Inner Peace (Through Superior Firepower) – Episode 030

[BLANDCorporatio]

So, anyway, if you offered me those (essentially) 2:3 odds, I'd totally take the short side of those odds, and in the bizarre case of somehow being able to repeat this bet on independent otherwise identical fictional narratives (ha), I think I would tend to win. 40% just isn't your best estimate.

It's a completely random key mash. So I wouldn't expect it to be the best at anything. I would like to know why you think it's not the best. Is it because it likely can't be* since it's random and thoughtless, or because, knowing other fantasy narratives, you intuitively keep some statistics on them?

EDIT *: damn you infinite regress!

Spoiler: show

assuming a "narrow" region for a "best choice", and uniform distribution of possible values, gives that a random choice from them isn't likely to be there. This works wonderfully if there's a finite amount of possibilities to chose from, it's more cumbersome with continuous variables but the basics are the same.

[MarbitChow]

He asked for numbers. What am I supposed to do?

It just seems to me that you're attempting to build in a significant number of additional variables into a prediction algorithm that we currently can't even conclusively eliminate pure random distribution from.
At best, we can say that "pure random selection from the group of non-popped caster types" appears to match the available data 'better' than "pure random selection from the group of all caster types". Attempting to add any additional parameters is pure conjecture, akin to asking "how do angels influence gravity?"

But I'm still going to continue to believe that it's actually Fate that determines when a caster pops, and what type of caster it is.

[BLANDCorporatio]

But I'm still going to continue to believe that it's actually Fate that determines when a caster pops, and what type of caster it is.

So far I've not even ventured into examining "when a caster pops". All analysis was, "assuming a caster was popped, what about the type?"

[0beron]

I do want to point out a possibly overlooked tidbit....we've been saying it would be hard for FAQ to have naturally popped so many casters without getting ten times as many warlords in the process...what's to say they don't?
We don't really know much about "court" but what's to say it's not a bunch of useless warlords who "drank the koolaid" and don't want to fight like Jillian?

[BLANDCorporatio]

Or heck, drank the kool-aid and went on to meet their makers?

Sorry. I really like that film, gigantic warts and all.

Besides, I always wonder what rations are made of. Maybe soylent green exists in Erfworld.

[Balerion]

To Marbit and Steve:

I agree entirely. There are all sorts of factors that reasonably could skew our running of the trial. GK probably is a bad example, and we don't actually know all Transylvito's casters, so they get tossed too. But even with just Unaroyal and Jetstone as addons, the % is going to dip under 10% (as each of them was about a 50% chance of no duplicates, with Unaroyal being a slightly better chance). That was part of my point about massive varriance on our sample; we likely have error bars of a good +-30% at least.

And people could be trading casters away, self-selecting into the no duplicate model. However, that is an event we haven't seen, nor have we ever heard of the event of a duplicate caster. So with the information we currently have, we have to make an assumption for which way this plays out. We don't have to hold to it very strongly but we should have it, and also know the new evidence that could dash it to pieces. And by the current information, pure random is a bad model. With a bit of extra info, it can become an acceptable model again.

[MarbitChow]

And by the current information, pure random is a bad model. With a bit of extra info, it can become an acceptable model again.

Based on the given data, "pure random w/ no dupes" seems to be the simplest model that matches the given data. I object to calling pure random a 'bad model' based on 2 samples, but since I don't believe it's even actually random, I'll refrain from further argument.

[0beron]

In the words of Einstein, "Make everything simple as possible, but not simpler". So in that spirit I'll have to agree with Marbit, Random w/ no Duplicates is the simplest model that still fits the (albiet limited) data.
As for OPINION, I think it's Random with non-equal probabilities for all caster types, no dupes....but the Titans "mess with the dice" every so often. But there's not enough data to get into something that complex, and I doubt there ever WILL be. This is probably one of those things we will never "figure out", it will either be told to us, or left a mystery.

[wrecan]

Maybe soylent green exists in Erfworld.

Soylent green is units!

[Balerion]

We know that soylent green is units; what do you think harvesting dwagons means happens to them

And yeah, calling pure random a bad model was probably a rhetorical overreach past what the numbers show; I will instead revise the opinion to "not a good model"

[MarbitChow]

We know that soylent green is units; what do you think harvesting dwagons means happens to them

Soylent Pink tastes like bubble gum!

[mortissimus]

Barbarian casters end up in the Magic Kingdom.

Casters that become barbarians modtly end up in the magic kingdom as they escape by portal. Of course they can also be out of the castle and being heir or in the personal stack of an heir. We have no indication that casters can pop randomly (I think) and we do have an indication that no casters pop in the Magic Kingdom as a text update talked about there being no shame in being a caster in MK unless you are a predictomancer (indicating that all predictomancers in MK have failed their sides before going to MK).

[LTDave]

I'm pretty sure I have no idea what you guys are talking about.

It comes down to this: FAQ doesn't have a think-a-mancer because they don't need one. All* their stacks and cities are in one little bubble within a move of each other. They don't need to send messages to far flung units or governors, which is what a think-a-mancer is for.
Even if they did pop a Think-a-mancer, they could quite easily sell them off or disband them or whatever.

*except for Jillian's mercenary stack, but Jillian has a magic message hat or equivalent.

[Kreistor]

Before anyone starts drawing any conclusions, you're overlooking one other assumption. You are all assuming that every caster type is equally powerful.

