Global Enchantment Casting Strategy?

[JonBrave]

Every time I think I'm getting confident at this (SP, vanilla), I find out there's yet another thing I don't know about

OK, so some AI guy has cast "The Eyes of God" ages ago. I'd like to cast it now (there are 4 globals up, 1 slot free, though I realise that isn't relevant here). So I generously add an extra 10 gems for 60, and confidently cast. Next turn, it failed. I'm 60 astrals lighter, and fed up.

I was expecting to win... am I just a wild optimist?

How the heck should I guess what he'll have spent? What are we using for "optimal" game theory strategy here?

I could have gone "Dispel" first, then cast my own, would that have been a better strategy, seemed to me I'd have spent more overall?

To conclude, I just started looking at docs. It does seem I had not understood the casters involved were relevant, can anyone summarise the math to say how much difference that makes?

[Squirrelloid]

Ok, first of all, there's absolutely no way to predict what the AI will have spent. Because the AI does not behave rationally.

Second, in this particular case, the appropriate strategy would have been to cast Fate of Oedipus, which not only automatically cancels Eyes of God, but also blinds the caster. Then you could cast your own =P.

Third, it is generally a better idea to overcast than to dispel and then cast. Well, its actually a better idea to kill their caster, but barring that...

Fourth, how overcasting and dispelling is calculated.
(1) *Naturally* having more of the path required to cast it than necessary gains you a benefit of +5 per extra path length to the check. So if an N7 caster was overcasting Gift of Health they'd get a +10 to teh check. Boosters do not count for this (but empowering does).
(2) Every excess gem you spend adds +1 to the check. (Note: if you have a site which decreases the cost of the appropriate path for the global, you will only benefit from gems spent in excess of its *normal* cost, not the cost with the site factored in. Ie, the excess you spend to make up the difference between the reduced cost and its normal cost is lost with no effect.)
(3) Both sides roll a drn and add their bonuses. Higher total wins, tie to the defender.

[JonBrave]

Don't have "Fate of Oedipus", looked it up and I don't really have any Fire at all. And I understand that killing the opposing caster would be best, but not an option.

So.... Given that we have "no idea" what the AI has spent, just what does anyone do in SP? I realise this may be a matter of the deepest disinterest for MPs, but any advice on what I should do now?

[Squirrelloid]

Um, I don't really play SP long enough for anyone to cast globals. My SP game durations are 10-20 turns long, and deal with testing expansion strategies.

[JonBrave]

Originally Posted by Squirrelloid

Um, I don't really play SP long enough for anyone to cast globals. My SP game durations are 10-20 turns long, and deal with testing expansion strategies.

I kinda realised this .
Try "Dawn of Dominions", all AIs Impossible, and let me know if you've won by turn #20

[Squirrelloid]

If its against the AI, i've won since pretender design. =P

Its like playing Tic-Tac-Toe against someone over the age of 6. Its a foregone conclusion.

[rdonj]

You can almost always beat the AI by casting a global at +20 gems and keep the global forever. Sometimes the AI will try and fight you for the spell though and you'll have to go a bit higher. Maybe +40. In 99% of cases though +20 should be enough.

[JonBrave]

Originally Posted by Squirrelloid

Its like playing Tic-Tac-Toe against someone over the age of 6. Its a foregone conclusion.

You obviously don't know the same 6 year olds as I've come across...

[don_Pablo]

Another question about globals.

Gem-producing GE's are the most popular for the whole course of game.
What is the reasonable limit of the cost for 20gem-producing enchantment?
(I mean late-game when you have many gems)

[Knai]

Originally Posted by JonBrave

You obviously don't know the same 6 year olds as I've come across...

Tic-tac-toe is solved. The foregone conclusion is a tie.