It doesn't use the same function for computing "power level" during the above calcoverkill x vs y check from what I can see, despite the fact that said check is still applied to these commander units for their contribution to the "enemy power level" (y) score. Their standard "power level" used in that y score is not considered during the "is this commander a thug/SC?" check.

Nothing else appears to be considered besides the overall unit power level check and the check for any one enemy commander that is "scary".

Many thanks for the extra info from the code iRFNA. I think it matches what's usually seen in practice and tests, but need to reread it with a fresher head than I have now before I can join the dots properly. Big thank you though, as you are one-of-a-kind right now in the community, as nobody else around can go code hunting like this AFAIK.