Balance is extremely difficult when considering disparate abilities. The folks at Blizzard had to drop kick their first attempt at Starcraft into the can, because they were trying to balance three completely different resource management styles, and couldn't do it. In the end, they went for the single model with different pricing you see in the modern game, delaying release for over a year.

Sides with Thinkamancers may simply have a higher probability of survival. If true, then even with a stochastic distribution, there would be more Sides with Thinkamancers than without.

Consider the Jetstone Dollamancer. What does he do? Creates units. Guess what? They can already create units through production. He doesn't add a new ability, only new units.

The Thinkamancer does something Units cannot -- permits direct diplomacy with the enemy, direct orders to expeditionary forces, etc. No known units in the Jetstone inventory do that. It's new functionality, and while not directly militarily powerful, indirectly it can be game changing. Even a Level 1 Thinkamancer might be able to end a War and save the Side with one two paragraph conversation. Could a Battle Bear do that?

Banhammer doesn't need a Thinkamancer because:
1) He doesn't want anyone knowing about his Side, so he doesn't want to talk to anyone
2) His Chief Warlord is with the expeditionary force and doesn't need better communications than a Hat
3) Charlie is always there to provide his Thinkamancy services to the Sides his expeditionary forces work with.

[0beron]

Well on the one hand you could argue that the ability of a thinkamancer to Link does actually offer them something. Linking the Math- and Money-mancers could make them even more efficient at maintaining the side on a small treasury, Linking Jack and Marie might make them more powerful, perhaps even the Lookamancer and Marie would be able to see things while linked that neither could otherwise witness.

On the other hand, the principle of "need" could be irrelevant. FAQ certainly doesn't need a Shock or Healomancer, but they have them anyway. A Shockamancer doesn't do much that a handful of archer stacks couldn't. And a Healomancer doesn't do anything that resting doesn't already do.

So while I'll agree that perhaps sides with Thinkamancers have better chances of survival, thus "artificially" making Thinkamancers themselves seem more common, the idea that FAQ doesn't have one because they don't need one doesn't hold much weight. My initial thought of "OH, they don't have a thinkamancer, this must mean something!" has changed to "Oh...they don't have a thinkamancer....or a caster of more than a dozen other disciplines either...oh well."

[drachefly]

So, anyway, if you offered me those (essentially) 2:3 odds, I'd totally take the short side of those odds, and in the bizarre case of somehow being able to repeat this bet on independent otherwise identical fictional narratives (ha), I think I would tend to win. 40% just isn't your best estimate.

It's a completely random key mash. So I wouldn't expect it to be the best at anything. I would like to know why you think it's not the best. Is it because it likely can't be* since it's random and thoughtless, or because, knowing other fantasy narratives, you intuitively keep some statistics on them?

Within this fantasy narrative in particular, I've developed a feel for the nature of its laws. Much as in real life, I would be very surprised to find a nonlocal force - even if it only applied to fundamental particles we had not previously created. The kinds of connections that are prone to exist.

But looking back... "It's a completely random key mash. So I wouldn't expect it to be the best at anything." Do you really think that a random key mash is the best way to generate estimates of prior probability in the absence of direct data?

How unlikely would P(data|u.i.d) have to be for you to reject it? Except by comparison to other things, there's no way to gauge that - after all, every time you add a datum, the probability drops. It's only relative probabilities that can actually make you change your mind. With something like 'randomly among those not already on the side'... what probability ratio would you have to see to put you in serious doubt?

That value is the value you implicitly are already using as your prior. You're just not admitting it.

[BLANDCorporatio]

Do you really think that a random key mash is the best way to generate estimates of prior probability in the absence of direct data?

Of course not. To claim that method X (say, random key mash) is the best would be a positive claim and may need backing. The whole point of the questions so far was to establish what the best method is. It looks to me like it either does not exist, or is sufficiently ill specified to allow great variation in priors and therefore, on the limited set we have, different conclusions.

How unlikely would P(data|u.i.d) have to be for you to reject it? {...} With something like 'randomly among those not already on the side'... what probability ratio would you have to see to put you in serious doubt?

That value is the value you implicitly are already using as your prior. You're just not admitting it.

One can work back from some arbitrary cut-off threshold like 0.05 chance of the evidence matching the hypothesis, yeah. Of course this just shifts the problem around aimlessly. Either justifying a cut-off or a certain prior belief, same thing - how? Or is it that this selection is fairly free, bound (or not) by other experience and subjective, complex and infinitely debatable?

[Kreistor]

And, of course, there's always the possibility that the author gnerated the Casters specifically to give us the widest possible understanding of magic. Repeating caster types does not introduce new info, so to get as many descriptions out as quickly as possible, he selects ones that have not been selected yet.

[0beron]

And, of course, there's always the possibility that the author gnerated the Casters specifically to give us the widest possible understanding of magic. Repeating caster types does not introduce new info, so to get as many descriptions out as quickly as possible, he selects ones that have not been selected yet.

Yeah that's definitely very likely, Rob has almost certainly chosen every single caster for important reasons (plot, joke reference potential, and showing us different aspects of magic) but at the same time he might have an idea of what governs caster popping under "normal" circumstances. But given the first part of my comment, it's unlikely we'll be able to guess those rules based on sample evidence from the comic.....

[Kreistor]

Yep, 0beron. But, in the end, as a player, I would probably want as much variety in my Casters as possible, so presuming I played Banhammer, I would Turn any Casters captured by the Expeditionary force, but far from FAQ so they have no knowledge of it, and trade with other Sides via the EF to get Casters of unique types.