Gem Usage in Battle

[Bananadine]

Originally Posted by Chazar

If I had only known those a single turn earlier! I guess that either knowing about the corner-self bless and/or the blood slave batteries would have saved my day.

I'm not sure, but it appears that the "corner" part of the corner-self-blessing idea was suggested because a unit by itself in the corner wouldn't be close enough to other sacred units to decide to bless them instead of itself. But if the unit is alone to begin with, there's no reason to put it in any particular spot just for the sake of blessing--it'll very likely hit itself on its first try no matter where you put it. (I think it's still possible for the blessing to miss, in spite of the maximal precision of the spell, but even if I'm right about that, it's still unlikely enough that you can safely ignore the possibility in most cases.)

Originally Posted by thejeff

I would assume the "Is this battle a big enough threat to use gems" calculation, applies to using gems to counter fatigue. I'm not 100% sure, but it looks like that is what happened. 25 PD wasn't enough of a threat, so the fatigue checks didn't even come into play.

My experience suggests that the 1-or-0 gem-use switch that shows up in the debug log is the extent of the influence that the enemy's strength can have on gem use in this situation. Especially with a self-targeting spell like Soul Vortex that works no matter what enemies are present, I wouldn't expect enemy strength to have anything to do with the per-spell gem use decision. But I'm only speculating with little evidence. Can anybody confirm this one way or the other?

[thejeff]

Originally Posted by Bananadine

I'm not sure, but it appears that the "corner" part of the corner-self-blessing idea was suggested because a unit by itself in the corner wouldn't be close enough to other sacred units to decide to bless them instead of itself. But if the unit is alone to begin with, there's no reason to put it in any particular spot just for the sake of blessing--it'll very likely hit itself on its first try no matter where you put it. (I think it's still possible for the blessing to miss, in spite of the maximal precision of the spell, but even if I'm right about that, it's still unlikely enough that you can safely ignore the possibility in most cases.)

I did extensive testing one this. It is not possible for a self bless to miss. (There's some evidence that long range 100 precision spells can miss, especially with low precision.) If there is no other unit within range, the blesser will always hit himself. Any other unit within range, even enemys or non-sacred units, makes it possible. The spell always hits it's target, but to choose that target the AI simulates casting it on the available targets, then picks the first target that got the best result. Since the AoE spread is random, the actual cast won't hit the same squares and thus maybe not the caster.
That's probably clear as mud, but I'm not sure how to say it better.

My experience suggests that the 1-or-0 gem-use switch that shows up in the debug log is the extent of the influence that the enemy's strength can have on gem use in this situation. Especially with a self-targeting spell like Soul Vortex that works no matter what enemies are present, I wouldn't expect enemy strength to have anything to do with the per-spell gem use decision. But I'm only speculating with little evidence. Can anybody confirm this one way or the other?

I think we agree. For some reason I was thinking that gems to avoid fatigue might ignore the gem-use switch. They don't of course.

[Bananadine]

Originally Posted by thejeff

I did extensive testing one this. It is not possible for a self bless to miss. (There's some evidence that long range 100 precision spells can miss, especially with low precision.) If there is no other unit within range, the blesser will always hit himself. Any other unit within range, even enemys or non-sacred units, makes it possible. The spell always hits it's target, but to choose that target the AI simulates casting it on the available targets, then picks the first target that got the best result. Since the AoE spread is random, the actual cast won't hit the same squares and thus maybe not the caster.
That's probably clear as mud, but I'm not sure how to say it better.

I can guess at an exact algorithm based on that, and it isn't what I'd vaguely imagined the actual algorithm probably was... interesting!

Anyway about the missing: I'm like 2/3 sure I've seen a blessing miss, but if you're right then it must have been because there were multiple units present... but surely not many, in fact maybe just two, because if there were many then I'd have expected the spell to miss some, so I wouldn't have been surprised when it did, and I wouldn't remember the incident now! But that's quite possible.

I thought I saw somebody say in some thread, though, that a 100 precision, single-tile spell could still straight-up miss. Like, somebody had seen a mind duel of all things miss, or something like that. I dunno. Of course magic duels tend to happen at much longer ranges than blessings.

[Samhain]

I'm going to make an educated guess here and say that your Oracle was burning pearls to boost his chances of victory in the duels rather than to reduce what little fatigue it would cost it. I haven't heard of this special case for Magic Duel before, and it is not listed in the Wiki page I linked to above. But, I've got to believe that is what was going on.

There is good nugget of advice in this little epiphany, if you are sending a mage in to battle with the intent of spamming Magic Duel, dump some pearls on him if you want to increase his probability of survival.

[Deathblob]

Magic Duel is based on the unboosted S levels of the mages.

[LDiCesare]

Yes, there would have been no point boosting the S level even if it worked: I was S9 and facing S2 adonim so had 100% chance of success. But to make sure I could retrat in time, I had tested the magic duel-returning combo in SP several times until I found out how many pearls the oracle actually used in battle.
I think that's the best thing to do: Setup a test game, check how many gems your mage will use to cast the spells you want to use. (Of course, magic scale affects fatigue casting, so the test may need to take thhat into account, but I haven't found the scale really relevant with reagrds to gem use).

[Samhain]

There would be no point if your mage knew the level of Astral of his opponents. But, the coders may not have written the AI code with the premise that your mage would know that. This could still be true, and consistent with some coherent concept of how battle magic in Dom3 works, even if the AI is written so that he targets the highest level opponent Astral first.

On the other hand, if Deathblob is right and that gem boosters don't help, then this could actually be a bug. I just can't think of any reason other than increasing the likelihood of success, even when the exact probability is not known, for your A9 Oracle to use Pearls for Magic Duel.

It is a good case for testing. I would be particularly interested if you set up a test with A9 astral mages on both sides of the duel but only one having the cache of Pearls. Run the test a few times and we should have a pretty good idea if the Pearls affect who wins.

[JonBrave]

May I ask (politely), if you care so much about discovering whether/when/how/etc gems are cast, why don't you run the turn with -dd debug ? Does that not tell you what you want to discover?

[Deathblob]

Because the question (or rather the answer) we all want to know is "OMGWTFBBQ WHY!?!?" And sometimes "Why not?".

Unfortunately neither of which is in the debug log.

[LDiCesare]

For magic duels, pearls and items don't affect who wins. Only the base, raw level of the caster.