10 Magic Strength Pretenders....

[Thufir]

archaeolept said:
currently the chance isn't even all that small - I think zen mentioned something like 17% w/ a 5 astral difference...

ie. that's huge.

That is huge. I think I'll put a post on the wishlist thread (I'm sure Gandalf/Endoperez will love me for it ) I don't suppose anyone has ever posted a formula for how magic duel works?

I like the role that Astral plays in the game mechanics - I think of it as the one path that allows you to gain magic superiority. As such, it would seem that the game would be richer and more fun if it were a viable strategy to go with a combat mage pretender, that included strong astral. By combat mage pretender, I don't necessarily mean a human caster as pretender (although making human pretenders more viable would be good as well, but that's a separate issue). Instead I mean that taking strong astral in a pretender shouldn't imply that it's necessary to hide your pretender from all exposure to combat, even (especially?) late game combat.

So, to that end what odds would be appropriate in order to make it viable to take astral on a pretender? With a 5 astral difference I'd guess that a 1% chance of defeat is a very low risk level. If you go with a s9 pretender, appropriately equipped, I'd think it would be expensive enough to get within 5 levels, that a 1% chance of success is far from worth the expenditure. So, probably 1% is overkill, yet the current 17% is underkill.

Whatever the right number is, it seems clear (at least to me) that tightening the standard deviation in the magic duel formula would improve the game. I'd bet that in current MP game play, it's mostly cheap mages boosted by communion that cast magic duel, and that certainly seems wrong.

I'd like to make a concrete proposal, but it's difficult to do that without knowing the formula for resolving magic duels. I imagine one could could set the formula in such a way that supports the most interesting game play by comparing the expected value of various astral hit squads, against mages with various astral levels. In other words, if you know how many hit squads it takes in order to reach a total of 90% chance that one of the squads succeeds, and you can calculate this for squads sent against a 6s mage, a 7s mage, etc, then you can look at a total gold cost for taking out a given level astral mage. At that point you could make a well informed decision on how to set the magic duel formula in order to make a given level of astral a viable choice for a pretender. I'm not sure exactly what that level should be, but I'm inclined to think that 6s+ should be a viable choice.

I certainly don't have enough game knowledge to do the above analysis (even if I knew the magic duel formula), but at least theoretically that would be a nice way to set the standard deviation correctly. Probably it's reasonable to tweak the formula without doing the calculation above, but any way you cut it, I'd like to see that tweak happen.

[Graeme Dice]

Thufir said:
I don't suppose anyone has ever posted a formula for how magic duel works?

Many times. 1d6 + skill vs. 1d6 + skill

The highest wins, the lower dies. On a tie, both die.

The probabilities are found in this thread:
Magic Duel and other dice roll charts

[Graeme Dice]

Graeme Dice said:

Many times. 1d6 + skill vs. 1d6 + skill

The highest wins, the lower dies. On a tie, both die.

The probabilities are found in this thread:
Magic Duel and other dice roll charts

If you are 5 points higher, then you will only die 2.7% of the time. That's about the same as having a MR of 22.

[Thufir]

Graeme Dice said:
If you are 5 points higher, then you will only die 2.7% of the time. That's about the same as having a MR of 22.

Excellent info and links, Graeme, thanks a bunch! And my apologies for not searching before asking

I find this encouraging, at a 2.7% chance with a 5 pt difference, it sounds to me like perhaps the astral combat mage pretender is a viable strategy. I still don't know enough about how magic duels unfold in MP game play to make an informed judgement, but it's quite helpful to know the formula.

[archaeolept]

i'm only going w/ the numbers zen mentioned. perhaps Magic Duel as currently implemented isn't working quite right. the numbers given by Saber Cherry were for non open-ended die rolls, for instance...

Frankly, I would suggest keeping your 500 pt Ghost King w/ 3 astral somewhere a long way from anyone else.

[Thufir]

archaeolept said:
i'm only going w/ the numbers zen mentioned. perhaps Magic Duel as currently implemented isn't working quite right. the numbers given by Saber Cherry were for non open-ended die rolls, for instance...

I haven't done the math (yet), but being open ended does not dramatically change the odds. It's certainly possible that this is not implemented the way the post indicates, but it seems unlikely that this is bugged (perhaps the implementation has been changed, or misinformation was propagated along the way). It's easy to test though, and I'll do that later today.

Frankly, I would suggest keeping your 500 pt Ghost King w/ 3 astral somewhere a long way from anyone else.

From a game design point of view, I think it's eminently reasonable that a 3s Ghost King is vulnerable because of having chosen astral. The thing I've been questioning is whether a 6s+ pretender is unduely vulnerable to astral hit squads.

[Thufir]

So, I've done some testing on magic duel now, and the results were a little bit confusing. I did 2 sets of 20 tests, one with S3 vs S8 and the second set with S7 vs S8

Here's what I've got:

S3 wins -- S8 wins

  0          20



S7 wins -- S8 wins -- Tie (no deaths) -- Tie (both die)

  2          16            1                   1

If my testing method is truly random (which I am begining to doubt), then the second set of results is surprising, in that S7 should win roughly 1/3rd of the time according to the formula for magic duel.

The one thing I'm uncertain of in doing my testing, is whether or not I am truly getting a random sample by reloading from saved files. Can anyone (ideally a dev ) tell me how Illwinter is setting the seed for their random number generator? I'd like to find a testing method where I'm sure that I'm getting a true random sample before doing more tests.

[archaeolept]

lol.

i especially like the "Tie (no deaths)" since, as presently understood, that shouldn't be possible

i also really suspect the random number generator is quite quirky anyways

but good work thufir. perhaps try a different situation in a different game...

also, if you're using 2.13, it could be that the devs have altered the method...

[Nagot Gick Fel]

Thufir said:
The one thing I'm uncertain of in doing my testing, is whether or not I am truly getting a random sample by reloading from saved files. Can anyone (ideally a dev ) tell me how Illwinter is setting the seed for their random number generator? I'd like to find a testing method where I'm sure that I'm getting a true random sample before doing more tests.

Don't bother with testing Magic Duel in 2.13. The formula is slightly bugged and you may not get accurate results.